IAMCCS V2V Bus To Pose Transfer

Pose transfer takes a driver video - the motion you want - and a reference image - the person/thing you want moved - and re-enacts the motion. V2V Bus To Pose Transfer is the adapter that unpacks the IAMCCS media and control buses into exactly those two things, plus prompts, FPS, and the resolution, so a pose-transfer backend can be dropped into the same bus-driven graph as every other V2V engine.

Like its siblings in the BusTo family, this node doesn't generate anything. It translates: the source video on the media bus becomes the driver_video, the source image becomes the reference_image, and the control bus's plan becomes concrete prompt/dimension/FPS outputs. The optional result_image input lets you feed a previous render back through, which is how the backend chains generations against the same source.

Inputs

• media_bus, control_bus - required.
• result_image (optional) - a prior result to re-inject.

Outputs

• driver_video - the motion source frames (IMAGE).
• reference_image - who/what gets the motion applied.
• positive_prompt, negative_prompt - conditioning text.
• fps, generation_width, generation_height - geometry for the backend.
• output_prefix, source_audio, source_video_info - naming and context (VHS-style video info).
• report - status.

Install

Ships in IAMCCS-nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/IAMCCS/IAMCCS-nodes.git

or via ComfyUI Manager (search "IAMCCS"), restart.

How to think about it

The interesting contrast with V2VBusToLTX is that pose transfer has no segment math - there's no segment_index, no continuation_loops. This backend is a single-pass re-enactment, so the bus adapter reflects that: it hands over the whole driver, whole reference, and the plan's prompts. Where people get tripped up is expecting pose transfer to be an identity transfer - it transfers motion, so the character in the reference comes out animated by the driver, not cloned from it. If the driver and reference have wildly different framing, the results look off; keeping them visually compatible is on you. And since this node leans on the control bus's backend routing, it usually sits behind V2VActiveBackendSwitch alongside SCAIL, WanAnimate, and LTX - one bus, four engines.