Issue evidence · Serve · Comfy-Org/ComfyUI

[Integration] Document/register vLLM-Omni as a remote MiniMax-H3 backend

GitHub issue: https://github.com/Comfy-Org/ComfyUI/issues/15281

Not run

The proposal asks ComfyUI to document or register the vLLM-Omni custom node as an external MiniMax-H3 video backend.

What Badgr ran

Nothing was run: this is a documentation and integration request, not a failure, and a MiniMax-H3 serve needs vLLM-Omni with the model weights on a large GPU.

What came back

  • The proposal lists T2VA, FL2VA and Ref2VA routing and a shared /v1/videos service.

A test would start vLLM-Omni's MiniMax-H3 service, call /v1/videos once, and pass its output through the custom node.

Checked 2026-10-14 in Badgr’s local development environment.

← All issue evidence