Ratatouille famously did this, with actual scene elements rather than digital watermarking.

There’s a scene with a poster in the background. Every copy of the movie had different digits on the poster, I think with a unique ID for each cinema they were sent to. When a leak came out they could check the ID and know exactly which avenue it was leaked from.

How are they different to monkrus? Am curious

Because AI art is shit at this kind of thing. Not incapable of, just absurdly fiddly to get obedience from. It’ll get you 75% of the way there, and then you actually need to know how to edit an image, which is where it lets you down.

Go ahead and try generating images of a family during benediction wearing ear bandages and you’ll find it really struggles with getting that last part right because it’s not typical subject matter.

Their docs are decent and will guide you through setup. For advanced stuff search for Trash’s arr guides