Stick-figure puppetry for Stable Diffusion pose control

What it does

This is a browser-based extension for AUTOMATIC1111’s Stable Diffusion WebUI that lets you manually assemble or extract human stick-figure poses and feed them straight into ControlNet. You can add skeletons from scratch, detect them from existing images, layer in a background for context, and export the result as a PNG or hand it off to the ControlNet extension without leaving the WebUI tab.

The interesting bit

The whole thing lives as a WebUI extension, so it turns the generative-AI pipeline into a closed loop: sketch a pose, send it to ControlNet, and iterate without juggling separate tools. The README also quietly notes that you should leave ControlNet’s preprocessor empty—implying the editor itself is doing the structural heavy lifting.

Key highlights

• Detects poses from uploaded images or lets you build them manually from scratch
• Supports background images for spatial context while editing
• Exports PNGs or passes skeletons directly to the ControlNet extension
• Runs entirely inside the AUTOMATIC1111 WebUI interface

Caveats

• The README is sparse on technical details; it is unclear exactly which pose-detection backend is used or whether it requires external model downloads
• A macOS SSL certificate workaround is documented, indicating the extension makes HTTPS requests that can trip up stock Python installs

Verdict

Grab it if you use AUTOMATIC1111’s WebUI with ControlNet and want precise pose control without leaving the browser. If you don’t generate people or don’t use that WebUI stack, this is just a niche stick-figure utility.