Flow Insert: add an object to a video, and the Sume API equivalent

To add an object to an AI video, Flow has an Insert tool, and on Sume you send the clip to Gemini Omni Flash 1.1 as a video_url edit with a prompt such as "Add a red umbrella leaning on the wall. Keep everything else the same." Insert is a Flow feature with no API on the page; the Sume route is an API job.

Google's Flow launch post for Veo 3.1 says Insert lets you "introduce anything you can imagine, from realistic details to fantastical creatures" while Flow handles shadows and scene lighting. The same post listed Remove as coming soon; I did not confirm on a current page whether it has shipped, so this page covers only adding. Read 2026-10-02.

What does Flow Insert do, and what does it cost?

Flow edits a generated or uploaded video. Google's credits page lists video editing at 40 credits per generation, regardless of clip length or resolution. Daily credits are 50 and do not roll over, so one edit fits inside a free day but a second does not.

The credits page does not say which editing tool each edit uses, so treat 40 as the figure for any edit, Insert included, only if you see it quoted that way in the Flow interface.

How do I add an object with the Sume API?

Send the source clip to the Video Router with model: "gemini-omni-flash-1.1" and a video_url. The docs say the prompt describes the edit, resolution is optional (default 720p), and aspect_ratio and duration are not accepted in edit mode. Native audio is always on, and billing is provider list times 1.25 per output second by resolution.

Describe the object, where it goes, and what stays unchanged.

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: insert-umbrella-001" \
  -d '{
    "model": "gemini-omni-flash-1.1",
    "prompt": "Add a red umbrella leaning against the left wall, lit by the same window light. Keep everything else the same.",
    "video_url": "https://example.com/street.mp4",
    "resolution": "720p",
    "mode": "async"
  }'

Can I use a photo of the object?

Not in the same request. In Sume, video_url is the edit source and cannot be combined with image_url, end_image_url, or reference_*_urls. So Insert-by-photo is not available on Sume today: you describe the object in words. For a branded product that has to look exact, generate a new clip with the product photo as <IMAGE_REF_0> instead, as in the product-swap recipe.

What should I check on the result?

Watch for shadows, scale, and whether the rest of the shot stayed fixed. Edit mode rewrites the clip, so small background changes are normal; if you need frame-exact fidelity outside the new object, trim the edited section and place it back in a Timeline 1.0 render over the untouched original.

Fetch the finished job with GET /v1/jobs/:id/result, as described in Jobs and results.