Edit an existing clip with Gemini Omni Flash 1.1: video_url on Sume

To edit an existing video with Gemini Omni Flash 1.1 on Sume, call POST /v1/video-router/generate with model: gemini-omni-flash-1.1, a video_url pointing at the source clip, and a prompt that states the edit. Resolution is optional and defaults to 720p; you cannot send aspect_ratio or duration in edit mode.

Google describes Omni Flash as a video generation and editing model on its pricing page. Sume's Video Router docs turn the editing side into one field, and this post covers its rules.

The rules from the docs

video_url is the edit source, not a reference. These are the constraints stated in the Video Router docs, read 2026-10-09.

• video_url cannot be combined with image_url, end_image_url, or any reference_*_urls field.
• The prompt gives the instructions for the edit, for example 'Replace the bottle with an apple. Keep everything else the same.'
• resolution is optional, with a default of 720p; aspect_ratio and duration are not accepted.
• Native audio is always on for this model; the API rejects generate_audio: false.
• Sume bills the provider list price times 1.25 per output second, as a function of resolution.

A request

Use an Idempotency-Key header so a retry of the same request does not create a second job.

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: omni-edit-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"}'

What it might cost

Sume's 720p rate is $0.125 per output second. If your edit comes back as an 8-second 720p output, plan on 8 x $0.125 = $1.00, and 1080p would be 8 x $0.1875 = $1.50. The docs do not state how the output length is determined in edit mode, so treat those as planning figures and read the finished job's usage. Source clips should be reachable over public HTTPS.

Handling the result

The following figures use the same list prices as the tables above.

• Poll GET /v1/jobs/:id/status or the job's polling URL; completion returns a result with the video URL.
• Use an Idempotency-Key per edit, so a network retry does not create a second edit job and a second charge.
• Keep the original: edits create a new job and a new video, and do not modify the source.
• If the edit is wrong, rewrite the prompt so it states what to change and what to keep, as in the docs example.

Limits of this comparison

This post covers only what the docs state. It does not cover how long an edit takes, how faithfully the model keeps the parts you ask it to keep, or what happens when the source clip is longer than 10 seconds, because the Video Router docs do not say. Run a few edits on short sources first. At 720p a 3-second test edit is about 3 x $0.125 = $0.375 if the output matches the source length, which is cheap enough to try five variants of a prompt for under $2.