I got tired of losing my Stable Diffusion prompts and workflows to broken ComfyUI custom nodes. Relying on PNG metadata is a highly fragile pattern for professional pipelines—the moment you update an environment or share an image, the workflow breaks or the data gets stripped entirely. My team and I built a free, 100% client-side local parser to fix this bottleneck. It reads hidden tEXt chunks from ComfyUI/A1111 PNGs directly in your browser, without requiring Python environments or specialized nodes. It extracts your raw prompt and saves it as a clean, indexable .deut text file. Key details: Zero-server: No images or text data are ever uploaded to any external server. Everything runs locally in your browser. Open Standard: We formalized the .deut format so studios can safely store their intellectual property in a transparent container instead of binary blobs. We are currently building a full-fledged professional prompt editor based on this format (with versioning and cryptographic verification), but the extractor is live and completely free to use right now. Would love to hear how you currently manage your AI generation workflows and any feedback on our approach!

Deut.li

Free client-side ComfyUI/A1111 PNG metadata extractor

Free client-side ComfyUI/A1111 PNG metadata extractor