# Account overview (/docs/account)
Your account area lives at **Account** in the app. To get there, open the account menu at the bottom of the sidebar (your name) and choose **Settings** or **Plan & Billing**, then move between sections with the account navigation. It has five sections:
| Section | What's there |
| -------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------ |
| [Profile](/docs/account/profile) | Your name, username, bio, and account activity |
| Community | Your public creator profile and the posts you've published — see [Community](/docs/collaborate/community#your-creator-profile) |
| [Notifications](/docs/account/notifications) | Your full in-app inbox |
| [Settings](/docs/account/security) | Active sessions — see where you're logged in and revoke devices |
| Billing | Credits, subscription, invoices — covered in [Plans & billing](/docs/plans-billing) |
Workspace settings — members, roles, tool access — are covered in [Collaborate](/docs/collaborate/workspaces).
# Notifications (/docs/account/notifications)
## The notification bell [#the-notification-bell]
The bell in the top bar shows your unread count and has two tabs:
* **Inbox** — your personal notifications. Click one to jump to its subject; mark items read, delete them, or **Mark all read**. Story invitations show **Accept / Decline** right in the list.
* **What's new** — product announcements, dismissible, sometimes with a link to try the feature.
**View all notifications** opens the full inbox at **Account → Notifications**, with an **Unread Only** filter. Read notifications are cleaned up automatically after 90 days.
## What you're notified about [#what-youre-notified-about]
| Event | In-app | Email |
| --------------------------------------------------------------------------------------------------------------------------------- | ------------------ | --------- |
| Someone shares a story with you or invites you to a workspace | Yes | Yes |
| Someone comments on your Auteur project | Yes | — |
| Your credit balance is low: under 150 credits, including zero (once each time you run low, not repeated until new credits arrive) | Yes | Yes |
| Chapa subscribers: your subscription renews in about 3 days (includes a renewal payment link) | Yes | Yes |
| Chapa subscribers: your subscription has expired | Yes | Yes |
| Product announcements | **What's new** tab | Sometimes |
Account emails arrive when you need them:
* the verification link after you sign up
* password reset links, and a confirmation after your password changes
* a notice when an administrator activates your account
Ehudai doesn't currently send notifications when a generation finishes or fails, or for payments and plan changes. Check the queue panel and the tool's history for generations, and [Billing](/docs/plans-billing) for payments.
Emails come from `noreply@ehudai.com` — add it to your contacts if messages land in spam. Chapa subscribers: your monthly **renewal payment link** arrives by email about 3 days before your period ends, so it's worth watching for.
# Profile (/docs/account/profile)
Open **Account → Profile** to manage your account details. Click **Edit Profile**, make your changes, then **Save Changes**.
## What you can edit [#what-you-can-edit]
* **First and last name**
* **Username** — generated for you at signup; change it here
* **Bio** — up to 500 characters
* **Location** and **website**
### Username rules [#username-rules]
* 3–30 characters
* Letters, numbers, and underscores only
* Can't start with a number
* Must not already be taken, and some names are reserved
## What you can't edit [#what-you-cant-edit]
* **Email** — your email address can't be changed once the account exists.
* **Account deletion** — there's no self-service deletion yet. To have an account removed, open the account menu at the bottom of the sidebar and choose **Help & Support**.
Your public name and bio in [Community](/docs/collaborate/community) are set separately, under **Account → Community**.
## Account activity [#account-activity]
The profile page also shows read-only details: when you joined, your account status, whether your email is verified, credits earned and spent, and stories created.
**Email verified** shows **Yes**, **Not verified**, or **No email on file**.
If **Email verified** shows **Not verified**, your 1,000 free signup credits haven't been granted yet. Click **Resend email** under it to get a new verification link, then open it from your inbox. After a resend, the button waits a minute before you can send another. See [Create your account](/docs/getting-started/create-account).
# Security & sessions (/docs/account/security)
## Active sessions [#active-sessions]
**Account → Settings** shows **Active Sessions**: every device where your account is logged in — browser, operating system, IP address, and when it was last active, with a **"This device"** badge on your current session.
* **Revoke** any individual session to log that device out.
* **Revoke All Others** logs out everything except the device you're on.
* Your current session can't be revoked — use **Log out** in the account menu instead.
If you see a session you don't recognize, revoke it and reset your password.
## Passwords [#passwords]
* **Forgot your password?** Use **Forgot password?** on the login page and enter the email address on your account. The reset link is valid for **1 hour** and works once. If your account has no email on file, contact support to reset it.
* A successful reset **signs you out everywhere** — all your active sessions end, and every device has to log in again with the new password. You'll also get a confirmation email with the time and IP address of the change.
* Passwords need at least 8 characters with an uppercase letter, a lowercase letter, and a number. The sign-up and reset password pages list these rules under the password field and tick each one off as you type.
* **Changing your password:** there's no in-app password change. To change it, log out and use **Forgot password?**.
* **Google sign-in accounts** have no password you know. To add one, run the same **Forgot password?** flow with your account email.
## Two-factor authentication [#two-factor-authentication]
Ehudai doesn't offer two-factor authentication (2FA) yet. Use a strong, unique password, or sign in with Google and protect your Google account with 2FA.
## Sign-in methods [#sign-in-methods]
You can sign in with your email address or username and your password, or with Google. If your Google account uses the same email as your Ehudai account, the two are linked automatically — either method works from then on.
# Changelog (/docs/changelog)
## September 24, 2026 [#september-24-2026]
### Assistant — a new model lineup [#assistant--a-new-model-lineup]
* **Nine current models in three tiers**: **Fast** (GPT-6 Luna, Gemini 3.5 Flash-Lite, Claude Haiku 4.5), **Balanced** (GPT-6 Sol, Gemini 3.8 Flash, Claude Sonnet 5) and **Premium** (GPT-6 Astra, Claude Opus 5.5, Gemini 3.1 Pro). The model menu groups them by tier.
* Older models have left the menu; chats that used one now continue on its current successor.
* New users start on a Balanced model.
## September 4, 2026 [#september-4-2026]
### Marketing Studio — now available to everyone [#marketing-studio--now-available-to-everyone]
* **Marketing Studio is open to all users**: describe your ad, add a product photo, pick an ad format (and a presenter, if you like), and launch a video ad or product shot from one prompt bar. The credit cost shows before you launch. See [Marketing Studio](/docs/tools/marketing-studio).
### Billing [#billing]
* **One Credit Top-up in both currencies**: 20,000 credits for $20 by card or 5,600 ETB through Chapa. Top-up credits never expire.
* **Top up at any balance**: you no longer need to use up your add-on credits before buying more. New top-up credits are added to whatever you have left.
## August 25, 2026 [#august-25-2026]
### Billing [#billing-1]
* **14-day grace period for Chapa renewals**: if a Chapa renewal isn't paid by your renewal date, you have 14 days to pay the renewal link and reactivate your plan. After that the subscription is cancelled, and you can subscribe again from the Pricing page.
## August 20, 2026 [#august-20-2026]
### AI Studio — a new AI director [#ai-studio--a-new-ai-director]
* **Start with a conversation**: AI Studio now opens on one question, *"What film are we making today?"* Describe your idea, attach a screenplay or reference images, and the agent opens your project and starts building.
* **Manual and Auto**: direct the agent one step at a time, or switch to **Auto** and give it a credit budget to plan and build the whole film on its own, with a live meter of what it has spent.
* **Writes, casts and designs**: the agent writes and edits scenes, designs shot lists with a length for every shot, and builds your cast with portraits, turnaround sheets and costume boards, plus locations and props.
* **Renders for you**: images for each shot using your cast and location references, then video from those images, with clips up to 30 seconds on Seedance 2.5.
* **See the plan**: bigger jobs show a task list that ticks off as the agent works; every action appears in the chat with its result.
* **Keeps going without you**: close the tab and the run continues; come back and the chat picks up where it is.
* **More ways in**: attach PDF, DOCX, TXT or Fountain scripts up to 2 MB and up to 10 images per message.
* The AI model and Agent/Chat pickers are gone: the agent chooses how to act from your message.
### Community — now available to everyone [#community--now-available-to-everyone]
* **Share your work publicly**: publish images, videos, music and Auteur work from the tools to the Community gallery, and browse what others are making. See [Community](/docs/collaborate/community).
## August 9, 2026 [#august-9-2026]
### New AI models [#new-ai-models]
* **Seedance 2.5** in the Video Generator — clips from 4 to 30 seconds at 480p or 720p, with optional audio. Leave the duration on **Auto** to let the model pick the length; you're charged for the length actually produced.
## August 7, 2026 [#august-7-2026]
### Billing [#billing-2]
* **Chapa plans renew on a fixed day**: ETB subscriptions now renew on the same day each month as the day you subscribed (month-end dates adjust in shorter months). Paying your renewal link early keeps your renewal day.
## August 3, 2026 [#august-3-2026]
### Auteur — our new AI filmmaking studio (private beta) [#auteur--our-new-ai-filmmaking-studio-private-beta]
* **Auteur is here, in invite-only early access**: turn a screenplay, treatment, or one-line idea into a full production — character bibles, identity plates, costumes, locations, storyboards, shot lists, and finished video clips — on one infinite canvas with a built-in production assistant. Unlock with an invite code from the dashboard, or join the waitlist.
* **Draft Mode**: build a cheap draft of your entire film hands-off. Segments render in parallel at the lowest-cost tiers, and everything lands marked DRAFT for you to review and re-render selectively.
* **New Angle**: re-shoot any finished clip from a different camera angle — the performance, timing, dialogue, and lip-sync stay frozen; only the lens moves. Target a single shot from your shot list to pay only for those seconds.
* **Assembly**: cut your film without re-rendering anything — reorder segments, trim in/out points, choose clip audio vs. stems, add crossfades, and set the score level. The cut plays live in the film player and carries into export.
* **Export to Editor**: ship an editorial package as one zip — every take, audio stems, storyboards, reference plates with their exact prompts, docs, and a ready-to-import EDL/FCPXML timeline for your NLE.
* **Score**: an optional continuous music bed composed to land its musical phrases exactly on your segment boundaries, played under the whole cut and ducked beneath dialogue.
* **Seamless film player**: gapless segment-to-segment playback, proper fullscreen, scene audio overlaid in sync, and the score running under the entire film.
* **Collaboration**: members draft and generate while directors approve and finalize; share projects with outside reviewers who can watch and comment; a review queue collects everything awaiting sign-off, plus per-card comment threads and a project activity feed.
* **Inpaint and Generate editing**: draw directly on any image (or drop stick figures to pose characters) and describe the fix, or re-render a card's subject from a prompt plus up to 3 of your own reference images. Every edit is saved as a switchable version.
* **Your references are in charge**: upload a reference image and ask for it to be used — including replacing a character's identity — and the model follows your instruction.
* **Costume looks**: characters with multiple outfits get one wardrobe board per look, automatically matched to the segments where each is worn.
* **Project aspect ratios**: choose 16:9, 9:16, 1:1, 4:3, 3:4, or 21:9 at setup — every plate and clip renders in your chosen frame.
* **Image resolution tiers**: pick 1K, 2K, or 4K output per project. Identity plates now default to the highest-quality model at 4K, since they anchor every downstream render (this raises their per-image cost).
* **Consistent character voices**: each character's voice — timbre, pace, accent, delivery — is defined once in their Character Bible and carried into every clip.
* **Storyboards follow your shot list exactly**: one panel per shot, in order, at the shot's framing — no invented coverage, and characters who are only heard are never drawn.
* **Budget control**: a per-project spend ledger with CSV export, and an optional credit budget cap that warns on manual runs and hard-pauses automated ones.
* **More creative control**: a target-runtime field, a "resegment" chat command, an exact-prompt editor on every generated image, mid-project Add Card for missing props/characters/locations, upload-your-own media on any card, and routing uploaded assets onto any card that can hold them.
* Plus a unified run-mode selector (Manual / Draft / Automate / Agent), a draggable floating chat panel, content mode for short-form pieces, continuous takes across segments, and a quality-tier picker for clips.
### New AI models [#new-ai-models-1]
* **Google Omni** — video generation with synchronized audio, from text or references.
* **Seedream 5 Pro** — high-end image generation and editing with up to 10 reference images.
* **VEED Lipsync v2** — a second lip-sync engine, selectable alongside Sync Lipsync v2 Pro.
* **Seedance 2.0 Mini** and **Nano Banana 2 Lite** — fast, low-cost video and image models, now in all the standard tools.
* **Kling 3.0 family** — Pro, Standard, 4K, Motion Control, and Reference-to-Video, with multi-shot prompts, element binding, and dialogue support.
* **GPT Image 2** and **GPT Image 2 Edit** — including inpainting with masks.
* **Nano Banana 2**, **Seedream 5 Lite**, and **Qwen Image 2 / Edit** across the image tools and moodboard.
* **Seedance 2.0 / 2.0 Fast** — text, image, and multimodal reference-to-video (up to 8 images + 3 videos + 3 audio clips) with native audio.
### Image & video tools [#image--video-tools]
* **Redesigned generation pages**: a floating prompt bar with your most-used controls inline, results in a chat-style feed, and older history loading as you scroll.
