/vt/ - Virtual Youtubers

Didn't see one of these yet so I made a big chart to see how well it works on hololive when given no extra instructions. Some worked better than others. A few that I didn't expect to really work (aloe, a-chan, kurokami) didn't work. Kobo I think just got unlucky. It's interesting how there's a split between standing far from viewer and standing closer to viewer.

Personally I have a really hard time with fubuki for some reason, which is strange since there's a lot of art of her and the other models do her well. You can obviously just prompt her features and it works fine but if you try to do her name only, which works well on so many others, there's a decent chance you get some totally random girl.

I'd also recommend always using "from hololive" with this, it makes them look a lot better with very little effort.
best quality,name from hololive, 1girl, standing

Negative prompt: lowres, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, jpeg artifacts, signature, watermark, out of frame, extra fingers, mutated hands, (poorly drawn hands), (poorly drawn face), (mutation), (deformed breasts), (ugly), blurry, (bad anatomy), (bad proportions), (extra limbs), cloned face, letterbox, censored, comic, (speech bubble:1.2), multiple perspectives, (monochrome)

Steps: 32, Sampler: Euler a, CFG scale: 9.5, Seed: 3330992480, Size: 512x768, Model hash: f57ffda8, Eta: 0, Clip skip: 2, ENSD: 31337

using hll-epoch-000015.safetensors