AI Video Generation
Render video clips from a shot description or starting image.
Overview
Sovran offers several models for video clips from text prompts or starting images. The model list includes Kling 4.0, MiniMax H3, Wan 3.0, Veo 3.1 Lite, Happy Horse 1.1, and Gemini Omni 1.2 Flash. Available models depend on the active catalog. Use AI footage for:
- Hooks — Eye-catching 3-5 second openers.
- B-roll — Supplementary footage that enhances your message.
- Product demos — AI-generated scenes showing your product in context.
Generating a Video
- Open AI footage under Prepare media in your project.
- Use Prompt or image and enter Describe the shot. Include the subject, action, camera movement, and final frame.
- Choose an available model. Add a starting image when the model requires or supports one.
- Check the available format, duration, and output settings. These depend on the model.
- Read the credit cost beside Render clip, then select it.
Follow progress under Recent clips. The time depends on the provider. When there are no earlier clips, this list shows No clips yet.
Select a recent clip to inspect it. A finished saved clip has View in Assets. If saving failed, use Retry save. Remove from Recent clips hides a saved clip from this list and keeps the asset in your project. Use the offered Undo action to restore the list item.
Model settings
Check the duration, format, and resolution controls after you choose a model. Models have different settings:
| Model | Clip length | Audio |
|---|---|---|
| Kling 4.0 | 3–30 seconds | Always included |
| MiniMax H3 | 4–15 seconds | Always included |
| Wan 3.0 | 2–30 seconds | Always included |
| Veo 3.1 Lite | 4, 6, or 8 seconds | Can be turned off |
| Happy Horse 1.1 | 3–15 seconds | Always included |
| Gemini Omni 1.2 Flash | 3–15 seconds | Always included |
Models with audio always included have no audio switch. Check the cost beside Render clip before each request. A new render uses credits. Retrying the save of an existing result does not create another clip.
Gemini Omni 1.2 Flash accepts a text prompt or a starting image. Choose 720p or 1080p. You can set the duration in steps of one second.
Tips for Better Results
- Be specific — "A woman in a white kitchen opening a blue box and smiling" works better than "someone opening a package."
- Include style cues — "cinematic lighting," "UGC-style handheld," "flat lay top-down" help guide the visual style.
- Iterate — Generate multiple versions and compare. Small prompt tweaks can significantly change the output.
Using Generated Videos
Generated clips are saved to your project's asset library. From there, you can:
- Tag them and add to your creative library.
- Edit them in the video editor (add text, captions, music).
- Assign them to sequence blocks for combinatorial testing.
Talking videos
Use AI Avatar to create a talking video from a saved photo and an approved custom voiceover.