Vidu movement_amplitude does nothing on Q2 and Q3; Sume uses prompts

Vidu's text-to-video reference, read 2026-10-01, describes movement_amplitude as a way to control how much objects move, with values auto, small, medium or large, and states: "This parameter does not take effect when using the q2 & q3 model". It works on the q1 model only.

Sume has no equivalent field and no Vidu model. The /v1/videos request body in the video docs has prompt, duration, resolution, aspect_ratio, frame_images, input_references, generate_audio, callback_url and a provider passthrough. Motion intensity is something you ask for in words.

How do I ask for small or large movement?

Put the amount and the subject of the motion in the prompt, and say what stays still. Compare "the camera holds still, only the steam moves" with "fast handheld orbit around the mug". If you need the motion copied from a clip, Sume's Higgsfield motion-transfer model takes one source video plus reference images, and minimax-h3 accepts reference videos.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: calm-motion-001" \
  -d '{
    "model": "seedance-2.5",
    "prompt": "Locked-off camera. Only the steam rises; the mug and table stay perfectly still.",
    "duration": 5,
    "resolution": "720p"
  }'

What does each side expose for motion?

Fields named in this post for controlling motion intensity.

Can I pass Vidu-style options through provider?

No. The model list returns allowed_passthrough_parameters, and in v1 that list is empty for every model, so a non-empty provider.options returns 400 unsupported_parameter. Unsupported fields fail loudly instead of being silently ignored.