Flow provides a range of features powered by various Veo models, in addition to Imagen 4 and Gemini. Different models support different features.
- Veo 2 is our default model and supports the full Flow experience. We recommend it for most uses.
- Veo 3.1 is our newest model and will offer the best quality and experimental audio.
Tip: Before generating, in the prompt box settings, check the active model and the latest credit costs.
Flow models & supported features
Important: If you select a feature not yet supported by Veo 3.1, Flow will notify you and automatically switch to Veo 2 to complete the request.
| Feature | Veo 2 - Fast | Veo 2 - Quality | Veo 3.1 - Fast | Veo 3.1 - Quality |
|---|---|---|---|---|
| Text to Video | Landscape only | Landscape only | Both aspect ratios | Both aspect ratios |
| Frames to Video: First | Landscape only | Landscape only | Both aspect ratios | Both aspect ratios |
| Frames to Video: First + last |
Landscape only | No | Both aspect ratios | Both aspect ratios |
| Camera Control* | Landscape only | No | No | No |
| Extend | Landscape only | No | Landscape only | Landscape only |
| Jump To | Landscape only** | Landscape only | No | Landscape only |
| Ingredients to Video | No | No | Landscape only | No |
| Insert or remove an object | Both aspect ratios | |||
* Requires first frame or first and last frame
** For best results, use Quality models, not fast.
Learn about experimental features
Audio Generation
Important: This is an experimental feature only available on the Veo 3.1 model.
You can now add sound effects and background noise, as well as speech in some cases.
- Add your guidance to the prompt itself. For example:
the sound of an ice cream truck can be heard in the backgroundthe captain turns towards the ocean and says ”we set sail at daybreak!”
- Sometimes Veo can produce low-quality audio, in which case your video will not be generated. Instead, your credits will be refunded. If you get this error, you can try again or try another prompt.
- Speech generation performs better with longer text transcripts in your prompt.
- We’re working on improving the following known issues:
- Speech is currently muted on generations depicting minors.
- Generated speech may incorrectly trigger on-screen subtitles.
Send us your feedback on this feature so we can iterate on it.