* **@-mentions in prompts**: type `@` to reference attached images, videos, audio, and elements.
* **Easier references**: paste images from the clipboard, upload multiple files at once, and drag any generated image in as a reference.
* **Polish prompt**: one free click rewrites your prompt into a more cinematic version for review before generating.
* **Generation queue**: a unified panel shows all in-flight generations with progress, and concurrent slots now scale with your plan (up to 5 on Enterprise).
* **Fair "auto" duration billing**: videos with automatic duration are billed for the length actually produced — the excess hold is refunded on completion.
* **Clearer feedback**: smooth progress, plain-language errors, and stalled generations now fail cleanly with automatic refunds.
### Moodboard [#moodboard]
* **Video on the canvas**: generate clips directly on the board, right-click any image to "Animate this," and play results inline.
* **Story cards and connectors**: six card types (scene, dialogue, character, shot, note, direction) with markdown, plus arrows to map your story's flow.
* **Groups**: gather items into named, draggable, recolorable frames.
* **A canvas-aware AI agent**: a chat assistant that sees your board and can add cards, connect nodes, generate images — and build out a treatment, characters, storyboard frames, and animated clips from a single idea, with your approval before anything paid runs.
* **Export**: canvas PNG, a storyboard PDF ordered by your connectors, and project JSON backup.
### Collaboration & workspaces [#collaboration--workspaces]
* **Private by default**: image, video, and moodboard sessions are visible only to you unless you share them.
* **Live workspace switching**: changing workspace updates every tool instantly.
* **Spendable credits at a glance**: the header shows what you can actually spend in the current workspace, including any member limit.
* **Reliable member limits**: workspace credit limits now account for in-flight generations.
### Billing & pricing [#billing--pricing]
* **Pricing updated to reflect provider costs** for Seedance 2.0 / 2.0 Fast video and high-resolution image generation. These generations now cost more credits than before.
* **Lip-sync is now billed per second** of your source video instead of a flat rate.
* **Subscription credits no longer roll over** past an unpaid billing period. Add-on credits are never affected.
* **Cancel from the billing page**, with the end date shown once scheduled.
* **1,000 free credits at signup**, granted once you verify your email.
### Account & support [#account--support]
* **Google Sign-In**, with automatic linking for existing email accounts.
* **Session management**: see every active login and revoke any of them from Settings.
* **First-run tour** for new users.
* **Faster support**: in-app tickets now reach the team through more channels.
## February 27, 2026 [#february-27-2026]
The 1.0 release rounded out the creative toolset and made billing self-serve:
* **Music Generator**: original audio generation and track extension, with an audio player and history gallery.
* **Storyboard view** with drag-and-drop scene reordering.
* **In-app support**: submit help tickets directly from the sidebar.
* **Unified Billing page**: usage, plan management, credit history, and invoices in one view.
* **Redesigned pricing page**: clearer plan comparisons with monthly credits and capacity estimates per tier.
* **Smoother checkout**: subscription and plan status appear immediately after payment, for both card and Chapa payments.
* **Dark mode refinements** across the app.
# Community (/docs/collaborate/community)
**Community** is Ehudai's public gallery. Publish your best work from the tools, show it under your own creator profile, and borrow ideas from other creators with one-click **Remix**.
Open it from **Community** under **Resources** in the sidebar.
## Browsing [#browsing]
The Community page opens with **Featured creators** at the top, followed by a grid of posts.
* **Sort** with the tabs: **Featured** (the default), **Recent**, or **Trending**. Trending favours posts that are getting views and likes right now.
* **Filter** with **All**, **Images**, or **Videos**. Music posts appear under **All**.
* Hover a tile to see its title, creator, views, and likes. **Load more** brings in the next batch.
Click a post to open it. You'll see:
* the image, video, or audio
* the title and caption, and the creator (click their name to visit their profile)
* the **Prompt** and model used, unless the creator kept them private
* tags, views, and likes
Click **Like** to like a post (click **Liked** to undo), or **Remix** to build on it.
## Publishing your work [#publishing-your-work]
You can publish finished generations from these tools:
| Tool | Where to find it |
| ---------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------ |
| Image Generator, Video Generator, [Marketing Studio](/docs/tools/marketing-studio) | Open an item from your history and click **Publish to Community** |
| Music Generator | **Publish** on a song |
| [Auteur](/docs/tools/auteur) | **Publish** in the full-screen viewer for an image, storyboard, video clip, or audio |
The **Publish to Community** dialog asks for:
* **Title** — required, up to 150 characters (pre-filled from your prompt)
* **Caption (optional)** — for example, how you made it
* **Tags (optional)** — press Enter after each one, up to 10
* **Show my prompt publicly** — on by default. Turn it off to keep your prompt and model private; others then can't remix the post.
Click **Publish**. The button in the tool then reads **In Community**. Each generation can be published once.
The first time you publish, the dialog also asks you to set up your creator profile — a **Handle** and a **Display name**, both pre-filled from your account. You can change them later.
## What's public when you publish [#whats-public-when-you-publish]
Published posts appear in the gallery in the app and in the public gallery on ehudai.com, where anyone can see them without an account. Each post shows:
* the image, video, or audio
* its title, caption, and tags
* your prompt and model, unless you turned off **Show my prompt publicly**. For Marketing Studio posts, the prompt is only the words you typed, not the format's built-in instructions.
* your handle, display name, bio, and profile picture
* its view and like counts
A post is a separate copy of your work. Deleting the original from your tool history doesn't remove the post — unpublish it instead.
If you're not on a paid plan, published images carry a small "Made with Ehudai" mark.
## Your creator profile [#your-creator-profile]
Every creator gets a public profile page at `ehudai.com/community/creators/your-handle`. It shows your profile picture, display name, @handle, bio, total posts, views, and likes, followed by your posts.
Edit it under **Account → Community**, in the **Creator profile** section:
* **Handle** — 3–50 characters: letters, numbers, dots, underscores, and hyphens. It must be unique. After you change your handle, you'll have to wait before changing it again; the page shows the date you can next change it.
* **Display name**
* **Bio** — up to 500 characters
Click **Save profile**. Your profile picture is copied from your account picture when your creator profile is first set up.
## Managing your posts [#managing-your-posts]
**Account → Community** also lists **My posts**, with each post's views and likes. To take one down, click **Unpublish** and confirm. The post leaves the gallery, and your original generation stays in your history. You can publish it again later.
Posts can't be edited after publishing — unpublish and publish again to make changes. If a post shows a status such as **In review**, it isn't visible in the gallery yet.
## Remixing [#remixing]
Found a post you like? Click **Remix** to open the matching tool with the post's prompt already filled in:
| Post | Opens | Pre-filled |
| ----- | ---------------------------------------------- | ---------------- |
| Image | [Image Generator](/docs/tools/image-generator) | Prompt and model |
| Video | [Video Generator](/docs/tools/video-generator) | Prompt and model |
| Music | [Music Generator](/docs/tools/music-generator) | Prompt |
The model is set only if the tool offers it; otherwise pick one yourself. Other settings and reference images aren't copied. Tweak the prompt and generate as usual — remixing costs the same credits as any generation.
**Remix** only appears on posts whose creator shared their prompt.
# Collaboration overview (/docs/collaborate)
Ehudai has two complementary ways to work with others, plus a public gallery:
**Workspaces** are your team's home. Members join with a role (admin, member, or viewer), see shared projects, use the tools with the workspace's credits, and — in Auteur — draft productions the director finalizes.
**Story sharing** works per project. Share a single story with specific people as editors or viewers, open it to your whole workspace, or create a public view link to a read-only page anyone can open without an account, and signed-in visitors can remix.
**Community** is where you show finished work to everyone. Publish images, videos, and music from the tools to a public gallery, build a creator profile, and remix other creators' prompts.
# Sharing stories (/docs/collaborate/sharing-stories)
Every story has its own sharing controls, independent of workspace membership. Open them with **Share** in the story editor's top bar.
Collaborators spend credits from the **story owner's** workspace when they generate — the Share panel says so at the top. See [Team credits](/docs/collaborate/team-credits).
## Inviting collaborators [#inviting-collaborators]
Type an address into **Invite by email**, pick a permission, and click **Invite**:
* **Editor** — can edit and share the story
* **Viewer** — can only view
If they already have an Ehudai account they're added right away and get an email plus an in-app notification with **Accept** / **Decline**. Otherwise they receive an email invitation, valid for 7 days, and join once they sign up or log in with that email address.
Owners and editors can invite. Once someone has access, click the *"N people with access"* line under the invite box to open **Who has access**, where the owner can remove collaborators.
## Project access [#project-access]
**Project access** sets who can open the story beyond the people you've invited. Only the story owner can change it.
| Level | Who can open the story |
| ---------------------- | ---------------------------------------------------------------------------------------------- |
| **Personal** (default) | Only you, plus people you've invited |
| **Workspace** | Everyone in the story's workspace, plus people you've invited |
| **Public** | Anyone with the public view link (a read-only page they can remix), plus people you've invited |
With **Workspace** access, workspace members join as viewers. The owner can switch any member between **Viewer** and **Editor** in **Who has access**.
## Public view links [#public-view-links]
Under **Public view link** — *"Anyone with this link can view and remix this project. To let someone edit, invite them by email."* — the owner clicks **Create link**, then **Copy** to copy the link.
The link opens a **read-only page** and doesn't add anyone as a collaborator. Anyone can open it without logging in. The page shows:
* the story's title and genre
* its cover image
* the story text
* when it was created and last updated
To give someone editing rights, invite them by email instead.
### Remixing from a public link [#remixing-from-a-public-link]
Signed-in visitors can click **Remix** on the public page to copy the whole project (characters, shots, generated media, and scripts) into a new project in their own personal workspace. The copy is theirs to edit and generate in; your original isn't affected. Visitors who aren't signed in are asked to log in first. Remix works only while the link does.
The link works while **Project access** is **Public**, and also right after you create it on a Personal story. To turn it off, set **Project access** to **Workspace**, or select **Personal** again. There's no button to disable or regenerate the link: its address stays the same, so setting access back to **Public** makes the same link work again.
## Duplicating [#duplicating]
From Home, open a story card's menu and choose **Duplicate** to copy the story — scenes, shots, characters, and scripts — into a new project called "Copy of …", in the workspace you're currently viewing. Useful for versioning before a big restructure.
# Team credits (/docs/collaborate/team-credits)
## The rule: the workspace owner funds the work [#the-rule-the-workspace-owner-funds-the-work]
When a workspace member (or a story collaborator) generates anything, the credits come from the **workspace owner's** balance — members don't need their own subscription to contribute. If the owner's balance can't cover a generation, the member sees *"Workspace owner has insufficient credits."*
The header always shows what **you** can actually spend in the current workspace — your own balance in your personal workspace, or the owner-funded amount (and any limit that applies to you) in a team workspace.
## Per-member spending limits [#per-member-spending-limits]
Owners can cap each member from the member list (**Set Spending Limit**):
* The limit counts everything the member spends in the workspace, **including in-flight generations** — a member can't blow past their cap with parallel renders.
* The member table's **Credits Spent / Limit** column shows each member's total spend against their limit (∞ when there's no limit), with a progress bar that turns yellow at 75% and red at 90%. Only owners and admins see this column; members and viewers can't see anyone else's spend or limit.
* No limit set means the member is bounded only by the owner's balance.
## Story collaborators outside the workspace [#story-collaborators-outside-the-workspace]
People you share an individual story with also bill the story owner's workspace when they generate — the share dialog says as much. Note that per-member limits apply to workspace members; an outside collaborator is bounded by your balance, so share editing rights deliberately.
## In Auteur [#in-auteur]
Workspace collaboration is built into Auteur's approval flow: members generate and draft (billed to the workspace), but their locks stay provisional until the director — the project owner, or the workspace owner or an admin — finalizes them. Each project keeps a ledger of the credits it has spent, and only a director can change its budget cap. See [Auteur](/docs/tools/auteur).
# Workspaces (/docs/collaborate/workspaces)
A workspace holds a team's projects, members, and spending. Every account starts with one workspace, named after you ("Abebe's Workspace", or "My Workspace" if you didn't give a first name). Switch between workspaces from the workspace switcher in the sidebar, and every tool follows — histories, projects, and credits are all workspace-scoped.
The switcher lists each workspace with its member count. On the current workspace, the gear icon opens its settings and, for owners and admins, the **+** icon opens the invite form.
## Creating workspaces [#creating-workspaces]
Choose **Create New Workspace** at the bottom of the switcher. The dialog asks for a **Workspace Name** and an optional **Description**.
* **Free accounts** have their one workspace. Creating more requires a paid plan — without one, you'll be prompted to upgrade.
* **Subscribers** can create as many workspaces as they need.
Workspaces can't currently be deleted — rename or repurpose one instead, or contact support.
