Seedance 2.5 targeted editing: vendor claim vs what Sume accepts
Together says Seedance 2.5 edits specific moments after generation. Sume lists no video_url edit for it; only gemini-omni-flash-1.1 does. The reroll workaround.

On Sume you cannot edit a moment inside a finished Seedance 2.5 or Wan 3.0 clip. Together's page describes Seedance 2.5 as supporting timestamp-level control and targeted edits after generation, and Morphic describes Wan 3.0 as letting you reselect a time interval and regenerate only that part. Sume's model cards say the opposite for both ids: Seedance 2.5 has no video_url edit, and wan-3.0 has no edit mode. The Video Router also rejects video_url edit for any model but gemini-omni-flash-1.1.
| Claim | Source | On Sume |
|---|---|---|
| Targeted edits after generation (Seedance 2.5) | Together page | No video_url edit on seedance-2.5 |
| Regenerate one time range (Wan 3.0) | Morphic page | wan-3.0 modes: t2v, i2v, r2v; no edit |
| video_url edit | Sume docs and validation | gemini-omni-flash-1.1 only, 3 to 10 s |
Direct time ranges at generation time
The Together page also says you can direct narrative, camera and pacing for specific time ranges in the prompt. That is a prompt technique and needs no special parameter: write the prompt as a schedule, such as 0 to 8 s a slow push-in, 8 to 20 s the door opens, 20 to 30 s a cut to the street. Sume sends the prompt as written, and nothing in its docs promises that a model obeys time ranges, so treat it as a request that you verify.
Reroll one segment instead
If the middle of a 30 s clip is wrong, you can cut around it. Import the clip, use video-trim to take the good start and the good end as new MP4s (each trim is $0.02 per job at the published rate), generate the replacement as a short job, and join the three parts in a timeline. A 10 s replacement on Wan 3.0 at 720p is $1.25.
The weak points are the joins. A replacement made from a first frame pulled from the end of the good opening, with a last frame from the start of the good closing, gives the model both ends. wan-3.0 and seedance-2.5 both list first and last frame support, so that is available on either.
Cost of a segment reroll
Rerolling the middle ten seconds of a 30-second Wan 3.0 clip at 720p costs $1.25 for the new segment, plus $0.02 for each trim and the timeline render. Rerolling the whole clip costs $3.75. If the failure is at a known point, the segment route is cheaper, but only if the replacement is a believable match, so check the join frame by frame.
On Seedance 2.5 the same ten seconds at 720p is 10 s x 1280 x 720 x 24 / 1024 = 216,000 tokens, or $5.78 on Sume, against $17.34 for the full 30 s. The shortest Seedance 2.5 job is 4 s, so a replacement shorter than that has to be trimmed afterwards.
When to use the model that does edit
gemini-omni-flash-1.1 is the one model Sume documents for video edit through video_url, and it is a 3 to 10 second model. For a 6 s product clip that needs one change, it is the direct route. For a 30 s clip it is not, because the source would exceed that range.
Sources
Related posts
More in Comparisons
- Seedream 5.0 Lite or Nano Banana 2.1 for product shots on Sume
Seedream 5.0 Lite bills $0.04375 per image on Sume against $0.10 for Nano Banana 2.1 at 1K. Ratios, tiers, references and when the extra cost is worth paying.
- Pilot STT in 3 languages: language_code hint vs auto-detect, 30 clips
MAI-Transcribe-2-Streaming advertises 60 languages; here is a 30-clip pilot on Sume STT that compares a language_code hint with auto-detect for about 60 cents.
- Tavus PAL Maker, API, Enterprise vs the Sume avatar API: what matches
Tavus's site lists PAL Maker, a Developer API (CVI) and Enterprise. Which of them overlaps with Sume's avatar clips, and which does not. Pages read 2026-10-07.
- Text-to-video or image-to-video: when the prompt alone is enough
Use text-to-video when the look is open, and image-to-video when a frame, a face or a product must match. The Sume models that take each, and a decision rule.
Written by Sume