Video generation workspace
Select Video at the top left of the workspace. The left panel lets you choose a model, creation mode, prompt, reference images, duration, resolution, and aspect ratio. Results appear on the canvas for playback, download, and History access.
Current models
| Model | Best for | Maximum reference images |
|---|---|---|
| Seedance 2.0 Fast (default) | Fast all-purpose reference | 9 |
| Seedance 2.0 | High-quality all-purpose reference | 9 |
| MiniMax H3 | High-quality audio-video generation | 9 |
| MiniMax H3 Max | Fast audio-video generation | 9 |
| Veo 3.1 Fast | Fast generation | 3 |
The available duration, resolution, and aspect-ratio options depend on the selected model. Seedance 2.0 and Fast provide a Generate audio switch (on by default); H3 and H3 Max generate native audio-video output.
Two creation modes
- All-purpose reference (Omni Reference): Generate from text without images, or add reference images. The model draws on subjects, scenes, composition, and overall visual features while the prompt guides camera movement, action, lighting, and changes in the scene. Seedance and H3 models accept up to 9 references; Veo 3.1 Fast accepts up to 3.
- Start / End (First/Last Frame): Generate from text without images, add one image as the first frame, or add two images as the first and last frames. The model creates a continuous transition between them. See First and last frame control.
You can also choose a lens preset. The Generate button shows the required credits before submission; check the current interface for the live amount.