Which video_url row to call: Omni edit, Recast or Motion Transfer
Three Sume rows accept video_url: gemini-omni-flash-1.1 edits by prompt, h3-max-recast swaps people, higgsfield-genjutsu moves motion. Limits and prices.

Only three Sume rows accept a source clip in video_url: gemini-omni-flash-1.1 edits a clip from a text prompt, h3-max-recast replaces the people in a clip with people from 1 to 4 photos, and higgsfield-genjutsu transfers motion onto reference images when its provider is configured. sume/auto never routes to the second or third, so you must name them.
The three rows side by side
Pick by what you want to change. Use the table's third column as the first filter, and the price column to estimate the bill. Per-second Sume rates are the fal list times 1.25, as documented in the Video Router page.
| Row | Use it to | Resolutions | Duration | Sume per second |
|---|---|---|---|---|
gemini-omni-flash-1.1 | Change objects, background, weather, style by prompt | 360p, 720p, 1080p, 4K | The source's length | $0.0375, $0.125, $0.1875, $0.375 |
h3-max-recast | Swap the main people for 1 to 4 photo references | 768p, 1080p | 5 to 30 s, no shot over 15 s | $0.375 (768p), $0.5625 (1080p) |
higgsfield-genjutsu | Motion Transfer: video_url + 1 to 8 reference images | 480p, 720p | 4 to 30 s | Read the live catalog |
Differences that decide it
Omni edit takes only a prompt and the clip, with native synced audio always on, and no reference images. Recast takes one photo per new person, keeps the source motion, camera, cuts and sound, and makes the prompt optional. Genjutsu appears in the catalog only when its provider is configured, so check GET /v1/video-router/models before you build on it.
Rule of thumb
If the change is an object, a colour or a setting, Omni edit is the cheaper per-second route. If the change is a person, Recast is the purpose-built route, at a higher per-second price. For anything else, generate fresh with a first frame instead of editing.
Sources
Related posts
More in Models
- Whistle's 86,174-utterance benchmark: test your own clips first
Cactus Whistle's card reports word error rates over 86,174 utterances. Score 20 of your own clips against a human transcript before choosing it over hosted STT.
- Whistle speech model: Apache 2.0, 16.9 MB, offline or Sume STT?
Cactus Whistle is a 16.9 MB Apache 2.0 speech-to-text model for seven languages. Run it on-device, or send the clip to Sume STT for other languages.
- An OpenRouter-compatible video API: sume/auto or a pinned model
Sume's POST /v1/videos follows OpenRouter's video generation API field for field. Let sume/auto pick the model, or pin a catalog id like seedance-2.5.
- Image generation API with reference images: POST /v1/images
Send a prompt plus public HTTPS reference images to Sume's POST /v1/images. Pin a catalog model or send sume/auto; the catalog lists each model's limits.
Written by Sume