## Roles [#roles]
| Role | Can do |
| ---------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Owner** | Everything: invite and remove members, change roles, set spending limits, and control tool access. The workspace bills against the owner's credits. |
| **Admin** | Invite members, resend or revoke pending invitations, edit the workspace name and description, and see every member's spend and limit. In Auteur, admins can also finalize work, like the owner. |
| **Member** | View and create projects, comment, and generate (spending the owner's credits). |
| **Viewer** | View and comment only. |
The owner can't be removed. From the **⋯** menu on a member's row, the owner can **Make Admin**, **Make Member**, **Set Spending Limit**, or **Remove Member**.
## Inviting people [#inviting-people]
Owners and admins open **Workspace → People** and click **Invite Members**, then enter an email and a role (**Admin**, **Member**, or **Viewer**). Members and viewers don't see the invite button. What happens:
* The invitee gets an email invitation — and if they already use Ehudai, an in-app notification too.
* Invitations are valid for **7 days**. Pending ones show in the member list with a **Pending** badge, where the owner or an admin can **Resend Invitation** (restarts the 7 days) or **Revoke Invitation**.
* The invitee must accept while signed in with **the exact email you invited**.
There are no shareable join links for workspaces — invitations are personal, per email address. Inviting people doesn't require a paid plan.
## Workspace settings [#workspace-settings]
The **Settings** tab holds the workspace name and description. Owners and admins can edit them and click **Save Changes**; members and viewers see them read-only. Owners also see a **Tool Access** card: switch Image Generator, Video Generator, AI Assistant, Moodboard, and Music Generator on or off for the whole workspace, then click **Save Tool Settings**. Generations with a tool that's switched off are blocked in that workspace.
## What's shared [#whats-shared]
* **Projects** — stories set to **Workspace** access are visible to all members (as viewers by default). Members always see their own projects and anything shared with them directly. See [Sharing stories](/docs/collaborate/sharing-stories).
* **Moodboards and Auteur projects** — private to their creator by default. Use **Share** in the tool to make one visible to the whole workspace or to share it with specific members.
* **Credits** — members spend from the owner's balance, within optional per-member limits. Only owners and admins see what each member has spent. Details in [Team credits](/docs/collaborate/team-credits).
# Core concepts (/docs/getting-started/concepts)
## Stories, scenes, and shots [#stories-scenes-and-shots]
Everything you make in AI Studio lives in a **story**. A story breaks down into **scenes**, each scene into **shots**, and each shot can carry generated **images and videos**:
```
Story
└── Scene — a unit of your narrative (INT./EXT., location, time)
└── Shot — one camera setup: size, movement, composition, lighting, sound
└── Image / Video — AI-generated visuals for that shot
```
You don't have to build this by hand. In [AI Studio](/docs/tools/ai-studio), the agent breaks an attached screenplay into scenes, extracts its characters, and designs the shot list for each scene.
## Credits [#credits]
Credits are Ehudai's single currency for AI generation, shared across every tool.
* Every generative action — an image, a video second, a music track, a chat reply — costs credits. The exact cost is always shown in the UI before you commit.
* As a rule of thumb, **1,000 credits ≈ $1** of plan value.
* When a generation starts, its cost is **held** from your balance. For most generations, the amount held is what you're charged when it finishes.
* A few generations settle to the real cost and refund the difference: AI chat replies (Assistant and AI Studio), Seedance videos with the duration set to **Auto**, and image batches that come back with fewer images than you asked for.
* Failed generations are refunded in full, automatically. So are generations in the Image, Video and Music Generators, Marketing Studio and the Moodboard tool that make no progress for 30 minutes.
* Credits come from your plan (monthly), one-time top-ups, or the 1,000-credit signup grant.
Full details in [How credits work](/docs/plans-billing/credits) and per-model prices in [Credit costs](/docs/guides/credit-costs).
## AI models [#ai-models]
Ehudai aggregates the leading AI models — image (Nano Banana, Seedream, Flux, GPT Image, Qwen, Runway and more), video (Seedance, including Seedance 2.5, Kling, Veo, Wan, Google Omni), chat (GPT, Claude, Gemini), and music. **All models are available on every plan**; they differ only in credit cost per generation.
## Workspaces [#workspaces]
A **workspace** is where your projects, your team, and your spending live. Every account starts with one workspace, and its owner (or an admin) can invite members into it with a role (admin, member, or viewer). Creating additional workspaces requires a paid plan. Inside a shared workspace, members generate using the **owner's** credit balance, with optional per-member spending limits.
More in [Workspaces](/docs/collaborate/workspaces).
## Generations and the queue [#generations-and-the-queue]
Generations run in the background — you can keep working while they render. Active generations appear in a queue panel in the bottom-right corner, showing how many are running out of your limit. How many can run at once depends on your plan (from 1 on Free and Starter up to 5 on Enterprise). When you hit the limit the generate button reads **"Queue full"**; wait for a generation to finish, or cancel one from the queue panel with its **×** button (shown on generations that can be cancelled). Cancelling refunds its credits.
# Create your account (/docs/getting-started/create-account)
## Sign up [#sign-up]
Go to the login page and click **Sign up** under the form. The page switches to **Create your account**. You can sign up two ways:
### Email and password [#email-and-password]
1. Enter your email address and a password, then type the password again to confirm it. Your first and last name are optional. Click **Create account**.
2. Your password needs at least 8 characters, with an uppercase letter, a lowercase letter, and a number. The rules are listed under the password field and tick off as you type.
3. Check your inbox for a verification email and click the link. The link is valid for 48 hours.
Your **1,000 free credits** are granted when you verify your email — not at signup. If your balance shows zero, verify your email first.
If your link has expired or was already used, opening it shows an error page with a **Send new verification link** form. Enter your email there to get a fresh link. You can also sign in and click **Resend email** on [Account → Profile](/docs/account/profile). You can request a few per minute at most.
Disposable email addresses are not accepted — sign up with a real inbox you can verify.
### Sign in with Google [#sign-in-with-google]
Click **Sign up with Google**. Google accounts are verified automatically, so your 1,000 free credits are granted immediately and your workspace is created on the spot.
If you already have an Ehudai account under the same email, signing in with Google links the two — you can then use either method. Linking doesn't count as verifying: if you never verified that account's email, verify it to get your free credits.
## What you get on day one [#what-you-get-on-day-one]
* **1,000 free credits** — roughly 7 images on a typical model, or 12–16 on lower-cost models such as Seedream or Qwen Image. Costs per model are always shown in the UI before you generate; see [Credit costs](/docs/guides/credit-costs).
* **A workspace** — named after you (for example "Abebe's Workspace", or "My Workspace" if you didn't give a first name), ready for your first project.
* **Access to every tool** — all AI models are available on every plan, including free. Plans differ in monthly credits and how many generations you can run at once, not in which models you can use.
## Logging in [#logging-in]
Log in with your email address **or your username**, plus your password, or with **Sign in with Google**. Forgot your password? Use **Forgot password?** on the login page and enter your account's email. You can review and revoke active sessions any time under [Account → Settings](/docs/account/security).
## Next [#next]
Head to the [quickstart](/docs/getting-started/quickstart) to make your first project.
# What is Ehudai? (/docs/getting-started)
Ehudai brings the whole creative pipeline into one place. Instead of switching between a writing app, an image tool, a video tool, and an editor, you work in a single studio where each stage feeds the next:
1. **Write** — describe your story or upload a screenplay. AI Studio breaks it into scenes, shots, and characters.
2. **Visualize** — build moodboards, design characters, and generate concept images with 15+ image models.
3. **Animate** — turn shots into video clips with the leading video models, from Seedance and Kling to Veo.
4. **Produce** — score your work with generated music, assemble clips, make product ads, and (in Auteur) export an editor-ready package.
5. **Collaborate** — share stories, invite your team into a workspace, control who can edit or view, and publish your best work to the Community gallery.
## The tools [#the-tools]
| Tool | What it does |
| ------------------------------------------------ | ------------------------------------------------------------------------------------- |
| [AI Studio](/docs/tools/ai-studio) | Create stories, screenplays & scenes with AI |
| [Assistant](/docs/tools/assistant) | Chat with AI to brainstorm & refine ideas |
| [Image Generator](/docs/tools/image-generator) | Generate visuals from text prompts |
| [Video Generator](/docs/tools/video-generator) | Bring your stories to life with AI video |
| [Moodboard](/docs/tools/moodboard) | Curate visual references & inspiration boards |
| [Music Generator](/docs/tools/music-generator) | Compose original scores & soundtracks |
| [Marketing Studio](/docs/tools/marketing-studio) | Make ready-to-post product ads — UGC, unboxing, product reviews, and more |
| [Auteur](/docs/tools/auteur) | Turn a screenplay into a film — storyboards, shots, and AI video clips (private beta) |
Every generative action is paid for with **credits** — one balance that works across all tools and models. New accounts get 1,000 free credits once they verify their email. See [How credits work](/docs/plans-billing/credits).
When you make something you're proud of, publish it to [Community](/docs/collaborate/community) — a public gallery where others can browse your work and remix your prompts.
## Next steps [#next-steps]
# Quickstart (/docs/getting-started/quickstart)
This walkthrough takes you from a blank account to a story with scenes, shots, and generated visuals. Budget 15–30 minutes and a few hundred credits.
## 1. Start from Home [#1-start-from-home]
After signing in you land on **Home** — the launchpad for every tool. If it's your first visit, a short tour points out the essentials. Your credit balance is always visible in the header at the top of the page.
## 2. Create a story in AI Studio [#2-create-a-story-in-ai-studio]
Click **AI Studio**. You have two ways in:
* **Describe your idea** — a sentence or two is enough. The agent develops it into a film with you through chat.
* **Attach a screenplay** — PDF, DOCX, TXT or Fountain, up to 2 MB. The agent reads it and builds the structure for you. See [Uploading a screenplay](/docs/guides/upload-screenplay) for format details.
Leave the mode on **Manual** to direct each step yourself, or pick **Auto** and set a credit budget to let the agent build the whole film on its own. See [Manual and Auto](/docs/tools/ai-studio#manual-and-auto).
## 3. Watch the agent build your structure [#3-watch-the-agent-build-your-structure]
You land in the story editor: chat on the left, your **Scene List** in the center, characters on the right. The agent starts working immediately — for bigger jobs its plan appears as a task list above the chat while scenes populate.
Ehudai structures every story the same way:
**Story → Scenes → Shots → Images & Videos**
## 4. Review and refine [#4-review-and-refine]
* Toggle between **List view** and **Storyboard view**.
* Drag scenes to reorder; expand a scene to edit its title, description, and shots.
* Ask the chat for changes in plain language: *"make scene 4 more tense"*, *"trim the dialogue in scene 2"*, *"analyze the pacing"*. See [AI Studio](/docs/tools/ai-studio) for everything the agent can do.
## 5. Build your cast on the Moodboard [#5-build-your-cast-on-the-moodboard]
Click **+** on the character rail to open the story's Moodboard. Upload reference images, generate new ones, or use **Auto-generate** to build a full board from your screenplay. For a consistent look across generations, you can [train a custom LoRA model](/docs/guides/consistent-characters).
## 6. Generate images and video for shots [#6-generate-images-and-video-for-shots]
The agent designs a shot list for each scene — ask it to *"design the shots for scene 3"* if it hasn't yet. Each shot carries real cinematography fields — shot size, camera movement, composition, lighting, sound — plus image and video prompts. Ask the agent to render them, or open a shot and click **Generate Image** or **Generate Video** yourself. Credits are held when a generation starts and refunded automatically if it fails.
## 7. Assemble and export [#7-assemble-and-export]
* **Review All Scenes** collects every generated video into a reorderable sequence you can combine into one cut.
* **Export PDF** produces a professional shot-breakdown document.
* **Share** your story with collaborators or publish a public read-only link. See [Sharing stories](/docs/collaborate/sharing-stories).
## Where to go next [#where-to-go-next]
# Consistent characters (/docs/guides/consistent-characters)
Getting the *same* character in shot after shot is the difference between a gallery and a film. Ehudai gives you three levels of control.
## Level 1: Reference images [#level-1-reference-images]
Every reference-capable image model accepts your images alongside the prompt. Generate or upload one strong character image, then attach it as a reference to every generation of that character — paste it, drag it from history, or select it on a moodboard canvas.
For the strongest identity carry-over, use edit-mode models (Seedream Edit, GPT Image 2, Qwen Edit): they treat the reference as the subject to keep, not just inspiration.
## Level 2: Character anchoring in stories [#level-2-character-anchoring-in-stories]
In AI Studio, build your cast on the story's **Moodboard** (the **+** on the character rail). When generating images for a shot, attach the character's moodboard image as an anchor — every shot generation of that character then starts from the same face.
In **Auteur** this is systematic: each character gets an identity plate rendered once at high quality, then referenced into every costume board, storyboard, and clip that character appears in.
## Level 3: Train a custom LoRA [#level-3-train-a-custom-lora]
For a look no reference image can pin down — a specific character across wildly different scenes, or a signature style — train a custom model from the Moodboard.
### Preparing training data [#preparing-training-data]
Training uses **before/after image pairs** in a ZIP:
```
training.zip
├── shot1_start.jpg ← input image
├── shot1_end.jpg ← what it should become
├── shot1.txt ← edit instruction for this pair
├── shot2_start.jpg
├── shot2_start2.jpg ← optional extra reference for the pair
├── shot2_end.jpg
├── shot2.txt
└── ...
