[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/vt/ - Virtual Youtubers

Search:


View post   

>> No.67019731 [View]
File: 3.02 MB, 1152x2016, ComfyUI_03757_.png [View same] [iqdb] [saucenao] [google]
67019731

>>67018656
Sorry, force of habit. Also, huge respect to your work. Has helped me out a lot.
And thanks, I had no intention of stopping. I might make it to my 5000th gen in my first week.

>>67018871
Do I look like a nerd that reads papers?
Cause I'm planning on reading it at some point, but I honestly haven't had the energy after work this week

Crop is pretty self explanatory then.

Any difference between width and target_width? And yeah, the resolutions are definitely important. Animagine probably has the same, or at least similar as base sdxl.
Another anon had the idea to try making images taller to avoid cramping (I ended up using 768x1680, and scaling that by 1.5 in the end) which worked amazingly with the old prompts. With these it seems to just make the pictures so big that any upscaling causes deformed torsos and limbs and multiple belly buttons in nearly all images.
So far I'm actually going under the recommended 768x1344 and using 688x1344, as I still think thinner images might be less prone to cramping for upright characters.

I haven't tested the same prompt in both enough to really be able to notice a major difference. I just found the idea of being able to signal how to apply piece of the prompt very intriguing. I'll test it a bit before calling it a night.

Navigation
View posts[+24][+48][+96]