OmniConsistency

Overview

OmniConsistency is a free plugin that helps AI keep faces and details right when switching styles. It comes from Show Lab at the National University of Singapore. Same folks behind stuff like PhotoDoodle and LayerTracer. They’re all about making AI art tools more controllable and higher quality.

You know how sometimes you try to make a sketch look like an oil painting but it loses the original look? This stops that from happening. You can throw it on top of other tools and it just works. 

Instead of mixing everything in one go this tool splits the job in two.
First it trains a style brain just for learning styles.
Then it trains another part to keep the face or shape consistent.

They even swap styles around while training so the model learns to adapt not just memorize.

How do you use it?

You can use it with things like Flux or ControlNet or anything built on diffusion models. Text prompts, images, even stylization modules work better with it on.

And it doesn’t care if it’s a style it’s never seen before. It’ll still keep the details right - it works with any Flux LoRA.

Try it on Hugginspace (see our Links section icons). Select a LoRA, enter a prompt, and upload an image to generate a new image with OmniConsistency.