Limits and moderation
AI Image Editor is in public beta, so costs and limits may change.
Three separate controls can stop an AI request, and every prompt goes through content moderation. For what each action costs, see AI Operations billing.
Limits
Each control returns its own error, described in Error handling.
Workspace request limit
- Each generate, edit, or upscale is one request, whatever its size.
- A request counts for 24 hours after it’s made, so the limit never resets at a fixed time.
- By default, the limit is the plan maximum: 50 requests on Pro, 200 on Business, and the contract value on custom plans. The workspace owner can lower it in Billing > Usage control.
Token operations
- Each AI request spends 1 from the token’s
uc.restrictions.limits.operationsclaim, whatever its AI Operations cost. - Uploads signed with the same token, for example through the File Uploader plugin, spend from the same limit: one per uploaded file. Set the limit to cover both.
- To issue tokens, see Signed AI image generation and Upload API authentication.
Project AI toggle
- AI is on for every project by default. Turn it off in Project settings > AI > Enable for this project.
- Requests also return
derivative_disabledwhen the plan doesn’t include AI Image Editor.
Moderation
- Every prompt goes through content moderation, including requests sent straight to the API.
- A blocked prompt fails with
content_moderated, and the editor shows the user a message about the content filter. See Error handling. - Blocked prompts aren’t billed. See What isn’t billed.
- To remove free-text prompts entirely, turn on presets-only mode: users pick from your presets and can’t type a prompt.