Create a video
Generate a video from a prompt, or animate one image. Every control on the Create Video page, what each setting does to the price, the sound switch and how long to wait.
For the complete documentation index, see llms.txt. This page is also available as Markdown.
#Before you start
- Sign in. Without an account the button reads Sign in to generate.
- Video costs more than images: 147 to 15,442 credits per generation, depending on the model, length and resolution. Read the price on the Generate button before you run.
#Generate a video
Step 1: Open Create video
Select Create video in the Create section of the sidebar, or go to /studio/create/video. To animate a photo, select Animate in the Tools grid (the Tools page calls it Animate a Photo), which opens the same page in image-to-video mode with the Start frame tile focused.
Step 2: Pick a model
The default is Kling 3.0. Three models only animate an image: Wan 2.6, Hailuo 2.3 Pro and Kling 2.5 Turbo.
Step 3: Write your prompt
Type in the Describe your video card: the shot, what moves and how the camera moves. With an empty prompt, Generate says "Describe your video first".
Step 4: Add a start frame if you want one
Select Start frame and choose a file. The tile reads Optional on text-and-image models and Required on the three image-only models. In Animate mode it reads "Add the photo to animate".
Step 5: Set the output
Choose aspect ratio, resolution and length where the model offers them. Where the model has sound, the Audio card shows On or Off with a switch. Then set Count.
Step 6: Check the price, then select Generate
A card with the model name and a running timer appears at once and updates in place when the video is ready.
#Every control
| Control | What it does | Default and limits |
|---|---|---|
| Model picker | Chooses the model. | Kling 3.0. 22 video models; see Video models. |
| Prompt | Describes the video. | Required for text-to-video. |
| Start frame | The image the video starts from. | Optional on text-and-image models, required on three image-only models. JPEG, PNG or WEBP, 20 MB at most, one image. |
| Aspect ratio | Shape of the video. | 16:9 where the model offers it. Veo 3.1, Veo 3.1 Fast, Veo 3.1 Lite, Luma Ray 2 and Wan 2.1 offer 16:9 and 9:16 only. Kling 3.0, Kling 3.0 Pro, Seedance 2.5 and Seedance 2.0 offer 16:9, 1:1 and 9:16. The rest offer all five ratios. |
| Resolution | Output quality, where a model has several. | Depends on the model, from 480p to 1080p. Veo 3.1 Fast offers 720p, 1080p and 4k, and Hailuo H3 offers 480P, 768P, 2K and 4K. Models with no choice show none. Higher resolution costs more. |
| Length | How long the clip is. | Depends on the model: Grok Imagine Video 1.5 offers 1 to 15 s, Kling 3.0 offers 3 to 15 s, Wan 3.0 and Wan 3.0 Prime offer 2 to 30 s. Seedance 2.5, Seedance 2.0 and Luma Ray 3.2 are fixed at 5 s, and Veo 3.1 at 8 s. Price scales with length. |
| Sound switch | Adds generated audio, where the model has it. | On by default in this page. 13 of the 22 models have it. |
| Count | How many videos one run makes. | 1 to 4. |
| Generate | Runs the generation. Shows the credit price. | Price includes sound when the switch is on and the model charges for it. Reads Sign in to generate when signed out. |
#Sound
The sound switch starts on wherever a model has one, so the first price you see includes sound. Turning it off lowers the price on the 6 models that charge extra for sound. See What actions cost for the multipliers.
| Sound | Models |
|---|---|
| Costs more | Google Veo 3.1 (x2), Seedance 1.5 Pro (x2), Veo 3.1 Fast, Kling 3.0 and Kling 3.0 Pro (x1.5), Veo 3.1 Lite (x1.67) |
| Included at no extra credit | Seedance 2.5, Seedance 2.0, Seedance 2.0 Mini, Seedance 2.0 Fast, Wan 3.0, Wan 3.0 Prime, FLUX 3 Video |
| No sound | Wan 2.6, Hailuo 2.3 Pro, Luma Ray 2, Kling 2.5 Turbo, Wan 2.1, Hailuo H3, Hailuo H3 Max, Grok Imagine Video 1.5, Luma Ray 3.2 |
#Image-to-video rules
- One start image only. There is no end frame, no multi-reference input and no video-to-video.
- A start image can change the price. Luma Ray 3.2 is 919 credits for a text-only 5 s clip and 276 with a start image. Grok Imagine Video 1.5 is 882 credits at its default settings and 992 with a start image. The button updates when you add the image.
- Image-only models follow the shape of your image, so the aspect-ratio choice does not apply. Kling 3.0 does the same when you give it a start image.
- If a model cannot use an image, the Studio says "<model> can't use a reference image — remove it or pick another model". If an image-only model has none, it says "Add an image to animate".
#While it runs
Videos are generated in the background. The card shows the model name and how long it has run. After 3 minutes it adds "Still rendering" and a note that the video will appear in your Library, so you can leave the page. A video that has not finished within 30 minutes is marked failed with "This video did not finish within 30 minutes", and its credits are refunded once. UOVA does not promise a render time, because it depends on the model.
#Limits
- 1 to 4 videos per run.
- No audio-only page. Speech, music and sound effects are made inside the Lock-in agent.
- Failures are refunded automatically and no substitute model runs. The failed card shows a short reason and Refunded with the amount.
- Video requests are limited per minute: 4 a minute per account on Free, multiplied by 3, 6, 12 and 24 on Basic, Pro, Studio and Creator, plus 8 a minute per network address that does not scale with your plan. Each video in a run counts as one request.
#Common questions
The sound switch is on by default. On Kling 3.0, for example, a 5 s clip is 772 credits with sound off and 1,158 with sound on.
The lowest price across all video settings is Grok Imagine Video 1.5 at 1 s and 480p, at 147 credits. Price depends on length and resolution, so compare models at the same settings in Model pricing.
No. Video upscaling and video-to-video are not available. The Studio tools work on images.