InvokeAI UI/UX

10 October, 2022

For the last month, among other things, I have been busy developing one of the forks of the Stable Diffusion, InvokeAI. I help make the interface, test it, translate press releases into Russian, and actively argue with programmers.

The development is spontaneous and somewhat chaotic, like the whole opensource (enthusiasts gathered, do what they want, and they want to put as many functions as possible), but both I and other participants are experienced enough to make something fast and convenient — which I personally lack in other implementations of UI's to SD.

And we're actually making a small neural network-based image editor so that it's possible to assemble a collage of multiple images to generate directly in Invoke, rather than having to open photoshop for each new picture change.

InvokeAI 2.0 was released today. It requires only ~3.5 GB of video memory to create 512x768 pixel images (and even less for 512x512), and is compatible with Windows/Linux/Mac with M1 and M2 (and it seems to be the fastest SD option for macs in general). Installation is simple, but we'll make it even simpler soon.


• • •