Video generation: nine modes, material requirements and pricing
From text-to-video to face swap, upscale and extend: what each mode needs, how to choose duration and resolution, how credits are calculated, and the most common failure causes.
Model overview
- Seedance 2.0 / 2.0 Fast / 2.0 Mini: fast, natural motion, great for short shots.
- Seedance 2.5: supports video edit and extend; reference audio can be submitted on its own.
- Wan 2.1 / 2.2: the widest mode set: keyframes, motion control, lip sync, video face swap and upscale.
Durations, resolutions and credit prices differ per model; the panel updates when you switch.
Modes and required materials
| Mode | Needs | Prompt |
|---|---|---|
| Text-to-video | prompt only | required |
| Image-to-video | 1 first-frame image | required |
| Keyframes | first + last frame | required |
| Omni reference | reference images / videos / audio | required |
| Motion control | character image + motion video | optional |
| Lip sync | character image + audio | optional |
| Video face swap | face photo + source video | none |
| Upscale | source video | none |
| Edit / Extend | source video + change description | required |
Slot counts are limited and the total duration of reference audio / video has a cap; both are checked before submission.
Duration, resolution and credits
Credits = resolution tier price × duration units × count, shown live at the bottom-right.
- Validate with 480p / 5 s first, then go to 1080p once you are happy: it saves a lot of credits.
- Edit and extend are priced by the source video length.
- Nothing is charged with insufficient balance; failed tasks are refunded.
Hard requirements for reference materials
- Materials must be uploaded to the Library (or referenced from canvas nodes); external links are not accepted.
- Face photos for video face swap must be clear frontal photos of real people.
- Keep upscale sources at 720p or below; 1080p sources fail upstream.
- Overlong reference audio is rejected; you can generate a voice track with the audio node first.
- Copyrighted music may be rejected upstream; use your own or royalty-free audio.
While generating, and results
Videos usually take 1–5 minutes; upscale takes longer. Progress is visible in My projects and on canvas nodes, and you can close the page while waiting.
When done, preview online, download the mp4, or drag it onto a canvas as the source for face swap, upscale or extend.
Common failure causes
- Parameter not supported by this model: the old ratio / duration no longer applies after switching models; pick again.
- Provider not enabled / quota issue: the model is temporarily unavailable; switch models, credits are refunded.
- Upstream generation failed: usually an unclear face, an oversized source video or copyrighted audio; adjust per the requirements above and retry.
- Total duration over the limit: use fewer or shorter reference audio / video clips.