Which Sume video rows take video_url: Omni edit, Recast or Genjutsu
Three Sume rows take a source video_url: Gemini Omni edit ($0.125 a second at 720p), H3 Max Recast ($0.375 and up) and Genjutsu ($0.3975 and up). Pick by task.

Only three rows accept video_url as a source clip: gemini-omni-flash-1.1 (edit a video with a text instruction), h3-max-recast (swap the people in a video for the people in your photos) and higgsfield-genjutsu (transfer motion from the source video onto your images). Send video_url to any other row and it is rejected.
sume/auto never routes to Recast or Genjutsu, so for those two you must pin the id. These are Video Router fields (/v1/video-router/generate); they are not part of the OpenRouter-shaped /v1/videos body.
What each row does
- Omni edit:
video_urlplus a prompt (for example, replace the bottle with an apple). Noaspect_ratio, and the source clip sets the output length; adurationyou send is only the reserve estimate, default 8 seconds. - Recast:
video_urlplus 1 to 4reference_image_urls, one photo per new person. The output keeps the source motion, camera, cuts and sound. 5 to 30 seconds, no shot longer than 15 seconds. - Genjutsu:
video_urlplus 1 to 8 images, 4 to 30 seconds, 480p or 720p, listed only when its provider is configured. Noaspect_ratioorgenerate_audio.
Cost of a 10-second source
Sume billable, list times 1.25, rounded up per job, from the Video Router docs. fal's Recast list is $0.30 at 768p and $0.45 at 1080p per its page, read 2026-10-05.
| Row | Tier | Sume per second | Sume 10 s |
|---|---|---|---|
| gemini-omni-flash-1.1 (edit) | 720p | $0.125 | $1.25 |
| gemini-omni-flash-1.1 (edit) | 1080p | $0.1875 | $1.88 |
| h3-max-recast | 768p | $0.375 | $3.75 |
| h3-max-recast | 1080p | $0.5625 | $5.63 |
| higgsfield-genjutsu | 480p | $0.3975 | $3.98 |
| higgsfield-genjutsu | 720p | $0.85125 | $8.52 |
Pick by task
If the people stay and one object or style changes, use Omni edit; it is the cheapest by a wide margin. If the people must change but the shot, camera and sound stay, use Recast. If you have a performance video and want your own character to move the same way, use Genjutsu.
Because the source length sets the bill, trim the source before you submit. A 30-second source on Recast at 1080p is a deliberate cost, and the person-swap post puts it per minute of finished video.
Mistakes that return 400
Do not combine video_url with image_url, end_image_url or reference_*_urls on Omni. Do not send exactly zero images to Genjutsu or Recast; see the Genjutsu 400 post for the count rules. Do not send aspect_ratio with an edit.
Sources
Related posts
More in Developers
- Which URL goes into video-captions after a trim: the new artf_
After video-trim, caption the new video_url from the trim result, not the source render. It is a new artifact on media.sume.com and a public HTTPS URL.
- Which video model takes 30 seconds? Filter the Sume catalog in Python
Ask GET /v1/videos/models which models accept duration 30 at 1080p instead of hard-coding ids. A short Python filter, and why Omni and Kling 3 drop out.
- Which video models accept 30 seconds on Sume: read the catalog first
Seedance 2.5, Wan 3.0, h3-max-recast and Genjutsu list 30 s on the Sume Video Router; MiniMax H3 stops at 15 s. Read the catalog before you set duration.
- Why a Kling motion control job shows type avatar_image_to_video
Sume stores Kling motion control as type avatar_image_to_video with model kling/3.0/motion-control. The model id, not the type, tells it apart from Fabric.
Written by Sume