Slice SDPose Keypoints

Slice SDPose Keypoints is a time-slice utility: it takes a pose sequence and returns just the frames you asked for, by start index and count. start_frame 120 with frame_count 80 gives you frames 120–199. That's the whole node - one input stream, two integers, one output.

Boring? Yes. Genuinely useful? Also yes. It's the piece that makes loop- and chunk-based long-video workflows work cleanly. WanAnimate and SCAIL-2 generation is built around fixed-size chunks - 81 frames is the WAN native context - and when you drive generation segment by segment you want the pose slice that matches the segment you're about to render, not a recompute or a manual trim of the whole JSON. Pull the slice for chunk N, feed it to your drawing or animation node, and let the loop advance the offset.

Inputs and outputs

• pose_keypoints - the POSE_KEYPOINT stream (required).
• start_frame (default 0) - where to begin, 0-based.
• frame_count (default 1) - how many frames to take.

Outputs are sliced_pose and input_frame_count (the length of the source sequence, useful for loop arithmetic like "start = previous_start + frame_count"). Out-of-range requests just clamp to what's available, so a short final chunk won't error - you get fewer frames than asked and the input count tells you how to react.

Install

Part of the ComfyUI-CustomNodeKit pack. ComfyUI Manager → search "ComfyUI-CustomNodeKit", or:

cd ComfyUI/custom_nodes
git clone https://github.com/user2318/ComfyUI-CustomNodeKit.git
cd ComfyUI-CustomNodeKit
pip install -r requirements.txt

Restart ComfyUI. No models, no downloads - it's list slicing with nicer naming.