Conversation

Notices

1. LS (lain@lain.com)'s status on Friday, 01-Sep-2023 20:00:08 JST LS
   I think 1.5 is still easier to use for 'artistic' stuff, but XL is amazing for realistic images.
   • Inertia (inertia@pone.social)'s status on Friday, 01-Sep-2023 20:23:42 JST Inertia

     in reply to

     @lain those are all awesome

     LS likes this.
   • guizzy (in exile) (guizzy@shitposter.club)'s status on Friday, 01-Sep-2023 21:25:58 JST guizzy (in exile)

     in reply to
     @lain Most SDXL checkpoints I used have been better with natural language than tags which tends to make it harder to pinpoint a style.
     LS likes this.
   • LS (lain@lain.com)'s status on Friday, 01-Sep-2023 21:28:34 JST LS

     in reply to

     • guizzy (in exile)
     @guizzy i read now that XL actually has two prompts, one that uses the old style and one that's more natural language, but automatic1111 only uses the natural language one.
   • guizzy (in exile) (guizzy@shitposter.club)'s status on Friday, 01-Sep-2023 21:59:35 JST guizzy (in exile)

     in reply to
     @lain Intarasting. Automatic1111 just updated yesterday, I'm looking to see what they added and if they added that second prompt
     LS likes this.
   • guizzy (in exile) (guizzy@shitposter.club)'s status on Friday, 01-Sep-2023 22:50:14 JST guizzy (in exile)

     in reply to

     • guizzy (in exile)
     @lain Searching the code I see references to both clip_l and clip_g in the SDXL Textual Inversion settings, but nowhere else. I'm not sure whether it just puts the text field in one of them or both of them at the same time. Looking at forums and guides and it seems most people recommend using the same prompt in both in ComfyUI.
     LS likes this.