2.x is just broken

#7945
by ramulloki - opened

I've compared model versions 1.5, 2.0 and 2.1 and I don't see any reason to use the latter. 1.5 gives the best results always. It's not about prompts and their wording, the guide doesn't help to make better queries. It's just that the 2.x model is broken and needs to be rethought. In an attempt to satisfy people who have nothing to do with image generation, but act as antagonists, opponents and critics,
stabilityai has ruined the experience for those who are supporters of their work.

1.5
222.png

2.0
72665dbe-9bec-49d7-97aa-9240282d2fa6.jpg

2.1
fddd5979-2c8b-460b-8150-1dcbbbe51a68.png

I agree. The change to a different data set and focus on negative prompting for 2.0 and 2.1 has provided me with less than optimal results compared to 1.5.

Generating images used to be more enjoyable. Sigh.

Yeah, it's completely broken. It makes horrible looking images compared to 1.5. What is to be done now? Are they just going to leave it like this? It's utterly ruined. It's a complete joke of what it was. It can't even regenerate an exact copy of an image anymore using the exact same seed. I can't believe how bad it is now. Are they going to fix this?

Sign up or log in to comment