Prompt expansion switch for Wan 3.0 and MiniMax H3 Max on Sume
Krea's H3 Max Turbo has prompt expansion modes. Sume's video model_params allowlist is empty in v1, so Wan 3.0 and H3 Max have no such switch.
There is no prompt expansion switch on Sume. The video request has a model_params object, but the OpenAPI description reads "Reserved for per-model knobs. Empty allowlist in v1 — omit or send {}." Sending a key inside it is not a way to toggle expansion.
What does Krea expose?
Krea's September 9, 2026 changelog says MiniMax H3 Max Turbo comes with "prompt expansion modes", alongside start and end frames and seed control. The snapshot read for this post does not list the mode names or any Wan 3.0 expansion option, so check Krea's API reference for those. They are Krea request fields; Sume does not mirror them.
What does Sume's schema say?
| Field | Rule |
|---|---|
model_params | Reserved for per-model knobs; empty allowlist in v1; omit or send {} |
model | wan-3.0 and minimax-h3-max are both in the catalog enum |
| Per-model limits | Read capabilities from GET /v1/video-router/models |
How do I control expansion myself?
Treat your prompt as the final prompt. If you want a model's own rewriting, you cannot ask for it here; if you want less guessing, write the shot out: subject, action, camera and setting. The Wan 3.0 prompt guide covers how to word it.
How will I know if a switch is added?
Check the Video Router docs and the schema for model_params; the v1 description says the allowlist is empty, so any later keys would be listed there. Until then, an unknown key is not part of the contract.
Sources
Related posts
More in Developers
- Remove filler words from a talking video with an API
Sume has no one-call filler remover. Transcribe with video inspect for word timings, then cut the clean ranges with video trim at $0.02 per job.
- Remove filler words from a video by API: cut at word timestamps
Descript's API lists Remove Filler Words as an Underlord edit. the Sume docs list no such op; here is how to cut um and uh yourself from words[] and a Timeline.
- OpenAI response_format json_schema on a Sume scheduled run
Sume accepts an OpenAI-shaped response_format as an alias for output_schema on schedule runs. Sending both returns 400, and the schema must be strict.
- Responses steering after a video job started: what Sume does
response.steer does not cancel started tools. A Sume video job past its start runs to completion; jobs_cancel returns 409 job_generation_already_started.
Written by Sume