// the find
comfyanonymous/ComfyUI_examples
Examples of ComfyUI workflows
A companion repo to ComfyUI that stores example workflows as image files with embedded metadata — drag the PNG into ComfyUI and the full workflow loads automatically. Covers the full range of models from SD1.5 basics through Flux, HunyuanVideo, Wan 2.2, and current video/audio models. Aimed at ComfyUI users who want working starting points rather than building pipelines from scratch.
The embedded-metadata trick is genuinely useful — each example image is also its own workflow file, so there's no separate JSON to lose track of or version-mismatch against the image. Coverage is kept current: Wan 2.2, Flux Kontext, Cosmos Predict2, and HiDream all have entries, which is more up-to-date than most community resources. The per-folder README pattern keeps examples organized without a monolithic doc that goes stale. Some folders (ltxv, cosmos, wan) include both JSON and rendered output side by side, which makes it easy to verify what a workflow should produce before running it.
This is documentation, not code — there's nothing to run, test, or install, so 'stars' and 'forks' mostly measure how many people bookmarked it. Many older folders (hypernetworks, unclip, gligen) have single images with minimal README text and no explanation of what the parameters are doing or why the workflow is structured that way. The tutorial VN linked in the README is marked 'early and not finished' and appears to have been in that state for years. No versioning or compatibility notes — a Flux workflow from 2024 may silently fail against today's ComfyUI and there's no indication which ComfyUI version any example was created against.