Local-first metadata workspace for ComfyUI

ComfyUI Meta Viewer

Browse generated image folders as a polished gallery, inspect prompts and model settings, open the original workflow graph, and create transparent PNG cutouts without uploading your work anywhere.

Get started View on GitHub

30+ REST endpoints

14 keyboard shortcuts

SQLite persistent local library

PNG / JPG / WEBP metadata extraction

Why it exists

Your ComfyUI archive should be searchable, inspectable, and reusable.

ComfyUI Meta Viewer turns a folder full of generated images into a fast local catalog. It keeps files in place, stores metadata in SQLite, renders thumbnails on demand, and gives every image a readable story: prompt, seed, sampler, model, LoRA, EXIF, and workflow.

Core workflow

Built for people who actually revisit generations.

Metadata extraction

Read PNG chunks, EXIF, ComfyUI prompt JSON, model names, LoRA weights, seeds, samplers, and resolution data.

Workflow graph

Inspect a color-coded SVG node graph with pan, zoom, links, node details, and workflow-aware categories.

Gallery and lightbox

Move through large folders with masonry thumbnails, image zoom, rotation, touch swipes, and metadata tabs.

Fuzzy search

Find images by filename, prompt text, model, sampler, folder, and extracted generation parameters.

In-place scanning

Scan folders recursively without copying originals. Incremental mtime checks keep repeat scans quick.

Object cutout

Select the subject in an image and export a transparent PNG while keeping cached results locally.

Interface

A desktop-feeling web app for local image libraries.

The UI is intentionally dense: a resizable sidebar for folders, a fast gallery surface, and a lightbox that keeps metadata close to the image instead of hiding it in raw JSON.

Gallery view with generated image thumbnails — Gallery view

Lightbox view with metadata panel — Lightbox metadata

Under the hood

Small stack, direct control, no documentation generator.

The app is a Flask service with a vanilla ES module frontend, Pydantic response models, SQLite WAL persistence, Pillow image handling, and vendored Fuse.js search.

Backend: Python 3.10+, Flask 3.1
Storage: SQLite with local uploads and cache folders
Frontend: Vanilla JavaScript modules and custom CSS

Install

Run it locally in a few commands.

Python 3.10+ and Poetry are the only manual prerequisites. Windows users can also run `start.bat`.

git clone https://github.com/Lotargo/ComfyUI-Meta-Viewer.git
cd ComfyUI-Meta-Viewer
poetry install --no-root
poetry run python -m app.main

Installation guide Feature reference Architecture Interactive API (Scalar) Development