Animate a mascot from one image: H3 start frame or Omni reference

A 10-second mascot clip made from one image costs $0.75 on MiniMax H3 at 768p when the image is the start frame, and $1.25 on Gemini Omni Flash 1.1 at 720p when the image is a reference. The difference is $0.50 a clip, and it buys a different behavior.

With a start frame, the first frame of the clip is your image. With a reference, the image tells the model who the character is and the scene is built around it. Prices come from the Sume price tool (read 2026-10-09).

Price of 10 seconds

Omni's 10 seconds is its ceiling. H3 can go to 15 seconds if the mascot needs a longer gag.

The Omni reference body

Refer to the image as <IMAGE_REF_0> in the prompt. One reference image with no first or end frame is a reference-to-video request, so you send only the list and the prompt.

{
  "model": "gemini-omni-flash-1.1",
  "prompt": "<IMAGE_REF_0> hops onto a skateboard and waves at the camera",
  "reference_image_urls": ["https://example.com/mascot.png"],
  "resolution": "720p",
  "duration": 10,
  "aspect_ratio": "9:16",
  "mode": "async"
}

Length changes the price more than the route

A shorter clip narrows the gap between the routes. The table prices 5, 8 and 10 seconds for both.

Limits and gotchas

Both routes take a prompt, so the cheapest test is one clip on each and a look at the result.

• Omni takes 3-10 seconds and 16:9 or 9:16 only. H3 takes 5-15 seconds and more shapes, including 1:1.
• Pick one: image_url is the start-frame request and reference_image_urls is the reference request.
• For H3, the body is the same as the Omni one without the token: image_url plus a prompt about the motion.
• Audio is native and always on for both.