> For the complete documentation index, see [llms.txt](https://docs.avastudio.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.avastudio.com/models-and-credits/video-models.md).

# Video models

Different video models are better for different jobs: speed, cinematic quality, reference consistency, motion control, or audio support.

| Model               | Best for                                                         | Typical use                                            |
| ------------------- | ---------------------------------------------------------------- | ------------------------------------------------------ |
| Kling 3.0           | Cinematic motion and prompt adherence.                           | Text-to-video, image-to-video, motion control, extend. |
| Kling O3            | Reference-heavy and multi-shot video.                            | Actor/product consistency, video edit/reference.       |
| Seedance 2.0        | Balanced text, image, and reference-to-video.                    | General UGC scene generation.                          |
| Seedance 2.0 Fast   | Lower latency and lower per-second cost.                         | Fast iteration and concept testing.                    |
| Veo 3.1             | Advanced cinematic generation with audio and resolution options. | High-value scenes and richer motion.                   |
| Veo 3.1 Fast        | Faster Veo-style outputs.                                        | Fast cinematic iteration.                              |
| Sora 2              | State-of-the-art video generation.                               | Prompt-led scenes in supported aspect ratios.          |
| Hailuo Standard 2.3 | High-quality video generation.                                   | General text or image-to-video scenes.                 |

For a simple from-scratch video path, start in Studio Mode, write one focused scene prompt, choose Seedance 2.0 or Seedance 2.0 Fast when available, generate 2 to 4 variants, then set the best output. Move to more specialized models when you need stronger reference control, motion control, or cinematic quality.

Model availability, cost, duration, and resolution can change. Always check the model menu and generation cost before starting important batches.
