/vt/ - Virtual Youtubers

Alright, LHC_alpha0.2.5. This one's trained on rouwei. https://huggingface.co/Jyrrata/LHC_XL/blob/main/alpha/LHC_alphav02_5-vpred.safetensors
Included chuubas are in the characters folder. it's 0.2.5 cause I started it just to test and haven't finished phase connect yet. Gonna do them and a few others and then bake 0.3. Hopefully done by the weekend.
Some examples:
https://litter.catbox.moe/rd7t1r.png
https://litter.catbox.moe/x7kgfy.png
https://litter.catbox.moe/qo78nl.png

I'm honestly not sure how I feel about it. Disregarding the fact that it's probably a bit underbaked, the weirdness of rouwei vpred has been pretty apparent in my tests so far.
However when I do have long prompts and really describe the light level, it's giving me some really nice results. Face detailer square is really obvious for this model though, so kinda unusable without manually fixing.

It did learn about the same as noob I'd say, though it basically doesn't know alias nono at all yet from my tests. Again, might be a bit underbaked cause I raised the batch size again, though I feel like it should know her at least a little bit at this point.

Probably gonna stay on noob and check in again for rouwei 1.0, cause it might just need to bake a little more.

>>89851343
Nice paizuri! I always forget how stacked michi actually is, kek.
>>89822521
Really nice dynamic here. Especially the facial expressions
>>89863181
Sexo!
I'm so sorry I haven't included Aia yet, I promise she'll be in 0.3
>>89864695
Whoa! Really love the composition on this one. All of them are great, but the sense of scale here is amazing
>>89867162
Kek
>>89884840
>>89886451
These are absolutely amazing. I really need to get on that traditional painter dataset