```
* Every pair needs a `_start` and `_end` image with matching names.
* Give each pair its edit instruction in its own `.txt` file, or fill in the dialog's **Default caption** to cover every pair that has no file. A pair with neither falls back to a generic caption, so caption every pair one way or the other.
* Extra references for a pair are named `_start2`, `_start3`, and so on.
### Training [#training]
Click **Train LoRA** in the Moodboard toolbar to open the **Train Model** dialog (the same dialog on the standalone Moodboard and on a story's Moodboard), then fill in:
| Field | What to enter |
| ------------------------- | ----------------------------------------------------------------- |
| **Model** | **Qwen Image Edit Plus — LoRA trainer** (the one option today) |
| **Training Name** | A name you'll recognise later |
| **Reference Image Count** | How many `_start` images each pair has — **1 to 3** |
| **Default caption** | Optional. The edit instruction for any pair without a `.txt` file |
| **Training Data (ZIP)** | Your ZIP file |
Start the training (80 credits). The number of training steps is fixed and can't be changed, so the quality of your pairs is what you control: consistent, well-captioned pairs matter most. Training runs in the background — you can close the dialog, and the new LoRA appears in your trained list when it finishes.
### Using it [#using-it]
On a **Moodboard**, either the standalone tool or your AI Studio story's board, click **Select LoRA** and pick up to 3 trained models under **Select Trained LoRAs**. Selected LoRAs appear as chips under **Selected LoRAs** and steer every generation on that board toward your trained look. Combine with a reference image of the character for maximum consistency.
On the standalone Moodboard, LoRAs run on **Qwen Image Edit Plus LoRA**. Select at least one canvas image as the base for it to edit; the model picker stays locked to that model, and shows its price, until you remove the LoRAs. See [Moodboard](/docs/tools/moodboard#generating-with-your-loras).
# Credit costs (/docs/guides/credit-costs)
Reference tables for credit costs across the platform. Two things to remember:
* **These tables are for planning.** The Image Generator, Video Generator and Marketing Studio show the price for your current options, and every charge is listed in **Account → Billing → Historical Usage**.
* Costs move with your settings — resolution, size, duration, and audio all change the price.
## Images (per image) [#images-per-image]
| Model | Credits |
| ------------------- | ----------------------------------------------------------------------------- |
| Nano Banana 2 Lite | 46 / 68 / 102 / 154 at 0.5K / 1K / 2K / 4K |
| Seedream 3 | 60 |
| Seedream 4 / 4 Edit | 60 |
| Seedream 5 Lite | 70 |
| Qwen Image 2 / Edit | 70 |
| Z Image | 20 |
| Nano Banana | 80 |
| Flux Kontext Pro | 80 |
| Seedream 4.5 | 80 |
| Nano Banana 2 | 90 / 134 / 202 / 302 at 0.5K / 1K / 2K / 4K |
| Runway | 100 at 16:9 or 9:16, 160 at 1:1 |
| Seedream 5 Pro | 135, or 270 at 2048×2048 |
| Flux 2 Flex | 240, rising with references (500 with 4) |
| Nano Banana Pro | 270 at 1K or 2K, 480 at 4K |
| GPT Image 1 | 160 at 1:1, 240 at other ratios |
| GPT Image 2 | 12 to 1,741 by quality and size ([table](/docs/tools/image-generator#models)) |
Reference images add about 1 credit each on the Nano Banana models. The Image Generator's model pill shows the exact price for your settings.
## Video (per second) [#video-per-second]
| Model | Credits/second |
| --------------------------------------- | ---------------------------------------------------------- |
| Seedance 1.5 | \~46 (480p) / \~104 (720p) with audio; \~23 / \~52 without |
| Wan 2.5 | 100 / 200 / 300 at 480p / 720p / 1080p |
| Kling 2.6 | 140 (silent) / 280 (audio) |
| Seedance 2.0 Mini | \~144 (480p) / \~309 (720p) |
| Kling 3.0 Standard | 168 / 252 (audio) / 308 (audio + voice) |
| Kling 3.0 Standard – Reference to Video | 168 / 210 (audio) |
| Seedance 2.5 | \~208 (480p) / \~462 (720p) |
| Kling 3.0 Pro | 224 / 336 (audio) / 392 (audio + voice) |
| Kling 3.0 Pro – Reference to Video | 224 / 280 (audio) |
| Seedance 2.0 Fast | \~225 (480p) / \~484 (720p) |
| Google Omni | 250 (from text) / 260 (with references) |
| Seedance 2.0 | \~281 (480p) / \~605 (720p) / \~1,361 (1080p) |
| Veo 3.1 | 200 (silent) / 300 (audio) |
| Kling 3.0 Motion Control | 336 |
| Kling 3.0 4K | 448 / 672 (audio) / 784 (audio + voice) |
Prices marked \~ are rounded: those models are billed by the amount of video they render, which doesn't come to a whole number of credits per second. A 5-second clip at about 200 credits per second costs about 1,000 credits.
Veo 3.1 with reference images (from a shot's **Generate Video** page in AI Studio) always renders 8 seconds and costs a flat **6,400 credits** with audio or **3,200** without.
## Everything else [#everything-else]
| Operation | Credits |
| ------------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------- |
| Music generation or extension | 120 flat |
| Image upscale (Moodboard) | 20 |
| LoRA training | 80 |
| Combining story clips into one video | 60 |
| Lip sync | Per second of your source video: \~167 (Sync Lipsync v2 Pro) / 140 (VEED Lipsync v2). In Auteur: 800 per started minute of the clip |
| Chat / agent messages | By actual tokens used; minimum 1 — typically a few credits, more for long premium-model conversations |
| Moodboard generation | Same as 4 images of the chosen model. With trained LoRAs, priced by output size and image count, shown next to **Selected LoRAs** |
| Marketing Studio ads | Shown next to the **Launch** button for your format and settings — see [Marketing Studio](/docs/tools/marketing-studio) |
## Billing mechanics [#billing-mechanics]
Estimated costs are **held** when a generation starts and become the charge on completion; failures and cancellations are refunded automatically. Seedance 2.0 and 2.5 videos with **Auto** duration are held at the model's maximum length — 15 seconds for Seedance 2.0, 2.0 Fast and 2.0 Mini, 30 seconds for Seedance 2.5 — and charged for the length actually produced. Details in [How credits work](/docs/plans-billing/credits).
# FAQ (/docs/guides/faq)
## Credits & billing [#credits--billing]
**I signed up but my balance is zero.**
Verify your email — the 1,000 free credits are granted when you click the verification link, not at signup. Can't find the email? Click **Resend email** on **Account → Profile**. Google sign-ups get their credits immediately.
**My balance dropped as soon as I clicked Generate.**
That's the credit hold — the estimated cost is reserved up front. For most generations the hold is the final charge; it comes back in full if the generation fails or you cancel it. See [How credits work](/docs/plans-billing/credits#holds-how-a-generation-is-charged).
**A generation failed. Where are my credits?**
Refunded automatically — the held credits go straight back to your balance. You can see it in the transaction history under **Account → Billing → Historical Usage**.
**My unused credits disappeared at renewal.**
Subscription credits don't roll over; each period starts fresh at your plan's amount. Credits from top-up purchases never expire — see [How credits work](/docs/plans-billing/credits).
**How do I upgrade or downgrade?**
Cancel your current plan — it stays active until the period ends — then subscribe to the new one once that period is over. You can't start a new plan while the current one is still active, and in-place switching isn't supported yet. See [Manage your subscription](/docs/plans-billing/manage-subscription).
**When do I get a low-credit warning?**
When your balance drops below 150 credits, including when it reaches zero. You get one warning, by email and in the app, each time you run low; the next comes only after new credits arrive and you run low again. See [Running out](/docs/plans-billing/credits#running-out).
**Why can't I buy a top-up?**
Top-ups require an active plan. Once you have one, you can buy a top-up at any balance — see [Credit top-ups](/docs/plans-billing/payments#credit-top-ups).
## Generations [#generations]
**It says my queue is full, or that I have too many active generations.**
You're at your plan's concurrency limit (Free/Starter: 1, Pro: 2, Business: 3, Enterprise: 5). Wait for a running generation to finish or cancel one.
**Can I close the tab while generating?**
Yes — generations run server-side and land in your history when done.
**A generation has been spinning for a long time.**
Generations that time out waiting for the AI provider are failed automatically and refunded. In the Image Generator, Video Generator, Music Generator, Marketing Studio and the Moodboard tool, a generation that makes no progress for 30 minutes is failed and its credits are refunded, and it stops counting toward your concurrency limit. If one still seems stuck, refresh and check your history, then contact support.
## Stories & files [#stories--files]
**What screenplay formats can I upload?**
PDF, DOCX, TXT or Fountain up to 2 MB in AI Studio (text PDFs, unencrypted); TXT or PDF — including scans — in Auteur. No direct FDX import yet; export to PDF first. See [Uploading a screenplay](/docs/guides/upload-screenplay).
**Can I export my screenplay?**
AI Studio exports a professional PDF shot breakdown. Auteur exports a full editorial package with an EDL/FCPXML timeline. Images, videos, and music download individually from their tools.
**Who pays when my collaborator generates something in my story?**
You do — collaborators and workspace members spend from the owner's balance. Owners can set per-member limits. See [Team credits](/docs/collaborate/team-credits).
## Access [#access]
**How do I get into Auteur?**
It's in private beta — redeem an invite code on the dashboard's Auteur banner, or join the waitlist there.
**Can I change my account email?**
Not currently. For account changes beyond what Profile offers — including deletion — contact support via **Help & Support** in the app.
**Something else?**
Open a ticket from **Help & Support** in the app sidebar, or email [hello@ehudai.com](mailto:hello@ehudai.com).
# Guides & reference (/docs/guides)
# Uploading a screenplay (/docs/guides/upload-screenplay)
## Supported files [#supported-files]
| | AI Studio | Auteur |
| -------------- | ------------------------------------------ | ----------------------------------------------------- |
| Formats | **PDF, DOCX, TXT or Fountain** | **TXT or PDF** (scanned PDFs work too), or paste text |
| Size limit | **2 MB** | Larger files supported |
| Encrypted PDFs | Not supported — export an unencrypted copy | Not supported |
Final Draft (`.fdx`) files aren't imported directly yet — export your script to PDF from your writing app first. Fountain files can be attached in AI Studio, and Fountain text can be pasted into Auteur directly.
## What happens on upload [#what-happens-on-upload]
Ehudai reads your script and builds a structured breakdown the whole platform works from:
* **Scenes** — from your scene headings (`INT.`/`EXT.`, numbered scenes, or structural breaks; non-English screenplays are handled)
* **Characters** — extracted from dialogue and action lines
* **Locations and time of day** — from your headings
* A first pass of **shots** per scene, with cinematography metadata
In AI Studio, attach the script in the chat composer (paperclip). The agent reads it and breaks it into scenes and characters, then you can start asking for changes.
## Tips for clean parses [#tips-for-clean-parses]
* Standard screenplay formatting parses best: sluglines, character names over dialogue, action paragraphs.
* One script per file. Title pages are fine — the title and author are picked up.
* If your PDF is a scan, Auteur reads it; AI Studio works best with true text PDFs.
* Big script? For AI Studio, stay under 2 MB — a text-only PDF export of even a feature-length script fits comfortably.
# Welcome to Ehudai (/docs)
Ehudai is one platform for the whole creative pipeline: write screenplays, generate images and video with the world's best AI models, build moodboards, compose music, make product ads, and — with Auteur — take a script all the way to a finished film.
These docs cover everything in the product. If you're new, start with the quickstart; if you're here for a specific tool or a billing question, jump straight to the section.
## Explore the docs [#explore-the-docs]
## For AI agents [#for-ai-agents]
These docs are machine-readable. Every page is available as plain Markdown — append `.md` to any page URL, or request it with an `Accept: text/markdown` header. A full index lives at [/llms.txt](/llms.txt), and the entire docs corpus at [/llms-full.txt](/llms-full.txt).
## Need help? [#need-help]
Open the account menu at the bottom of the app sidebar (your name) and choose **Help & Support** to open a ticket, or email [hello@ehudai.com](mailto:hello@ehudai.com).
# How credits work (/docs/plans-billing/credits)
Credits are the single currency for all AI generation on Ehudai. Every tool and model draws from the same balance.
## Two pools [#two-pools]
Your balance is the sum of two pools that behave differently:
| | Subscription credits | Add-on credits |
| --------------------- | -------------------------------------------------------------- | -------------------------------------------------------------------------------------- |
| Come from | Your monthly plan, the signup grant, credits granted by Ehudai | One-time [top-up purchases](/docs/plans-billing/payments#credit-top-ups); some refunds |
| Expire? | **Reset each billing period** — unused credits don't roll over | **Never expire** |
| Survive cancellation? | No — cleared when the subscription ends | Yes |
**Subscription credits are always spent first.** Your purchased top-up credits are protected until the monthly allowance is used up.
