Gemini Omni video edit on Sume: video_url price by output second

To edit an existing clip on Sume, send gemini-omni-flash-1.1 a video_url and a prompt that states the change. The resolution is optional and defaults to 720p, and the request takes no aspect_ratio or duration. At 720p Sume bills $0.125 per output second, so a 6-second edit costs $0.75.

There is no Sora remix endpoint to port to. The closest thing in the Sume docs is this edit mode, and it is available through the Video Router, not as a Sora-style remix call.

The request

The edit goes to POST /v1/video-router/generate. video_url is the edit source, not a reference, so it cannot be sent with image_url, end_image_url or the reference_*_urls fields. Native audio is always on for this model, and the API rejects generate_audio: false.

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: edit-clip-001" \
  -d '{
    "model": "gemini-omni-flash-1.1",
    "prompt": "Replace the bottle with an apple. Keep everything else the same.",
    "video_url": "https://example.com/clip.mp4",
    "resolution": "720p",
    "mode": "async"
  }'

Price by output second

The provider list is $0.03, $0.10, $0.15 and $0.30 per second at 360p, 720p, 1080p and 4K. Sume bills list x 1.25. The columns show a 3-, 6- and 10-second output; the model's window is 3 to 10 seconds.

Limits to plan around

• The request takes no duration or aspect_ratio, so the price you pay follows the length of the output. Test with a short source first.
• Because the call is async, poll the job or use a webhook as for any other video job, and send an Idempotency-Key so a retry returns the original job.
• Check the result for what the prompt said to keep. An edit that changes more than asked is a prompt problem, so state what must stay the same.

Edits, remixes and the Sora habit

A Sora remix took an existing video and a new instruction. The Sume edit mode has the same shape: one source video and a prompt. What differs is the surface, the model, and the price table. The docs do not describe an extend call for this model, so do not plan one. If you need a longer clip, generate on a model with a longer window, such as seedance-2.5 or wan-3.0, and compare the result.

Prompt the edit as a change list. Say what to replace and then say what to keep. The docs' own example does both in two short sentences, and that is a good pattern to copy.

Budget a test round

Run the same edit at 360p first. A 6-second test costs $0.225 at 360p against $0.75 at 720p, so three failed prompts at 360p cost less than one 720p render. Move to 720p only when the edit does what you asked.

When not to use edit mode

If the change is a trim, use a trim tool rather than a generation. The Sume docs list a video trim tool at $0.02 per job, which is a fraction of the cheapest edit render here. Edit mode is for changes a trim cannot make, such as replacing an object or changing a style.