R to @ostrisai: The difference is impressive. I had also noticed that using longer prompts worked better especially with finetunes and personalisation. Nice work!

R to @okarisman: Yes, just the 1.0 base model. It also works well on 1.5 models, but SDXL presents a much greater jump in quality.

R to @ostrisai: I assume this is sdxl 🤔

This is the way

R to @ifndefgt: Aradigim cevap

R to @okarisman: We have this light festival in Helsinki and I hope it was them just testing the lights :D

R to @TakoTreba: Any klu about what that is?

R to @okarisman: Nice drawing indeed!

R to @dgut_: It’s a painting by my friend (iz.denizart on insta)

R to @okarisman: #3 without the purple. What's the sketch on the wall?

Choose your side

R to @benfurkankilic: Benim 5.5 ama ekrana yeteri kadar yaklasirsam gozluksuz de oluyor 😂

R to @okarisman: Yaklaşık 4 numara miyobum var. Gözlerimi tam anlamıyla kör etmek gibi bir neyim yok, teşekkürler 😄

R to @benfurkankilic: Bence soyle bir setup yeterli: https://x.com/okarisman/status/1651165653166960640?s=46&t=TrXWEuh6yE79N25lpK_SLg

R to @Ethan_smith_20: I've tried to implement a bunch of these generative models from papers, and couldn't get the same results they showed. Had me thinking I was the problem at first. But then I realized, these models are really good for some specific things but fall short everywhere else. Super disappointing, honestly.

R to @anotherjesse: It should work in theory. If the unet shape is the same you can actually try to add the normal control net residuals 🤔

RT by @okarisman: made the landing page in hours 🎉 👉 http://shipixen.com how was it even possible? 1. I built it with... Shipixen itself 😃 2. as always, I used Indie design kit by @MrNick_Buzz & @aliszu. this was one of my all time best purchases To everyone that wants to try it: I'll get in touch shortly. Thank you so much for expressing interest, I can't wait for you to try it out 🫶

My real secret 🔥