On each successful renewal your subscription pool is **reset** to the new period's amount — unused subscription credits from last month expire (you'll see this as an "Expired … unused subscription credits" line in your transaction history). If a subscription ends unpaid, its remaining subscription credits are cleared. Add-on credits are never touched by any of this.
The 1,000 free signup credits live in the subscription pool — when your first paid month starts, the plan's allowance **replaces** whatever remains of them. Use them.
## Holds: how a generation is charged [#holds-how-a-generation-is-charged]
1. When you start a generation, its **estimated cost is held** — your available balance drops immediately.
2. When it completes, the hold becomes the charge. For most tools the estimate *is* the price, so you pay exactly what was held.
3. If the generation **fails or you cancel it, the full hold is returned** to your balance automatically.
In the Image Generator, Video Generator, Music Generator, Marketing Studio and the Moodboard tool, a generation that makes no progress for 30 minutes is failed automatically and its hold is refunded in full.
A few generations settle to less than the hold when they finish, and the difference comes back to you:
* **Seedance 2.0 and 2.5 videos with "Auto" duration** are held at the model's maximum length and charged for the length actually produced.
* **Chat and AI Studio messages** are charged for the tokens actually used.
* **Image batches** that come back with fewer images than you asked for are charged only for the images delivered.
## Running out [#running-out]
* When your balance drops **below 150 credits**, including all the way to zero, you get a low-credit warning by email and in the app. You get it **once each time you run low**: it isn't repeated while you stay low, only after new credits arrive (a renewal, top-up or grant) and you run low again. It's never sent more than once a day.
* When a generation would cost more than you have, it's blocked before any work starts, and you're shown how many credits it needs.
* Workspace members can also hit their per-member limit — a separate cap set by the workspace owner. See [Team credits](/docs/collaborate/team-credits).
## Tracking your usage [#tracking-your-usage]
**Account → Billing** shows:
* **Current Period** — your available credits, total earned and spent, a usage bar, and your next plan refresh date.
* **Historical Usage** — every transaction: what was charged, refunded, released, granted, or expired, with your balance after each.
* **Invoices** — your payment history across Stripe and Chapa.
Per-model prices are listed in [Credit costs](/docs/guides/credit-costs).
# Plans (/docs/plans-billing)
Ehudai pricing is credit-based: your plan grants a monthly pool of credits, and every generation draws from it. **All AI models and features are included on every plan** — plans differ in how many credits you get and how many generations you can run at once.
The pricing page picks your currency from your location: in Ethiopia you see **ETB** plans paid through Chapa; everywhere else you see **USD** plans paid by card through Stripe. There's no currency switch on the pricing page. See [Payments](/docs/plans-billing/payments).
The figures below describe the current plans. If anything here differs from the **Pricing** page in the app, the Pricing page is what you'll be charged.
## Free [#free]
| | |
| ---------------------- | ------------------------------------------------------------- |
| Price | $0 |
| Credits | 1,000 one-time signup credits (granted on email verification) |
| Concurrent generations | 1 |
| Workspaces | 1 |
Free accounts can use every tool and model until their signup credits run out. Top-up purchases require a subscription.
## USD plans [#usd-plans]
| | Starter | Pro | Business |
| ---------------------- | ------------------------ | -------------------------- | ------------------------------------ |
| Price | **$9**/month | **$35**/month | **$90**/month |
| Monthly credits | 9,000 | 35,000 | 90,000 |
| Roughly | \~60 images / \~8 videos | \~250 images / \~33 videos | \~660 images / \~85 videos |
| Concurrent generations | 1 | 2 | 3 |
| Workspaces | Unlimited | Unlimited | Unlimited |
| Support | Email | Email | Priority + dedicated account manager |
Pro is the most popular plan. USD plans renew automatically each month by card and can be cancelled anytime.
## ETB plans (Ethiopia) [#etb-plans-ethiopia]
| | Starter | Pro | Business |
| ---------------------- | ------------------------ | -------------------------- | -------------------------- |
| Price | **2,000 ETB**/month | **10,000 ETB**/month | **20,000 ETB**/month |
| Monthly credits | 7,000 | 36,000 | 72,000 |
| Roughly | \~50 images / \~6 videos | \~260 images / \~34 videos | \~530 images / \~68 videos |
| Concurrent generations | 1 | 2 | 3 |
| Workspaces | Unlimited | Unlimited | Unlimited |
ETB plans renew via a monthly payment link — see [Payments](/docs/plans-billing/payments) for how Chapa renewals work.
## Enterprise [#enterprise]
Custom credit packages, priority support with an SLA, and custom terms. Concurrency up to 5 parallel generations. Contact [hello@ehudai.com](mailto:hello@ehudai.com).
Enterprise plans are set up by Ehudai rather than bought on the Pricing page, so they have no self-serve checkout or cancellation — see [Manage your subscription](/docs/plans-billing/manage-subscription#plans-managed-for-your-organization).
Per-member spend limits aren't Enterprise-only: any workspace owner can set them. See [Team credits](/docs/collaborate/team-credits).
## The image/video estimates [#the-imagevideo-estimates]
The "\~N images" figures assume a typical image model (about 135 credits per image); the video figures assume a typical 5-second clip (about 1,050 credits). On the **Pricing** page, each plan card shows the same two estimates in its **Images** and **5s videos** boxes. Real consumption depends entirely on which models and settings you use — a Z Image generation costs a few credits while a 4K flagship image costs hundreds. See [Credit costs](/docs/guides/credit-costs) for per-model numbers.
# Manage your subscription (/docs/plans-billing/manage-subscription)
Everything lives at **Account → Billing → Subscription**: your current plan, its price, and your renewal date.
## Cancelling [#cancelling]
Click **Cancel plan**, optionally tell us why, and confirm with **Cancel subscription**. Cancellation always takes effect **at the end of your billing period**:
* You keep full access and your remaining subscription credits until the period ends.
* The button changes to **Cancellation scheduled**, and the page says *"Your plan ends on … and will not renew."*
* When the period ends: unused subscription credits expire, your add-on credits remain, and your account returns to Free limits (1 concurrent generation). Workspaces you already have are kept; you just can't create new ones until you subscribe again.
## Switching plans [#switching-plans]
Ehudai doesn't yet support in-place plan changes or proration. While a plan is active — including after you've scheduled its cancellation — subscribing to another plan is blocked with a message to cancel first. A new plan can start only **after your current period ends**.
Schedule the cancellation, use your remaining subscription credits (they don't roll over), then subscribe to the new plan from the **Pricing** page once the current period has ended.
## Renewals [#renewals]
* **Stripe** — automatic. Your card is charged monthly and the new period's credits are granted on payment. If a renewal payment fails, your subscription credits are removed right away and Stripe retries the charge; if a retry succeeds, you get the full new period's credits.
* **Chapa** — manual, via the renewal link sent by email and in-app notification about 3 days before your period ends. The billing page shows when your next payment is due, with a **Pay renewal now** button once the link is ready. Paying an old link after the plan has ended doesn't bring it back, and the payment will be refunded. See [Payments](/docs/plans-billing/payments).
## Plans managed for your organization [#plans-managed-for-your-organization]
Enterprise plans can be set up for your organization by Ehudai. The billing page shows these as *"plan, managed for your organization"*, with the date the plan renews. These plans renew automatically each month with a fresh credit allowance, and there's no self-serve cancellation — contact your account manager to make changes.
## Plans granted by an admin [#plans-granted-by-an-admin]
Some accounts have their plan granted directly by an administrator. These accounts show *"granted by your administrator"* on the billing page and have no self-serve billing or cancellation — contact your admin or [hello@ehudai.com](mailto:hello@ehudai.com) for changes.
# Payments (/docs/plans-billing/payments)
Ehudai supports two payment providers. The pricing page picks one from your location: Chapa (ETB) in Ethiopia, Stripe (USD) everywhere else.
## Stripe — international (USD) [#stripe--international-usd]
Credit and debit cards, processed by Stripe.
* Clicking **Get Started** on a plan opens Stripe's secure checkout in a new tab. Promotion codes are accepted at checkout.
* Subscriptions **renew automatically** each month on your card.
* After payment your plan activates and your monthly credits are granted immediately.
## Chapa — Ethiopia (ETB) [#chapa--ethiopia-etb]
Local payments in Ethiopian Birr, processed by Chapa.
* Clicking **Get Started** opens Chapa's hosted payment page in a new tab; on success your plan is active and your credits are granted.
* **Renewal is manual.** ETB plans don't renew automatically. Chapa subscriptions renew **monthly on the same day you subscribed** (a June 15 subscription renews July 15, August 15, and so on; month-end dates adjust in shorter months). About 3 days before your period ends, you receive a **renewal payment link** by email and in-app notification — pay it to keep your renewal day.
* **Account → Billing** shows *"Your next payment is due on …"*, and a **Pay renewal now** button once your renewal link has been sent.
* If the period ends unpaid, your remaining subscription credits are cleared and you're notified that the plan has expired. You then have a **14-day grace period**: paying the renewal link within 14 days of your renewal date reactivates the plan with a fresh full month counted from the payment date, which becomes your new renewal day.
* After 14 days unpaid, the subscription is cancelled. To come back, subscribe again from the **Pricing** page.
* **A cancelled plan stays cancelled.** Paying an old renewal link after the plan has ended (whether you cancelled it or the grace period ran out) doesn't reactivate it, and no credits are granted. The same applies during the grace period if you've already started another plan. You'll see **Plan not renewed**; the payment will be refunded. Contact support from **Help & Support** if you have questions, or click **Start a new subscription** to subscribe again.
There's no auto-charge with Chapa. If you miss the renewal link, your plan's unused monthly credits are cleared when the period ends, and the plan is cancelled 14 days later. Your add-on credits are never affected.
## Credit top-ups [#credit-top-ups]
Need more credits mid-cycle? Buy a one-time **Credit Top-up**:
| | |
| ------- | ------------------------------------------------------------ |
| Credits | **20,000** |
| Price | **$20** (Stripe) or **5,600 ETB** (Chapa) |
| Expiry | **Never** — top-up credits survive renewals and cancellation |
One rule: top-ups require an **active plan**.
Top-ups stack — you can buy one at any balance, and the new credits are added to whatever you have left.
Buy from **Account → Billing → Current Period → Purchase Add-on Credits**: choose a payment method (card or Chapa) and click **Purchase credits**. Checkout opens in a new tab.
## Where to see your payments [#where-to-see-your-payments]
**Account → Billing → Invoices** lists every payment with its **Date**, **Description**, **Amount**, **Gateway**, and **Status**.
# AI Studio (/docs/tools/ai-studio)
AI Studio is the narrative heart of Ehudai. Describe an idea or attach a screenplay, and the AI Studio agent develops it with you: it writes the scenes, designs the shots, builds the cast and locations, and renders images and video for each shot.
## Starting a film [#starting-a-film]
Open **AI Studio** from the sidebar. You'll see *"What film are we making today?"* and a single composer:
* **Describe your idea**: a sentence or two is enough, for example *"A heist thriller set in 1970s Addis Ababa"*.
* **Attach files** (paperclip): a screenplay, reference images, or both. See [Attachments](#attachments).
* **Genre** (optional): a hint such as Drama, Sci-Fi or Commercial. Leave it unset and the agent works the genre out from your idea.
* **Mode**: **Manual** or **Auto**. See [Manual and Auto](#manual-and-auto).
Press **Enter** to start (**Shift + Enter** adds a new line). The agent opens your new project and gets to work.
## Manual and Auto [#manual-and-auto]
| Mode | What happens |
| -------------------- | --------------------------------------------------------------------------------------------------------------------- |
| **Manual** (default) | You direct each step. The agent does what your message asks, reports back, and waits for your next message. |
| **Auto** | The agent plans the whole film and keeps working through its plan on its own until it's done or your budget runs out. |
In **Auto**, you set a **credit budget** (default 1,000 credits; anywhere from 50 to 200,000). A meter shows credits spent against the budget. When the budget is used up, the agent stops starting new work, sums up what's finished and what's left, and ends the run. An Auto run can last up to 90 minutes.
You can switch mode on any message.
## The project workspace [#the-project-workspace]
| Panel | What it does |
| ----------------------- | -------------------------------------------------------------------- |
| **Chat** (left) | Your conversation with the agent, its plan, and its results. |
| **Scene List** (center) | Your story's structure, in list or storyboard view. |
| **Characters** (right) | Your cast's reference images; the **+** opens the project Moodboard. |
The top bar holds **Export PDF**, **Share**, and your credit balance.
### Following the agent's work [#following-the-agents-work]
* **The plan**: for bigger jobs the agent writes a task list, shown above the chat, and ticks tasks off as it goes.
* **Tool cards**: each action the agent takes (writing a scene, generating a portrait, rendering a clip) appears in the chat as a card with its result: images inline, and before-and-after changes for edited scenes.
* **Stop**: stops the current run after the step in progress. Renders that already started still finish.
* **Leaving is fine**: a run keeps going if you close the tab. Come back to the project and the chat reconnects to it.
