Runway lists Seedance 2.5 with video input: what Sume accepts
Runway says seedance2_5 takes text, image or video. On Sume, seedance-2.5 takes video as a reference, not an edit source. Which model edits a clip.

Runway's API model guide lists seedance2_5 as taking text, image or video input. On Sume, seedance-2.5 accepts video as a reference through input_references (video and audio references are honored by the Seedance 2.x models), but it has no video_url edit mode. To edit an existing clip, Sume offers gemini-omni-flash-1.1 through the video_url field.
What each Sume model does with a video (read 2026-10-04)
The rows come from the Sume docs and video catalog.
| Model | Video as reference | Video as edit source | Notes |
|---|---|---|---|
| seedance-2.5 | Yes | No | 4 to 30 s, text, image, end frame, image/video/audio references |
| wan-3.0 | Yes, up to 5 (15 s total) | No | 2 to 30 s |
| gemini-omni-flash-1.1 | Yes, up to 3, each up to 3 s | Yes, video_url | 3 to 10 s; edit cannot combine with image_url or references |
| higgsfield-genjutsu | Source video required | Motion transfer | Plus 1 to 8 reference images, 4 to 30 s |
| h3-max-recast | Source video required | Person swap | Plus 1 to 4 person photos, 5 to 30 s |
Reference, not edit
A reference video guides motion or style. The model renders a new clip. Send it as reference_video_urls on the Video Router. If you need the original frames kept and one thing changed, the edit path is the one that does that.
curl -X POST https://api.sume.com/v1/video-router/generate \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: seedance-ref-video-001" \
-d '{
"model": "seedance-2.5",
"prompt": "Match the camera move of the reference, new subject: a red bicycle",
"duration": 8,
"resolution": "720p",
"reference_video_urls": ["https://example.com/move.mp4"],
"mode": "async"
}'Edit an existing clip
Edit mode lives on the Video Router. Send the source in video_url with Gemini Omni Flash 1.1, and no image_url or reference_*_urls in the same request. The prompt describes the edit, and aspect_ratio and duration are not accepted.
curl -X POST https://api.sume.com/v1/video-router/generate \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: video-router-edit-001" \
-d '{
"model": "gemini-omni-flash-1.1",
"prompt": "Make the sky overcast. Keep everything else the same.",
"video_url": "https://example.com/source.mp4",
"resolution": "720p",
"mode": "async"
}'Inputs must be public
Media inputs must be public HTTPS URLs. Check supported_input_references and the Video Router capabilities for the model before you submit.
Sources
Related posts
More in Models
- Scribe v2, Realtime or Medical: which model for interview clips
ElevenLabs lists Scribe v2 with up to 32 speakers and 90+ languages, Realtime at about 150 ms, and Medical with 35% fewer errors. Which fits interview clips.
- Seedance 2.0, Fast and Mini next to 2.5: still callable on Sume
Sume still serves seedance-2, seedance-2-fast and seedance-2-mini by id next to seedance-2.5. 15-second ceiling, per-token prices, 4 s and 5 s job costs.
- Seedance 2.5 on BytePlus excludes the US: check access first
BytePlus lists Seedance 2.5 on ModelArk in supported markets excluding the United States. What that means for a US team, and how to read the Sume catalog.
- Seedance 2 mini, Seedance 2.5 or Wan 3.0 for holiday ad drafts?
At 480p a 5-second draft costs $0.3125 on Wan 3.0, $0.4394 on Seedance 2 mini and $1.3434 on Seedance 2.5. Here is when each is worth it.
Written by Sume