Gemini Omni task values mapped to Sume request fields

On Sume you do not send a task. gemini-omni-flash-1.1 is one catalog id that Sume routes by the shape of the request: prompt alone is text-to-video, image_url is image-to-video, reference lists are reference-to-video, and video_url is an edit. Extend has no Sume equivalent.

Google's guide lists five allowed task values. The mapping below uses Sume's Video router docs, read 2026-10-01.

What does Google's task field do?

The task parameter in video_config states the intended behavior explicitly. If it is not set, the model infers it from the prompt. Google recommends relying mainly on prompting and using task only when prompting does not work, because the field adds constraints. Allowed values: text_to_video, image_to_video, reference_to_video, edit, extend.

Which Sume field matches each task value?

Why is extend missing?

Sume's docs state that previous_interaction_id and extend are not exposed. To lengthen a clip on Sume you submit a new job per clip; see Gemini Omni extension to 40 seconds and the Sume alternative.

What does an edit request look like?

Send video_url with a prompt that describes the edit, to POST /v1/video-router/generate. The video_url is the edit source, not a reference, so it cannot be combined with image_url, end_image_url or reference_*_urls. Mixing them is the usual way to get a rejected request.

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: video-router-edit-001" \
  -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",
    "mode": "async"
  }'