You can have one run going per project at a time; send your next message when it finishes, or stop it.
## What the agent can do [#what-the-agent-can-do]
The agent works directly on your project. Ask in plain language.
**Story and script**
* Write or rewrite scenes: *"Write the scene where they finally confront each other"*, *"Make scene 4 more tense"*
* Make exact line edits: *"Change Marcus's line 'I'm leaving' to 'I'm already gone'"*
* Break an attached screenplay into scenes and extract its characters
* Find every line a character speaks, or summarise the structure so far
* Set the project's look and tone; the agent keeps a style guide (format, tone, visual style) and applies it to every image and video it makes
**Shots**
* Design a shot list for a scene. Each shot gets a size, camera movement, action and a length in seconds, planned around what the video model can render in one clip.
* Revise shots, and delete scenes you no longer need
**Cast, locations and props**
* Create characters and generate their **portrait**, a **turnaround sheet** (front, profile, back, three-quarter, expressions) and **costume boards**
* Create locations and props, and generate reference images for them
* Portraits and location images are pinned to the project Moodboard automatically
**Images and video**
* Generate images for shots, using your cast and location references to keep faces and places consistent
* Render video for a shot, typically from the shot's image first so the clip matches the frame you approved. Clips run up to 15 seconds, or up to 30 seconds at 480p or 720p on Seedance 2.5.
* Check on renders that are still in progress; finished clips attach to their shots
**Moodboard**
* Generate mood images, pin an image you've shared, or add text notes to the board
## Attachments [#attachments]
Click the paperclip in the composer. You can attach up to 10 files per message.
| Type | Formats | Limit | What the agent does with it |
| ---------------------- | ------------------------ | ---------- | --------------------------------------------------------- |
| **Screenplay or text** | PDF, DOCX, TXT, Fountain | 2 MB | Reads it as your story's source text |
| **Images** | PNG, JPG, WebP, GIF | 10 MB each | Looks at them, and uses them as references for generation |
PDFs must be text PDFs and unencrypted; scanned pages can't be read. See [Uploading a screenplay](/docs/guides/upload-screenplay).
## Scenes and shots [#scenes-and-shots]
Expand any scene to edit its title and description (rendered with screenplay formatting) and manage its shots. Each shot carries real cinematography fields:
* **Shot size** (CU, MS, WS, ECU), **camera movement**, **composition**, **lighting**, **sound**
* An **action** description of what the camera sees
* Dedicated **image** and **video prompts** for generation
Add shots by hand with **Add Shot**, or ask the agent to design them.
From any shot, **Generate Image** and **Generate Video** open dedicated generation pages with model options and a chat sidebar for iterating by hand.
## Assembling and exporting [#assembling-and-exporting]
* **Review All Scenes** collects every shot video into a reorderable sequence you can combine into a single cut. Each scene also has its own combine page.
* **Export PDF** generates a professional A4 shot-breakdown document: cover page, every scene with its shots and cinematography details, in industry-standard Courier.
The export is a **shot breakdown** for production planning, not a page-for-page formatted screenplay. Images and videos are downloaded individually from their generation pages.
## Sharing [#sharing]
Use **Share** in the top bar to invite collaborators as editors or viewers, open the story to your workspace, or publish a read-only public link. See [Sharing stories](/docs/collaborate/sharing-stories).
Your chat with the agent is your own: collaborators see the story, scenes and shots, not your conversation.
## Credits [#credits]
* **The agent's thinking and writing** is billed by actual usage for each step it takes.
* **Images and video** the agent generates are billed at the same per-model rates as the standalone generators. See [Credit costs](/docs/guides/credit-costs).
* In **Auto**, everything the run spends counts against the budget you set.
If you run out of credits mid-run, the chat tells you and links to **Get More Credits**.
# Assistant (/docs/tools/assistant)
The Assistant is a general-purpose AI chat, separate from any story. Use it to brainstorm, draft, research, or think out loud — with the model of your choice.
## Sessions [#sessions]
Conversations are saved as named sessions in the left sidebar. Start one with **New Chat**, reopen any session from **History** to continue where you left off, or delete ones you no longer need. Sessions belong to the workspace you're in.
## Models [#models]
Models are grouped into three tiers in the model menu:
| Model | Maker | Tier | Notes |
| --------------------- | --------- | -------- | ------------------------------------------------------------------- |
| GPT-6 Luna | OpenAI | Fast | Fastest GPT-6, built for everyday chat |
| Gemini 3.5 Flash-Lite | Google | Fast | Google's most cost-effective model; reads images & PDFs |
| Claude Haiku 4.5 | Anthropic | Fast | Fastest Claude, with near-frontier intelligence |
| GPT-6 Sol | OpenAI | Balanced | Strong reasoning at a working price |
| Gemini 3.8 Flash | Google | Balanced | Google's most intelligent Flash; reads images & PDFs |
| Claude Sonnet 5 | Anthropic | Balanced | The best combination of speed and intelligence |
| GPT-6 Astra | OpenAI | Premium | OpenAI's most capable model, for the hardest reasoning and research |
| Claude Opus 5.5 | Anthropic | Premium | For long, demanding work and deep knowledge tasks |
| Gemini 3.1 Pro | Google | Premium | Google's flagship; reads images & PDFs |
New users start on a **Balanced** model. After that, your last-used model is remembered per device.
## Attachments [#attachments]
Attach **images** or a **PDF** with the paperclip. Every model can read images. PDFs are read by the Gemini models, so pick one of those when your question is about a PDF.
## Credits [#credits]
Messages are billed by actual tokens used — longer conversations and longer replies cost more, with a minimum of 1 credit per message. **Fast** models cost a fraction of **Premium** ones; for most messages the cost is a few credits. An estimate is held when you send and settles to actual usage when the reply completes.
The Assistant is not connected to your stories. To have AI modify a story's scenes and shots, use the chat inside [AI Studio](/docs/tools/ai-studio).
# Auteur (/docs/tools/auteur)
Auteur is Ehudai's filmmaking studio. It takes a screenplay, a treatment, or a one-line idea and carries it through an entire production pipeline: cast design, locations, storyboards, shot lists, and finished 15- or 30-second video clips, all on one infinite canvas, with a production assistant in the chat rail.
Auteur is in invite-only early access. On the dashboard's Auteur card, enter an invite code and click **Unlock**, or join the waitlist there and we'll be in touch. Without access, the tool redirects back to Home.
## Choose your medium [#choose-your-medium]
Every project starts by picking the look you're directing in. It sets the visual language for every character, scene, and clip (you can change it later):
| Medium | Look |
| ------------ | ----------------------------------------------------------------------- |
| Cinematic | Live-action realism: anamorphic light, real-world physics, film grain |
| 3D Animation | Sculpted form, soft global illumination |
| Anime | Cel-shaded 2D: bold linework, flat colour, expressive motion |
| Stylized | Art-directed graphic worlds: shape, palette, bold direction |
| Experimental | Avant-garde, mixed media |
| Custom | Describe your own style and add reference images; Auteur synthesizes it |
| Content | Short-form non-film work: trailers, ads, social, explainers |
**Content** is a format you choose when you create the project. It can't be switched on or off later, and the top bar shows it read-only as **Format: Content**.
## Set up the project [#set-up-the-project]
Only a **Title** is required. Anything you leave out is generated in the pipeline:
* **Plot**: a logline or short synopsis.
* **Screenplay**: paste it, or upload `.txt` / `.pdf` (scanned PDFs are read too).
* **Treatment**: a director's treatment as `.txt`, `.md`, or `.pdf`.
* **Character bible**: `.txt`, `.md`, or `.pdf`. Authoritative if provided; otherwise characters are extracted from the screenplay.
* **Reference images**: up to 12 PNG, JPEG, or WebP images, each tagged **General**, **Style**, **Character**, **Location**, **Costume**, or **Moodboard**.
* **Frame**: the project aspect ratio: 16:9, 9:16, 1:1, 4:3, 3:4, or 21:9. Locked once the project builds; every plate and clip renders in it.
* **Runtime**: a stepped slider from **Auto** through 0:30, 0:45, 1:00, 1:30, 2:00 and on up to 120:00 (two hours). A set runtime overrides auto-detection.
* **Engine**: **Seedance 2.0** builds 15-second segments and offers the Mini, Fast, and Standard quality tiers up to 1080p. **Seedance 2.5** builds 30-second single takes at 480p or 720p, with more references and higher face and lighting fidelity, at a higher price. Segment length is locked when the project is built.
* **Generation**: **Just-in-Time** (the default) generates only what each segment needs, when you reach it, for a fast first video. **Upfront** builds every character, location, and asset as you go. You can switch later from the top bar.
## The production canvas [#the-production-canvas]
Your screenplay is broken into segments: 15 seconds each on Seedance 2.0, 30 seconds on Seedance 2.5. Cards are laid out across four columns:
| Column | Cards |
| ------------------ | ---------------------------------------------------------------------------------- |
| **01 · Character** | Character Bible, Character Identity, Costume Design, Technical Sheet, Voice Anchor |
| **02 · World** | Location, Props, Cinematic Style, Score |
| **03 · Coverage** | Shot List, Blocking / Floor-Plan, Storyboard |
| **04 · Output** | Scene Audio, Seedance Prompt, and the clip itself (for example **15-Second Clip**) |
Technical Sheet, Voice Anchor, Score, and Scene Audio are optional. In Just-in-Time mode, cards are grouped per segment, and a segment opens once the previous segment's clip is approved and locked. In Upfront mode, shared assets sit in one pool.
Cards flow left to right: identities and world plates anchor the look, coverage plans the camera, output renders the clip. Every card can be inspected, edited, re-run, uploaded over with your own media, and **approved & locked**. Locking is what feeds a card's result into everything downstream. Missed a prop or character? **Add Card** creates it mid-project.
### Card tools [#card-tools]
* **Technical Review**: an independent check of an image or text card against its references. The verdict and any suggested correction are posted in the chat.
* **Renderer**: image cards that are ready to run let you pick the model: **Auto**, **NB2 Lite**, **Nano Banana 2**, **GPT Image 2**, or **Seedream 5 Pro**.
* **Inpaint**: draw directly on an image (or drop stick figures to pose characters), describe the fix, and click **Apply edit**.
* **Generate**: re-render the card's subject from a prompt plus up to 3 of your own reference images. Your references are authoritative, including when replacing a character's identity.
Every edit is kept as a switchable version.
## Run modes [#run-modes]
| Mode | What it does |
| -------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Manual** | You drive: run and approve each card yourself. |
| **Draft Film** | Builds the whole film in parallel at the cheapest tiers. Choose **Full draft** or **Fast draft** (skips boards and audio). Everything lands marked DRAFT for review. |
| **Automate** | Generates every needed card straight through to the end, hands-off. |
| **Agent** | Runs the pipeline semi-autonomously, with an independent technical judge reviewing outputs. |
In Manual mode, **Lean Plan** skips optional assets (costume, technical sheets, blocking) the film doesn't need. Lead characters are never skipped. Draft Film and Automate run it automatically.
## Quality and cost controls [#quality-and-cost-controls]
* **Quality**: **Mini · cheapest** (the default), **Fast · cheaper**, **Standard · max quality**, or **2.5 · next-gen**, set for the project.
* **Resolution**: 480p, 720p (the default), or 1080p. 1080p is available on Standard only; 2.5 renders at 480p or 720p.
* **Image tier**: **Auto** (the default), 1K, 2K, or 4K. On **Auto**, each card renders at its own default: character identity, technical sheet, costume, and location plates at 2K, everything else at 1K. Pick a tier and every image card renders at it, overriding those defaults; choose **Auto** again to go back. Draft Film always renders images at 1K. Character identity plates default to Seedream 5 Pro in the Cinematic and Content mediums and to Nano Banana 2 Lite in the others, because every downstream render depends on them.
* **References per clip**: up to 9 images, 3 videos, and 3 audio files on Seedance 2.0; up to 30 images, 10 videos, and 10 audio files on Seedance 2.5.
* **Ledger and budget cap**: the credits chip in the top bar opens the project's spend ledger (with CSV export) and an optional budget cap. You're warned at 80% of the cap. A manual run that would go over asks you to confirm; Draft Film, Automate, and Agent runs stop.
* **Paid confirmations**: a manual clip render always opens **Render clip — paid** first, with the quality, resolution, and length, the **Estimated cost**, and your **Available** credits. If you don't have enough, **Render & spend** is disabled. With reference videos attached, the estimate is marked **≈**: their exact length is measured when the render starts, so the charge can differ slightly. On a workspace project someone else owns, the owner pays, so the modal shows **Billed to the workspace** instead of your balance.
* While a clip renders, its card names the Seedance version it's rendering on (for example *Recording video with Seedance 2.5…*). This is the tier actually used, which can be cheaper than the top bar shows during a Draft Film run.
## Finishing [#finishing]
* **New Angle**: re-shoot any finished clip from a different camera; the performance stays, only the lens moves. When a segment has a shot list, you pick a single shot to cover, so you pay only for those seconds.
