Veo 3.1 object insert and remove vs Sume's Omni video edit

Google's Veo page lists object insertion and removal among the Veo 3.1 features, next to scene extension, first and last frame transitions and a maximum clip length of 8 seconds at 1080p or 4K. Sume's v1 video catalog has no Veo row. If you need to add or remove an object in a finished clip on Sume, the supported route is the edit mode of Gemini Omni Flash 1.1: send the source clip as video_url on the Video Router and describe the change in the prompt.

What each side documents

The Veo page lists the 3.1 capabilities. Sume's video guide lists Veo 3.1 as not in the v1 catalog: there is no provider registry entry, fal builder or pricing constant for it.

Omni edit is documented in the Video Router guide. The video_url is the edit source, and only three rows accept that field: Omni, higgsfield-genjutsu and h3-max-recast. It cannot be combined with image_url, end_image_url or reference lists.

Run an edit on Sume

Edit uses the Video Router path, not the OpenRouter-shaped /v1/videos body. The source clip sets the output length and framing, so you send no aspect_ratio (that is a 400) and no real duration; a duration is only a reserve-estimate hint, default 8 seconds. generate_audio: false and reference_audio_urls are also rejected, because Omni always produces native synced audio.

Steps:

• Host the source clip at a public HTTPS URL.
• Post to /v1/video-router/generate with model set to gemini-omni-flash-1.1 and a video_url.
• Describe the change and say what to keep, for example: replace the bottle with an apple, keep everything else the same.
• Poll the shared job status and result routes and download the file.

curl -X POST https://api.sume.com/v1/video-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -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"}'

Cost and limits

Omni is billed per second by resolution: $0.0375 at 360p, $0.125 at 720p, $0.1875 at 1080p and $0.375 at 4K on Sume. Sume's docs flag that the provider does not return an output duration for edits, so the reserve is a hint rather than a probe. At the 720p rate, a 10-second edit would be about $1.25.

Edits can drift: faces, text and small props sometimes change. Review the result before it ships.

Writing a good edit prompt

Name the object, say what replaces it, and state what must stay unchanged. A prompt that lists only the new object invites the model to restyle the whole scene. Keep the source clip short and well lit, and run a draft at 360p first at $0.0375 a second to check that the change lands before paying the 1080p rate.

What Sume does not do

Sume does not offer Veo 3.1, does not expose scene extension, and does not promise that an edit leaves every other pixel untouched.