Just Launched ML Blocks - A No Code Tool to Build Custom Image Processing Workflows

Hi everyone! 👋

I'm Neil and I'm just starting out building my app, ML BLocks 🧩, in public. I started working on it 2 months ago, so I should have written this earlier but better late than never!

ML Blocks is a product that was born to solve a problem I had with bulk processing images at a previous startup. We did some very repetitive image editing work and decided we wanted to speed up our workflow by running an automated sequence of image processing actions on images like remove background, crop, resize to multiple aspect ratios, and AI caption generation.

We found some APIs online like removebg that could bulk process one piece of the workflow but we almost always needed to write our own code to glue everything together, post-process the images, and build a complete workflow.

Ultimately we realized this is a common issue in any kind of image processing task. Especially now when there are so many hosted AI models to experiment with, having to write these pipeline endpoints really slows down prototyping.

To solve this, ML Blocks lets you create custom image processing workflows that can handle multi-step image generation or analysis pipelines using node-based workflows. Just connect a bunch of blocks like remove background -> crop -> upscale and you get a full image handling workflow in a few minutes.

Here's a screenshot of a simple ML Blocks workflow. The cool thing about this is that you can take 1 input image and generate as many outputs as you need in a single click.

We're also building a way to deploy your entire workflow as a single API endpoint to integrate into apps and workflow automation tools.

If you want a more detailed explanation, I made a video tutorial here:
https://youtu.be/3kNX6iCgGoo

We're starting with about 30 blocks, 8 of which are AI blocks. For example:

• Stable Diffusion XL
• GPT4v
• Remove BG
• Upscale
• Segmentation

The rest are basic image processing functions like crop, blur, saturation etc. There are nearly infinite ways to connect them and I feel like we're just scratching the surface on the kinds of workflows you could build with this. If you're into Blender for 3D rendering, you'll notice the parallels to shader nodes.

Some workflows you can build with ML Blocks are:

1. Detect faces and blur them
2. Crop image to the best-fit square, then add a text overlay
3. Remove the background and recenter the subject in the frame

I've always struggled with sharing my work online, but the stories on IH are too inspiring to ignore so I've decided to give this a real shot and "just do it". I would love some feedback on the app if you have a minute and if you know someone who might find this useful, please share it :)

Here's a link to ML Blocks - https://mlblocks.com

Thanks a ton!