Just got round to trying the Stable unCLIP model, it's surprisingly bad. These are the image variations for my usual Ghibli house input, they're sort of terrible!

For reference the same output for my Image Variations model is in the tweet below πŸ‘‡

---
RT @Buntworthy
πŸ“£2️⃣ Stable Diffusion Image Variations v2 release! 2οΈβƒ£πŸ“£

It took a while but I released the updated checkpoint for the image variations model. It was trained longer and mor…
https://twitter.com/Buntworthy/status/1600518238240055296

Justin Pinkney on Twitter

β€œπŸ“£2️⃣ Stable Diffusion Image Variations v2 release! 2οΈβƒ£πŸ“£ It took a while but I released the updated checkpoint for the image variations model. It was trained longer and more carefully than the original and the image quality and similarity is better.”

Twitter
And this is the Matisse one, looks more like a crayon drawing?...
@justinpinkney I thought my bad results had to do with not giving it images made with SD, but I didn't investigate very hard