* **Lip-sync**: on a finished clip, pick an MP3, WAV, or M4A file (up to 25 MB). A **Lip-sync — paid** confirmation shows the price, 800 credits per started minute of the clip, and your balance. Nothing is uploaded or charged until you click **Lip-sync & spend**. The clip is then re-synced to your audio and the result is added as a new version; the original stays untouched.
* **Score**: an optional continuous music bed composed so its phrases land on your segment boundaries, ducked under dialogue.
* **Assemble**: cut the film as pure metadata: reorder, trim, per-clip audio, cross-dissolves, music level. Nothing re-renders.
* **Play**: plays your cut end to end.
* **Editor**: exports one zip with every take (including New Angle takes), audio stems, storyboards and blocking plans, reference plates with their exact prompts, docs, and an EDL / FCPXML timeline ready for Resolve, Premiere, or Final Cut. Lock at least one clip first.
* **Publish**: open an image card's lightbox and click **Publish** to share it to the Community gallery.
## Collaboration [#collaboration]
Auteur uses your [workspace roles](/docs/collaborate/workspaces). Members run cards and generate (billed to the workspace), but they can't approve; their automated locks stay provisional until a director signs off. Directors (the project owner and workspace admins) approve and finalize.
On a workspace project, the owner or an admin can **Share** it: keep it private, open it to the whole workspace, or share it with specific members. Track everything awaiting sign-off in the review queue (**N to review**), leave notes on cards, and follow the project's **Activity** feed. Simultaneous edits are last-writer-wins, so coordinate who's driving.
## Credits [#credits]
Auteur is the most credit-intensive tool, which is what the ledger and budget cap are for. Planning steps, Lean Plan, and Technical Review bill like chat. Plates and boards bill per image. Clips bill per second at Seedance rates by quality tier and resolution. Lip-sync bills 800 credits per started minute of the clip. Clip renders and lip-sync show their price before you confirm.
# Image Generator (/docs/tools/image-generator)
The Image Generator is a chat-style workspace: your prompt bar sits at the bottom, results flow upward, and older generations load as you scroll. Click any image to open it full-screen, where you can **Download** it, delete it, or **Publish to Community**.
## Generating [#generating]
Type a prompt and pick your options from the pills on the prompt bar:
* **Model**: the pill shows a per-image credit guide for the selected model.
* **Aspect ratio**: each model lists the ratios it supports (Nano Banana 2 and 2 Lite offer 14). The ratio pill is hidden for models whose shape comes from a size preset or the input image.
* **Count**: 1 to 4 images per generation.
* **Settings**: per-model options such as resolution, size, quality, seed, negative prompt, guidance, steps, and output format.
## Reference images [#reference-images]
Models that support references let you steer a generation with existing images. Three ways to add one:
1. The image button on the prompt bar.
2. **Paste** an image straight from your clipboard.
3. **Drag** a file from your desktop, or any previously generated image from your history.
Multiple references are allowed. GPT Image 2 switches into edit mode automatically when a reference is attached and uses up to 4 references. Seedream 5 Pro uses up to 10.
## Inpainting (GPT Image 2) [#inpainting-gpt-image-2]
With GPT Image 2 and a reference image attached, **Settings** shows a **Mask** upload: a PNG with transparency, matching the reference's dimensions. Transparent areas get repainted from your prompt; opaque areas are preserved. When a mask is attached, only the first reference image is used.
## Models [#models]
| Model | Credits per image | References |
| ------------------------------- | ------------------------------------------- | -------------------- |
| Nano Banana 2 *(default)* | 90 / 134 / 202 / 302 at 0.5K / 1K / 2K / 4K | Yes |
| Nano Banana 2 Lite | 46 / 68 / 102 / 154 at 0.5K / 1K / 2K / 4K | Yes |
| Nano Banana | 80 | Yes |
| Nano Banana Pro | 270 at 1K or 2K, 480 at 4K | Yes |
| Seedream 5 Pro | 135, or 270 at 2048×2048 | Yes (up to 10) |
| Seedream 5 Lite | 70 | No |
| Seedream 4.5 | 80 | No |
| Seedream 4 Edit / Text To Image | 60 | Edit: yes |
| Seedream 3 | 60 | No |
| Flux Kontext Pro | 80 | Yes |
| Flux 2 Flex | 240, rising with references (500 with 4) | Yes |
| GPT Image 2 | 12 to 1,741 by quality and size (see below) | Yes (up to 4) + mask |
| GPT Image 1 | 160 at 1:1, 240 at other ratios | No |
| Runway | 100 at 16:9 or 9:16, 160 at 1:1 | Yes |
| Qwen Image 2 / Edit | 70 | Yes |
| Z Image | 20 | No |
The model pill shows the exact credits for your current settings, image count and references, so what you see is what you're charged. Multiply the table by your image count. Reference images add about 1 credit each on the Nano Banana models. Failed generations are refunded automatically.
**GPT Image 2** is priced by quality and size. **Auto** quality is billed as High.
| Size | Low | Medium | High |
| ---------------------- | --- | ------ | ----- |
| 1024×1024 | 12 | 106 | 422 |
| 1536×1024 or 1024×1536 | 12 | 82 | 330 |
| 2048×1152 | 15 | 123 | 495 |
| 2048×2048 | 27 | 219 | 880 |
| 3840×2160 (4K) | 53 | 433 | 1,741 |
**Model highlights:**
* **Nano Banana 2**: resolution tiers from 0.5K to 4K (1K by default), a **Thinking Level** control (*High* does better on complex prompts), and **Search Grounding**, which pulls real-time facts from Google Search into the generation.
* **Seedream 5 Pro** and **Seedream 5 Lite**: an **Image Size** preset in **Settings**, up to 2048×2048 (XL) on Pro and 3072×3072 (HD) on Lite. Seedream 5 Pro takes up to 10 reference images.
* **GPT Image 1**: renders at Medium quality.
* **GPT Image 2**: sizes up to 3840×2160, **Quality** (Auto, Low, Medium, High), PNG/JPEG/WebP output, **Background** (Auto or Opaque; transparent backgrounds aren't supported), and **Moderation** (Auto or Low).
* **Z Image**: the budget workhorse for fast drafts.
## Tips [#tips]
* Start cheap. Iterate composition with Z Image or a Lite model, then re-run the winning prompt on a premium model at high resolution.
* Drag a generated image back into the prompt bar as a reference to iterate on it.
* Deleting an image removes all images from the same generation.
* A generation that's still running can be cancelled from the queue panel in the bottom-right corner; its credits are refunded.
# Tools overview (/docs/tools)
Ehudai's tools live in the sidebar of the app and share everything that matters: your workspace, your history, and your credits.
## Behavior shared by every tool [#behavior-shared-by-every-tool]
**Workspace-scoped.** Each tool shows the history of the workspace you're in. Switch workspaces from the top bar and every tool follows instantly.
**Private by default.** What you generate in the Image, Video, Music, and Marketing Studio tools is visible only to you. In a workspace, Moodboard boards and Auteur projects can be shared with your whole workspace or with specific members. To show work publicly, use **Publish to Community** on an image, video, track, or ad.
**Credits up front, refunds on failure.** Starting a generation holds its estimated cost, and in most cases that estimate is what you're charged. The exception is **Auto** duration on the Seedance video models, which settles to the length actually produced. If a generation fails, your credits come back automatically.
**Cancel anytime.** Running generations appear in the queue panel in the bottom-right corner. Click the **×** next to one to cancel it and refund its credits.
**Concurrency by plan.** Free and Starter run 1 generation at a time, Pro 2, Business 3, Enterprise 5. At your limit, the generate button shows **Queue full** until a slot frees up.
**Costs are visible before you commit.** The Image Generator, Video Generator and Marketing Studio show the price for your current options, and it updates as you change quality, size, duration, resolution and references.
# Marketing Studio (/docs/tools/marketing-studio)
Marketing Studio turns a product photo and a short description into a finished ad. Pick a format, attach your product, choose a presenter, and launch. It's built for social and commerce: creator-style UGC clips, hands-on reviews, unboxings, how-tos, polished TV spots, CGI product motion, and packs of product stills.
Open **Marketing Studio** from the **Tools** section of the sidebar. It's available to every signed-in account.
## Make an ad [#make-an-ad]
1. **Pick a format.** The format pill on the prompt bar opens **Ad format presets**, a set of cards with a looping sample of each format. Hover a card for a one-line description.
2. **Attach your product.** Use the **Product** button next to the prompt box to upload a photo of the item you're advertising.
3. **Choose a presenter** (presenter formats only). Click **Select Avatar** and pick one of the built-in presenters, or upload your own presenter photo with the **Character** button.
4. **Add a style reference** (optional). The **Reference** button takes an image for look, palette, or scene inspiration.
5. **Describe the ad.** Say what should happen, where, and in what mood. The format supplies its own camera language and structure, so a sentence or two is often enough.
6. **Set the options.** Aspect ratio, plus resolution and duration for video formats.
7. Check the credit estimate next to the button, then click **Launch**.
Results appear in the feed above the prompt bar, newest at the bottom, each tagged with its format (for example **UGC**). Click a result to preview it (the preview names the model that made it), **Download** it, or **Publish to Community**. If you leave or reload the page, a running generation keeps going and reappears when you come back.
To remove a result, click its trash icon (**Delete result**) on the card or in the preview, then confirm **Delete this result?**. Everything that generation produced is removed from your history, and this can't be undone. A result can't be deleted while it's still generating. Posts you've already published to Community stay up.
When you publish, the post shows only the words you typed, not the instructions the format adds behind the scenes.
## Formats [#formats]
| Format | What you get | Output | Default |
| ------------------ | ---------------------------------------------------------- | ------ | -------------- |
| **UGC** | Face-to-camera, creator-style clip with selfie framing | Video | 9:16, 12s |
| **Product Review** | Hands-on demo: hold, use, and react to the product | Video | 9:16, 12s |
| **Unboxing** | First-touch reveal: the package opens, the product emerges | Video | 9:16, 10s |
| **Tutorial** | Step-by-step how-to-use with clear instructional beats | Video | 9:16, 15s |
| **TV Spot** | Cinematic commercial framing with polished camera language | Video | 16:9, 10s |
| **Hyper Motion** | CGI-style product motion; pure spectacle, no presenter | Video | 9:16, 6s, 720p |
| **Product Shots** | A pack of 4 still images: closeup, lifestyle, editorial | Images | 1:1 |
Every format offers 9:16, 1:1, and 16:9. Switching format resets the pills to that format's defaults.
* **Presenter formats** (UGC, Product Review, Unboxing, Tutorial, TV Spot) feature a person. They need a starting image: attach a product photo or your own presenter photo, or choose a built-in presenter. Until you do, **Launch** stays disabled with *"Add a product photo or choose a presenter to generate this format."*
* **Hyper Motion** has no presenter. Choose 4 to 15 seconds at 480p, 720p, or 1080p.
* **Product Shots** always returns 4 images per launch.
Video ads are silent: there's no voiceover or scripted dialogue.
## Attachments [#attachments]
Three buttons sit next to the prompt box. Each holds one image (JPEG, PNG, or WebP); uploading again replaces it.
| Button | Use it for |
| ------------- | -------------------------------------------------------------------------------------------------------- |
| **Product** | The item you're advertising. It's reproduced as faithfully as possible: shape, label, colours, branding. |
| **Character** | Your own presenter photo. It replaces any built-in presenter you picked. |
| **Reference** | Style or scene inspiration: look, palette, energy. |
Attached images show as chips above the prompt; click the **×** on a chip to remove it. For a clean result, send a clear, well-lit product photo on a plain background.
## Presenters [#presenters]
**Select Avatar** opens **Avatar presets**, a grid of 12 built-in presenters. Click one to select it; click it again, or use **Clear selection**, to go without.
| Presenter | Style |
| --------- | ------------------ |
| Selam | warm & trusted |
| Dawit | confident expert |
| Maya | energetic Gen-Z |
| Jonas | friendly everyman |
| Adaeze | polished lifestyle |
| Kenji | tech reviewer |
| Lucia | fashion-forward |
| Omar | fitness coach |
| Linh | calm & mindful |
| Priya | beauty creator |
| Marcus | bold hype |
| Elsa | mature & credible |
Presenters apply to the presenter formats only. To feature a specific person, such as yourself or a brand ambassador, upload their photo with **Character** instead.
## Credits [#credits]
The estimate next to **Launch** (shown as **cr**) updates as you change format, aspect ratio, resolution, and duration. That amount is held when you launch.
* **Presenter formats** are billed per second of video, so longer clips cost proportionally more. Check the live estimate for the exact figure.
* **Hyper Motion** costs \~281, \~605, or \~1,361 credits per second at 480p, 720p, or 1080p: from about 1,125 credits (4 seconds at 480p) to about 20,412 (15 seconds at 1080p). The default, 6 seconds at 720p, is about 3,629.
* **Product Shots** is a pack of 4 images at 80 credits each, plus about 1 credit per reference image. If fewer images come back, you're charged only for those delivered.
Failed generations are refunded automatically. To stop a running generation, click the **×** next to it in the queue panel (bottom right); its credits are refunded. Marketing Studio shares your plan's generation limit with the other tools.
