Gemini Omni Flash 1.1 API: text, image, and reference video modes

To call Gemini Omni Flash 1.1 through the Sume API, send POST /v1/videos with model: "gemini-omni-flash-1.1". The one id covers text-to-video, image-to-video, and reference-to-video, chosen from the inputs you send, for 3–10 second clips at 360p, 720p, 1080p, or 4K with native synced audio.

The facts below come from Sume's Video generation and Video Router docs and from the model catalog the API serves, read on 2026-09-26. The submit-and-poll flow every video model shares is in An OpenRouter-compatible video API.

Which mode will my Gemini Omni Flash 1.1 request run in?

You never pick an endpoint: Sume routes the one catalog id by the shape of the request. On POST /v1/videos that works like this:

• A prompt alone runs text-to-video.
• frame_images with a first frame, and optionally a last frame, runs image-to-video. A last frame without a first frame is refused.
• input_references without frames run reference-to-video. A single reference image with no first or end frame still counts as a reference, not as an opening frame.
• Frames and references together run image-to-video, because frame_images takes precedence.

What are the limits?

A duration, resolution, or aspect_ratio outside these lists is refused at submit with 400 unsupported_capability; video generation API 400 errors explains the code.

How do I point the prompt at a reference?

Name each reference with a token: <IMAGE_REF_0> for the first image, <VIDEO_REF_0> for the first clip, counting from 0 in list order. Stills and clips go in input_references as image_url and video_url entries, at public HTTPS URLs, and a clip may run at most 3 seconds.

curl -X POST "https://api.sume.com/v1/videos" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: omni-flash-ref-001" \
  -d '{
    "model": "gemini-omni-flash-1.1",
    "prompt": "The mug from <IMAGE_REF_0> spins like the cup in <VIDEO_REF_0>",
    "input_references": [
      { "type": "image_url", "image_url": { "url": "https://example.com/mug.png" } },
      { "type": "video_url", "video_url": { "url": "https://example.com/spin.mp4" } }
    ],
    "resolution": "1080p",
    "aspect_ratio": "9:16",
    "duration": 6
  }'

Can I turn the audio off?

No. The model always produces native synced audio, so leave generate_audio out. Sending it as false to POST /v1/videos returns 400 unsupported_capability, and the message says to omit the field.

Can it edit an existing clip?

Yes, but not on POST /v1/videos: the docs expose its edit mode through the Video Router's video_url field. Edit a video with a prompt covers that request and its rules.

How is Gemini Omni Flash 1.1 billed?

Per output second, at a rate set by resolution: the model's pricing_skus hold one per-video-second-<resolution> rate each for 360p, 720p, 1080p, and 4K, and usage.cost on the poll response is the Sume billable amount. How those rates relate to the provider's list price and the 5.5% default agent fee is covered in List video generation models via API.