Compare the body shape of a new character against an existing character library.
Detect when different LoRA characters have quietly converged on similar bodies.
Inspect per-segment breakdowns of skeleton length and body build similarity.
| fuguesegue/bodyrec | aarav90-cpu/fetchit | adammartinez271828/x4-analysis | |
|---|---|---|---|
| Stars | 10 | 10 | 10 |
| Language | Python | Python | Python |
| Setup difficulty | hard | easy | easy |
| Complexity | 4/5 | 2/5 | 2/5 |
| Audience | vibe coder | developer | general |
Figures from each repo's GitHub metadata at analysis time.
The app installs easily, but the ComfyUI measurement server needs a GPU and several specific extensions to set up.
BodyRec is a desktop tool for people who train character LoRAs, a type of custom AI model used to generate images of a specific character. When you train many different character models over time, some of them end up rendering bodies that look surprisingly alike without anyone noticing. BodyRec answers a simple question offline: given one character, which other characters in your collection have a similar body shape. To build its library, you give BodyRec a folder of full body images for each character. It measures each image and stores a compact body fingerprint per character. Once that library exists, picking any character instantly ranks every other character by similarity, using only local calculations with no internet connection needed for that part. Rather than combining everything into one similarity score, BodyRec reports two separate numbers. One covers skeleton proportions, such as the relative length of the arms, legs, and torso, measured using a 3D pose estimation model. The other covers build, meaning thickness and girth, measured using a body shape model. The tool deliberately keeps these separate because two characters can share a skeleton shape while having a very different build, or the other way around, and combining the numbers would hide that difference. The measurements themselves come from a separate server you run yourself using ComfyUI, a popular tool for running AI image and pose models, along with several specific add on node packs the README lists and links to. This measurement server only needs to run when you add a new character to the library, not during everyday browsing. Setting up the app itself on Windows is simple, just double clicking a launch file that installs its own small set of dependencies. Setting up the measurement server is the harder part, requiring a working ComfyUI installation with a GPU and several specific extensions, which the author describes as an afternoon of work for someone already familiar with ComfyUI. The application code is released under the MIT license, but the underlying pose and shape models it depends on carry separate research only licenses, so using the measurement pipeline for commercial purposes may not be allowed under those models' own terms.
A desktop tool that measures body proportions across character AI models and shows which ones share a similar body shape, all offline.
Mainly Python. The stack also includes Python, PySide6, ComfyUI.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
double-check against the repo, no cap.