PoseData to pose_keypoint

Every node in this pack speaks one dialect of pose data - POSE_KEYPOINT - and PoseDataToPoseKeypoint is the translator that lets poses from another family walk in the door. Specifically, it converts the POSEDATA that WanAnimate's "Pose and Face Detection" preprocess node produces into the pack's keypoint format. If you're coming from a WanAnimate workflow and want to proportion-edit its poses, this is the handoff node.

This matters more than it sounds. WanAnimate is a whole pose-driven animation ecosystem (and the pack's own lineage runs through it - the detector here is a port of kijai's WanVideo UniAnimate detector), but it stores poses as POSEDATA objects with their own internal structure. You can't just wire one into a ProportionChanger node and expect the sockets to agree. This node does the boring-but-essential work of reading the pose_metas off the POSEDATA, extracting the 25 body points, 70 face points, and 21 points per hand, and repacking them as POSE_KEYPOINT frames tagged with the canvas size you specify.

The inputs

• pose_data - the POSEDATA output from WanAnimatePreprocess's "Pose and Face Detection". That's the only sensible source.
• width / height (defaults 512×768) - the canvas the converted keypoints get expressed in. Set these to match your downstream generation resolution, because everything downstream assumes the coordinates live on this canvas.

Output is pose_keypoint, ready for ProportionChanger Reference, Params, or Pose Render.

One implementation detail that's worth knowing: frames where the body confidence is all zero get emitted as empty frames rather than dropped, so your frame count and timing stay aligned with the source video. That's exactly what you want when the pose data is driving a video and you don't want sync drift.

Install and notes

Part of the one-pack install - ComfyUI Manager search "ComfyUI-ProportionChanger", or clone + pip install -r requirements.txt + restart. The node itself has no extra dependencies.

The thing to double-check is the canvas size. The defaults are 512×768, which is WanAnimate's typical resolution, but if your target workflow is a different resolution you need to pass the right width/height or your edited pose will render stretched or off-center. And remember what you lose in the translation: WanAnimate's POSEDATA and this pack's keypoints are both skeletons, so conversion is lossless-ish - but a sloppy source detection produces a sloppy pose either way. Render it before you build on it.