Seamless looping food video with AI: same first and last frame

Use one image as both the first and the last frame. Google's Gemini Omni 1.1 Flash post says specifying the starting and ending frames makes it suitable for seamless looping clips, and Sume's Video Router takes a start frame as image_url and an optional end_image_url for gemini-omni-flash-1.1, with 3 to 10 second clips.

Vendor wording is from the Google post, Sume's from the Video Router docs, both read 2026-10-01.

How do I set up the loop?

The Video Router docs say the image_to_video capability takes image_url plus an optional end_image_url. For a loop, send your food still as both, and write a prompt where the motion returns to rest: steam rising and fading, sauce settling, a slow turntable that ends where it began.

// POST /v1/video-router/generate
{
  "model": "gemini-omni-flash-1.1",
  "prompt": "Steam drifts up from the bowl and fades; the camera holds still.",
  "duration": 6,
  "image_url": "https://media.sume.com/your-dish.png",
  "end_image_url": "https://media.sume.com/your-dish.png"
}

What does the catalog confirm for this model?

Is the loop point guaranteed to be invisible?

No. Matching end frames makes the jump between the last and first frame small, but motion, lighting and steam can still disagree at the seam, so play the clip on repeat before publishing. Google also lists 360p previews for iteration; check Sume's catalog entry for which resolutions you can request. If the loop is too short, increase video length by looping explains repeating it in a player or timeline.