Artist's resources:

https://www.pixiv.net/en/users/53398478

https://x.com/PwNziJfhlnoA7Lw

ANIMA INFO:

Sorry @LivingFossil I couldn't make it any faster.

No TW required.

Sampling method - ER SDE

Schedule type - Beta

Steps - 20~

NOOB V-PRED INFO:

CURVY WOMEN

All pics in dataset had either "sweat" or "very sweaty" tag so you probably want to put those to negatives.

Recommended samplers:

• DPM SDE++ ★

• DPM++ 2S a

• DPM++ 2S Ancestral CFG++

• Euler Ancestral CFG++

• DPM++ SDE CFG++

• Res Solver

• DEIS

• DPM++ 2M SDE Turbo

• DPM++ 2S Ancestral Comfy

Recommended checkpoint:

NAI-XL vpred1.0 2d accelerated "Final" version

Base V-pred noob is kinda fine too but not really recommended.

Recommended upscale options:

Upscaler: foolhardy Remacri or Lollypop or Fatal Anime

Upscale by: from 1.25 to 2.

Hires steps - half of your Sampling steps.

Denoising strength - I usually set the value to 0.33, but it depends on what Hires sampling method you use.

Hires sampling method: DPM++ SDE CFG++ for highlighting details or DDPM for a nice smoothing.

If you found the LoRA interesting or somehow useful, consider supporting me by joining my Telegram channel. I post thoughts there and do streams of LoRA making.