ComfyUI ID-V2V vs reference video and face swap on Sume

ID-V2V is a ComfyUI open-source model addition for identity-preserving video-to-video, described as working on "Wan 2.1 and VACE based models". Sume has no ID-V2V node. The nearest hosted options are video references on models that accept them, and the Beta face-swap endpoint when you have a ready avatar.

ComfyUI facts are from its changelog entry dated September 29, 2026, read 2026-10-01; the entry is one line and I read nothing else about ID-V2V. Sume facts are from Video generation and Face swap (Beta).

What does the ComfyUI changelog say?

Under "New Open-Source Model Support" for September 29, 2026 it lists "ID-V2V: Identity-preserving video-to-video on Wan 2.1 and VACE based models". It gives no node names, inputs or limits, so check the ComfyUI docs before building a workflow.

Which Sume path fits which job?

Is a video reference the same as ID-V2V?

Not by the docs. A video reference guides generation; the docs do not describe it as identity-preserving. Face swap is the identity-specific path, and it needs a ready Avatar 1.0 identity first. See Avatar Face Swap API (Beta) and which endpoint to choose.

How do I check a model before sending a video reference?

Call GET /v1/videos/models and read supported_input_references for the model id; the docs example lists image_url, video_url and audio_url. A model that does not list a type will not take it. If you already run ComfyUI partner nodes, MiniMax H3 in ComfyUI vs the hosted API compares the two routes for one model.