This is a massive, highly curated dataset consisting of thousands of native ComfyUI workflows extracted directly from Civitai metadata.

Unlike traditional metadata dumps, this archive contains zero image payloads. Every single file has been scraped and stripped down into its raw, clean .json node graph format. This means a microscopic download size for an immense amount of functional data.

Key Features & Details:

• Pure Data: 100% .json format—no bloated images, no hidden junk files.

• Searchable Filenames: Files are dynamically named using the unique Civitai Image ID followed by a shortened snippet of the workflow's internal positive prompt (e.g., civitai-127819953-1girl,a girl cosplaying as hoshimi miyabi,asymmetr.json). This allows you to instantly search and filter the archive via keyword for specific styles, characters, or concepts right from your OS file explorer.

• Tested & Verified: Spot-tested across various categories; files load cleanly and directly into ComfyUI without node graph corruption.

• Massive Variety: Features an extensive cross-section of modern AI generations, including heavy distributions of SDXL, Illustrious, Pony, Flux, Wan2, and specialized FP8/BF16 pipelines.

• Manifest Included: Includes folder_file_counts.txt inside the archive, giving you a full breakdown of exactly how many workflows are included per model type.

⚠️ Note on Duplicate Workflows / Image IDs: The numerical string at the beginning of each filename is the specific Civitai Image ID the workflow was pulled from. Because creators often generate multiple images or batches using the exact same prompt and node configuration, you will encounter some identical workflows with different Image IDs. This preserves the exact lineage of the image it came from and ensures a complete dataset.