AI video generator from text and images

One creation form accepts your brief and images. Images can guide people, scenes or an ordered progression. Review the quote before generating; results are saved privately and can be attached to storyboard shots.

Open tool

Updated Oct 16, 2026

What you provide: A description and optional images: up to 9 with H3 or Seedance 2.0, or 30 with Seedance 2.5. Compatible completed videos from your account can also be referenced.
What you get: The original playable, downloadable MP4, a poster and a separate task record.

How to use it

Sign in with an account eligible for video creation and choose a model.
Describe the action and add images if needed. Explain their roles or the order in which the scene should progress.
Set duration, resolution, aspect ratio, audio and subtitles as supported, then review the quote. There is no separate creation-mode selector.
Generate, then preview or download the result. In the drama workspace, save it as a shot version; completed H3 768P results also offer a separate 2K version action.

Before you start

Access follows the Pro assistant entitlement and global switch. Gift credits and unverified or test payments do not unlock video.
H3: Images: PNG, JPEG or WebP, at most 16 MiB each; minimum short side 256 px, maximum long side 5760 px and aspect ratio 0.4–2.5.
If your brief explicitly assigns a start frame, its image controls the aspect ratio for that operation. Otherwise choose an aspect ratio. A 2K version retains its source format.
H3: Video references must be your completed H3 generation tasks: up to 3 clips totalling at most 15 seconds. Uploaded video and audio references are not supported.
H3: Reference seconds and images beyond the first 5 are billed separately. 2K regeneration is a paid task retaining the original inputs and duration.
Private videos are kept for 30 days. Uncertain purchases are not automatically repeated; unavailable or expired references cannot be used.
Images may describe multiple stages, not just endpoints. Large differences in framing, pose or scene can still produce unstable transitions; exact timing and pixel matches are not guaranteed.
Seedance 2.0 supports 4–15 integer seconds, 480p/720p/1080p/4K, synchronized audio and up to 9 images. Images must be at least 300 px on the short side, at most 16 MiB and 5760 px on the long side. This form does not accept uploaded video/audio or provide editing and extension operations.
Seedance has no plan discount. The estimate includes reference video time; final billing follows actual supplier usage and may be higher or lower.
Seedance 2.5 supports 4–30 whole seconds, 480p/720p/1080p, up to 30 images and 10 of your completed same-model clips (30 seconds total). Seedance 2.0 supports 4–15 seconds, optional 4K, 9 images and 3 same-model clips (15 seconds total). All videos use actual-usage billing without plan discounts.
Subtitles are guided by the prompt. An explicit setting takes priority; final text depends on the model output.
Manual optimization accepts up to 9 images and 7,000 characters within its input limit. Reduce the material or keep generating with your original prompt.

Common questions

Can I generate from a storyboard shot?

Yes. Open video generation from a shot to bring in its prompt, dialogue and linked images. Confirm the model, settings and quote. Image and video versions remain separate.

Do I need to choose text, image or reference mode?

No. Enter a description and optional images in one form. Explain each image’s role or the intended progression. Two images are not automatically treated as first and last frames.

Will my prompt be optimized automatically?

No. Only the Optimize prompt button starts optimization. Each successful request costs 0.1 Credits. Preview the result and choose whether to add its guidance to your original text. Generating a video does not trigger optimization.

How to use it

Common questions

Related tools

Practical guides

Model reference