Limits and things to know
The rules and limits that apply across Studio: sign-in, credits, file sizes, outputs per run, what is not offered, retirement dates and licensing.
For the complete documentation index, see llms.txt. This page is also available as Markdown.
#Account and credits
- Every generation needs a signed-in account. There is no anonymous generation in Studio.
- Every model is available on every plan, including Free. Credits are the limit.
- One credit pool pays for Studio, the site generator and the editor's AI chat. See How credits work.
- If you do not have enough credits, the run does not start and nothing is charged.
- Generation requests are rate limited per address and per account. Paid plans get higher account limits.
#Per run
| Limit | Value |
|---|---|
| Outputs per image or video run | 1 to 4 |
| Reference or start image | One image, JPEG, PNG or WEBP, 20 MB at most |
| Images from a run that has a reference image | 1 |
| Default aspect ratio | 4:3 for images, 16:9 for video |
| Image aspect ratios | 16:9, 4:3, 1:1, 3:4, 9:16 |
| Video length | From 1 s to 30 s depending on the model; some models are fixed at 5 s or 8 s |
| Sound switch | On by default where a model has one (13 of 22 video models) |
| Video that never finishes | Marked failed after 30 minutes and refunded once |
#Files and sizes
| Where | Accepted | Limit |
|---|---|---|
| Reference image or start frame | JPEG, PNG or WEBP | 20 MB. The picture is scaled to 2,048 px on its longest side before it is sent. |
| Upscale | PNG, JPG or WebP | 20 MB. The result is 2x or 4x each side. |
| Remove Background | PNG, JPG or WebP | 20 MB. The result is a PNG. |
| Edit Image | An image | Scaled to 1,024 px on its longest side when it enters the editor. Expand adds 0 to 300 px on each side. |
| Lock-in agent message | Up to 6 images, 8,000 characters | Up to 4 images are read as vision input. |
| Spaces | Images and videos on a media node | A List node holds up to 10 text items and 12 media items. |
#Rate limits
Generation requests are limited per minute, once per account and once per network address. The account limit is the base number below on Free and is multiplied by 3 on Basic, 6 on Pro, 12 on Studio and 24 on Creator. The per-address limit is the base number for everyone, and whichever limit you reach first applies. Each output in a run counts as one request. When you reach a limit the card or page says "Too many requests".
| Action | Per account, Free | Per address |
|---|---|---|
| Image generation | 10 a minute | 10 a minute |
| Video generation | 4 a minute | 8 a minute |
| Edit Image | 8 a minute | 8 a minute |
| Upscale | 5 a minute | 5 a minute |
| Remove Background | 10 a minute | 10 a minute |
#Messages you may see
| Message | Meaning |
|---|---|
| Describe your image first / Describe your video first | The prompt is empty. Nothing was sent. |
| "<model> can't use a reference image — remove it or pick another model" | The chosen model is text only. Remove the image or switch model. Nothing was charged. |
| Add an image to animate | The chosen video model needs a start image. |
| <model> doesn't offer that resolution, length or aspect ratio | The option is not in the model's list. Nothing was charged. |
| Use a JPEG, PNG or WEBP image / That image is over 20 MB | The reference file was refused. |
| Not enough credits | Your balance cannot cover the run. Nothing was charged. |
| Your session expired | Sign in again. |
| "Connection lost — check your Library and credit balance before retrying" | The request may have been charged. Look in the Library before running again. |
| Model unavailable | The provider could not make the result. The charge is refunded. |
#Lock-in agent
- Up to 6 images and 8,000 characters of text per message.
- Up to 4 proposals per message, each with up to 4 outputs.
- Auto is free for your first 50 messages each UTC day, then 5 credits.
- A temporary chat is dropped when you leave and lasts at most 24 hours.
#What Studio does not offer
- More than one reference image, an end frame, video-to-video and video upscaling.
- An audio page. Audio is made inside the Lock-in agent only.
- 3D model generation, voice cloning, face swap and lip-sync.
- Posting your generations to a public feed.
- A public API or API keys.
#Model changes
- GPT Image 1.5 retires on 2026-12-01.
- A model's price can change. The price on the Generate button is always the live one.
#When a run fails
A failed generation is refunded automatically and the card shows the reason. The Studio runs the exact model and prompt you chose and never switches to another model. If the automatic refund fails, the message tells you to contact support. See Help and support.
#Your content and licensing
- You keep ownership rights in your content. The Terms grant you a license to use AI-generated content from your account for personal and commercial purposes.
- UOVA makes no representation about copyright ownership of AI content, and outputs are not guaranteed to be unique.
- Prohibited uses include deepfakes, non-consensual intimate imagery, impersonation of real individuals and misleading political content.
- You are responsible for labelling AI content where the law requires it.
See Data, security and IP and the Terms of service.
#Common questions
Check the file: JPEG, PNG or WEBP, 20 MB at most, one image. Also check the model: if it cannot use a reference, the Studio says so and asks you to remove it or pick another model.
See Report a bug.