## Tips [#tips]
* **Lead with the product photo.** It's the strongest signal in every format. Use the **Reference** slot for mood, not for the product itself.
* **Describe the scene, not the camera.** The format already handles framing and pacing. Spend your words on setting, action, and tone: *"on a sunny kitchen counter, morning light, upbeat"*.
* **Keep it short.** The box holds up to 2,000 characters, and less on formats that reserve part of the prompt for their own instructions. A counter appears as you near the limit, and the box stops accepting text once you reach it.
* **Match the ratio to the placement.** 9:16 for Reels, TikTok, and Shorts; 1:1 for feeds; 16:9 for YouTube and TV-style placements.
* **Prototype short.** Try a shorter duration or 480p to test an idea, then launch the final version at full length.
# Moodboard (/docs/tools/moodboard)
The Moodboard is an infinite, zoomable canvas for building visual boards: character looks, location references, tone and palette studies, or a whole film's visual bible. Boards are saved as sessions you can switch between; the most recent opens automatically.
## Working the canvas [#working-the-canvas]
* **Add images**: **Add Image** lets you upload from disk or pick from the built-in **Inspirations** library.
* **Generate**: prompt from the bottom bar and images land directly on the canvas, using the same model catalog as the [Image Generator](/docs/tools/image-generator).
* **Prompt library**: the **Select a Prompt** button on the left toolbar drops a curated prompt into the prompt box.
* **Text**: add labels and notes with font, size, color, and style controls.
* **Arrange**: drag, resize, multi-select with a drag-rectangle, reorder, delete. Zoom from 0.1× to 5×; hold **Space** and drag to pan.
* **Crop**: trim any image to a drawn rectangle.
* **Upscale**: enhance a selected image in place (20 credits).
* **Download**: one image or everything selected.
* **Export as PNG**: in the board's title menu, saves the whole board as a single image.
Uploading, arranging, text, cropping, and exporting are free. Only generation, upscaling, and training cost credits.
## References from the canvas [#references-from-the-canvas]
When your selected model supports reference images, any images you have **selected on the canvas** are used as references automatically. They appear as chips above the prompt. This makes iterating on a look natural: select the frames that work, prompt the next variation.
## Custom model training (LoRA) [#custom-model-training-lora]
Train a custom model on your own image pairs, for example a consistent character, product, or style edit.
1. Click **Train LoRA** on the bottom bar. The **Train Model** dialog opens. It's the same dialog on this tool and on an AI Studio project's Moodboard.
2. Pick the **Model** (**Qwen Image Edit Plus — LoRA trainer**) and enter a **Training Name**.
3. Upload **Training Data (ZIP)** of image pairs: `photo_start.jpg` + `photo_end.jpg` (before/after). Add extra references per pair as `photo_start2.jpg`, `photo_end2.jpg`, and so on, and set **Reference Image Count** (1 to 3) to the number of references per pair.
4. Give each pair its edit instruction: a caption file such as `photo.txt`, or type one instruction in the optional **Default caption** field to cover every pair without a file. Pairs with neither fall back to a generic caption, so give every pair one or the other.
5. Click **Start Training** (80 credits). It runs in the background; you can close the dialog.
When training completes, it's listed under **Select Trained LoRAs**. A full walkthrough is in [Consistent characters](/docs/guides/consistent-characters).
## Generating with your LoRAs [#generating-with-your-loras]
1. Click **Select LoRA** on the bottom bar and pick up to 3 trained LoRAs. They appear under **Selected LoRAs**.
2. Select at least one image on the canvas. LoRAs run on **Qwen Image Edit Plus LoRA**, an edit model, so your selected images are the **Base images** it edits. Generating without one is blocked.
3. Prompt and generate as usual.
While LoRAs are selected, the model picker shows **Qwen LoRA** with its price and is locked. The price per generation is also shown under **Selected LoRAs**. Remove the LoRAs to pick another model again.
## Sharing [#sharing]
Boards are private by default. In a workspace, the board owner can open **Share** from the board's title menu to make a board visible to the whole workspace or to specific members.
Besides this standalone tool, every AI Studio project has its own Moodboard for cast and reference images, opened with the **+** on the Characters rail. Same canvas, scoped to that project.
# Music Generator (/docs/tools/music-generator)
The Music Generator composes original audio. Controls on the left, your history on the right with inline players, download, and delete.
## Generate [#generate]
Provide at least one of:
* **Prompt** — describe the music: mood, instrumentation, tempo, scene.
* **Tags** — comma-separated genre and style tags, e.g. `pop, upbeat, electronic`.
* **Lyrics** — optional; must be accompanied by a prompt or tags.
Options:
| Option | Values |
| --------------- | -------------------------------------------------------------------- |
| Instrumental | On / off |
| Prompt Strength | 0.5 – 5.0 (default 2.0) — how strictly the music follows your prompt |
| Output Format | MP3 (default), WAV, FLAC, OGG, M4A |
## Extend [#extend]
Take any completed song from your history and add to it:
* **Direction** — append to the end, or prepend to the start.
* **Duration** — 1 to 85 seconds of new music (default 30).
* The same prompt / tags / lyrics / strength controls shape the new section.
## Credits [#credits]
A flat **120 credits per generation**, for both Generate and Extend. The cost is shown before you submit.
For a continuous music bed composed to match your film's segment boundaries, see the Score card in [Auteur](/docs/tools/auteur).
# Video Generator (/docs/tools/video-generator)
The Video Generator shares the Image Generator's chat-style layout: prompt bar at the bottom, results above, history on scroll. Each clip previews inline. Click one to open it full-screen, where you can **Download** it, delete it, or **Publish to Community**.
It opens on **Seedance 2.5** with **Auto** duration, 720p and audio on. Pick another model from the model pill; a Community **Remix** opens on the post's model instead.
## Generating [#generating]
* **Prompt**: a live character counter tracks the model's limit. It turns amber near the limit and blocks generation over it.
* **Model**: the pill shows a live `~credits` estimate that updates as you change options. Models marked **Needs image** in the picker require a reference image.
* **Aspect ratio** and **duration**: pills on the prompt bar. The aspect-ratio pill lists only the ratios the selected model supports. If you switch to a model that doesn't offer your current ratio, it falls back to that model's default.
* **Settings**: **Resolution** (only on models that offer a choice) plus each model's own options (audio, end image, negative prompt, and so on).
* **Reference image**: on most models, the image button (or dropping an image on the prompt bar) sets the starting frame. On the Seedance 2.x models the image button adds a reference instead; drop an image on the bar to set the starting frame.
## Models [#models]
In picker order:
| Model | Duration | Resolution | Notes |
| --------------------------------------------- | ----------------- | ------------------- | ------------------------------------------------------------------------------------------ |
| Seedance 2.5 | Auto or 4–30s | 480p / 720p | Longest clips; image, video and audio references; audio on by default |
| Seedance 2.0 | Auto or 4–15s | 480p / 720p / 1080p | Image, video and audio references; audio; optional End Image |
| Seedance 2.0 Fast | Auto or 4–15s | 480p / 720p | Cheaper, faster |
| Google Omni | 3–10s | Fixed by provider | Video with synchronized audio; an attached image is used as a reference, not a start frame |
| Seedance 2.0 Mini | Auto or 4–15s | 480p / 720p | Cheapest of the 2.0 family |
| Seedance 1.5 | 4–12s | 480p / 720p | Audio toggle |
| Veo 3.1 | 4 / 6 / 8s | 720p / 1080p | **Generate Audio** switch (on by default); 16:9 or 9:16 only |
| Wan 2.5 | 5 / 10s | 480p / 720p / 1080p | |
| Kling 2.6 | 5 / 10s | | Audio, negative prompt |
| Kling 3.0 Standard / Pro / 4K: Text to Video | 3–15s | 720p / 1080p / 4K | Resolution is fixed by tier |
| Kling 3.0 Standard / Pro / 4K: Image to Video | 3–15s | 720p / 1080p / 4K | Needs image; optional End Image |
| Kling 3.0 Standard / Pro: Reference to Video | 3–15s | 720p / 1080p | Up to 4 reference images or elements |
| Kling 3.0 Motion Control | From source video | 720p | Needs image; a reference video drives its motion |
Seedance 2.5 durations are Auto, 4, 5, 6, 8, 10, 12, 15, 20, 25, or 30 seconds. The Seedance 2.0 models offer Auto, 4, 5, 6, 8, 10, 12, or 15 seconds.
### Seedance references [#seedance-references]
The Seedance 2.x models (2.5, 2.0, Fast, and Mini) show three attach buttons on the prompt bar: image, video, and audio. On top of the starting frame, you can attach these references (JPEG, PNG, WebP or GIF images; MP4, MOV or WebM videos; WAV or MP3 audio):
| Model | Images | Videos | Audio files |
| -------------------------------- | ------ | ------ | ----------- |
| Seedance 2.5 | 29 | 10 | 10 |
| Seedance 2.0, 2.0 Fast, 2.0 Mini | 8 | 3 | 3 |
Audio needs at least one image or video alongside it.
With references attached, type `@` in the prompt to insert a reference token: `@Image1`, `@Image2`, `@Video1`, `@Audio1`, and so on, numbered in the order you attached them. The starting frame doesn't get a token.
### Veo 3.1 audio [#veo-31-audio]
Veo 3.1 has a **Generate Audio** switch, on by default. With audio, a clip costs 300 credits per second; switch it off for a silent clip at 200 credits per second.
On a shot's **Generate Video** page in [AI Studio](/docs/tools/ai-studio), adding Veo 3.1 **Reference Images** switches it to reference-to-video (unless the shot also has a **Last Frame**). That mode always renders 8 seconds, whatever length you pick, and costs **6,400 credits** with audio or **3,200** without. The estimate on the model shows this price.
### Kling 3.0 options [#kling-30-options]
* **Generate Audio**: off by default. **Voice / Dialogue Control** makes spoken lines follow your prompt and moves the clip to a higher billing tier.
* **Multi-shot Prompts**: up to 6 shots, each with its own prompt and duration. The total duration is what's billed.
* **Elements**: up to 4 reference images or videos you can cite in the prompt as `@Element1`, `@Element2`…
* **Negative Prompt**, **CFG Scale** (0 = loose, 1 = strict), **Shot Type**, and an optional **End Image** on Image to Video.
On Reference to Video, **Reference Images** and **Elements** share one cap of 4.
### Kling 3.0 Motion Control [#kling-30-motion-control]
Give it a still image plus a **Reference Video**, and the video's motion is transferred to your image's character. **Character Orientation** sets the source video cap: *Image* (up to 10s) or *Video* (up to 30s). **Keep Reference Audio** preserves the reference video's sound, and a **Facial Element** image helps preserve identity when orientation is *Video*.
## "Auto" duration billing [#auto-duration-billing]
On the Seedance 2.x models, **Auto** lets the model choose the clip length. The estimate and the credit hold assume the model's maximum: 15 seconds for the Seedance 2.0 models and 30 seconds for Seedance 2.5. When the clip completes, you're charged for the length actually produced and the difference is refunded.
With a fixed duration, the estimate you see is what you're charged.
## Credits [#credits]
Video is billed per second and scales with model, resolution, and audio options: from **\~23 credits/second** (Seedance 1.5 at 480p without audio) to **\~1,361 credits/second** (Seedance 2.0 at 1080p).
| Model | Credits/second |
| --------------------------------------- | ---------------------------------------------------------- |
| Seedance 2.5 | \~208 (480p) / \~462 (720p) |
| Seedance 2.0 | \~281 (480p) / \~605 (720p) / \~1,361 (1080p) |
| Seedance 2.0 Fast | \~225 (480p) / \~484 (720p) |
| Seedance 2.0 Mini | \~144 (480p) / \~309 (720p) |
| Seedance 1.5 | \~46 (480p) / \~104 (720p) with audio; \~23 / \~52 without |
| Google Omni | 250 (from text) / 260 (with references) |
| Veo 3.1 | 200 (silent) / 300 (audio) |
| Wan 2.5 | 100 / 200 / 300 at 480p / 720p / 1080p |
| Kling 2.6 | 140 (silent) / 280 (audio) |
| Kling 3.0 Standard | 168 / 252 (audio) / 308 (audio + voice) |
| Kling 3.0 Pro | 224 / 336 (audio) / 392 (audio + voice) |
| Kling 3.0 4K | 448 / 672 (audio) / 784 (audio + voice) |
| Kling 3.0 Standard – Reference to Video | 168 / 210 (audio) |
| Kling 3.0 Pro – Reference to Video | 224 / 280 (audio) |
| Kling 3.0 Motion Control | 336 |
Attaching a reference video lowers the per-second rate on the Seedance 2.x models. The live estimate on the model pill is the number to trust. More in [Credit costs](/docs/guides/credit-costs).
Closing the tab doesn't cancel a render. It finishes in the background and lands in your history. To stop one, click the **×** next to it in the queue panel (bottom right): the generation is cancelled and its credits are refunded. A generation that makes no progress for 30 minutes is failed automatically and its credits are refunded.