Hey HN! I’ve been working with AI models like Stable Diffusion and DALLE-3 for a while now and realized that building real applications with them usually requires building a multi-step image processing workflow, either to chain several models or combine AI results with basic image processing functions like cropping and resizing.<p>For example, you might want to take an input image, enhance lighting, upscale it, crop to square, and then add a text overlay. You can use a hosted AI model to solve some pieces of the puzzle, but ultimately you need to write your own code to glue everything together. I got pretty frustrated doing this for every project and decided to build ML Blocks.<p>ML Blocks lets you create custom image processing workflows that can handle end-to-end image generation or analysis pipelines using simple node-based workflows. Just connect a bunch of blocks and you can instantly deploy your entire workflow as a single API endpoint.<p>This initial version has around 20 blocks to play with. I will continue to add blocks as requested and I can see a possible future where this becomes like a Zapier for Image Processing, where blocks are community-contributed with a standardized API interface. I’d love to get your thoughts on this!