PixlStash

Local AI image manager that searches, tags and scores photos, with desktop and self-hosted options and two-way ComfyUI integration.

PixlStash is an image manager for photographers culling shoots, creators reviewing AI images and people preparing datasets. It helps you find relevant pictures and review which ones to keep in large collections.

Content and face search: Find untagged images with CLIP text search, group people or characters with face recognition, and find similar faces or images through reverse search.
Tags and captions: Choose PixlStash Tagger, WD14, Florence-2 or JoyCaption, review uncertain labels in a queue and check individual tags on a health board.
Image review: Sort by quality, character likeness or malformed-anatomy scores, detect duplicates and segment objects.
ComfyUI workflows: Process selected images in ComfyUI or use its nodes to search, load and save library images; generated results return with tags and project, person and picture-set associations.
Folder-based libraries: Use existing folders without moving, renaming or copying files, keep independent libraries and export images to folders.

It runs as a desktop app or a self-hosted server with a browser interface, with Docker also supported. Models run on your hardware using a CPU, NVIDIA CUDA or experimental AMD ROCm support. Library data stays in local storage, and AI processing needs no cloud API or external account. Initial model downloads require internet access; later starts use cached models. A REST API supports other integrations.

PixlStash is open source under GPL-3.0. The default buffalo_l face-recognition weights restrict use to non-commercial research; AuraFace weights use Apache 2.0 and permit commercial use. It doesn't include OCR or mobile apps with background sync.