đź› PixlStash 2.0: Image Library and LoRA Training in One Application
This open-source application (v2.0.0 released November 10, 2026, v2.0.1 released November 16) groups faces by similarity to build a character dataset, trains LoRA models via AI-toolkit on a local GPU, and attaches the finished model to the character—making it available in ComfyUI through custom nodes. It runs on Windows, Linux, and macOS (macOS build pending), and features a headless mode.
🌍 The personalization pipeline—from photo to finished LoRA in ComfyUI—is consolidated into a single tool, lowering the barrier to entry for personalized models.
👤 Training LoRA on your own characters or photos? Data selection by face clusters and training initiation require no manual file copying. The application is free and open-source. Note: grouping faces without the consent of the depicted individuals requires legal review.
Source 1: http://pixlstash.dev/whatsnew.html Source 2: https://github.com/Pikselkroken/pixlstash
