marduk191's Pseudo-Modern Anime (lightning)版本v1.0 (ID: 355519)

there is a non-lightning version available above.

This is an early version of an Anime model I have been working on lately that finally got to a starting point. Trying to fuse the darkpizza oldschool Japanime look with some modern anime looks. It's missing some concepts and cohesion still, but this is a start. All cover images are random prompts thrown at it with basic adetailer. Yes, the hands somewhat suck lol. Please use the embedded VAE. This is put together for simple text prompts that turn out heavily stylized. You will need external control of some kind if you try to get very intricate with it. The NL version is more generic and may be better for traditional works. Just try em out. This is mainy for me, I shared it because I liked where it is heading.

~

This still needs face detailer currently for eyes, nothing aggressive.

~

Settings:

CFG: 1-2

Steps 4+ (5)

Sampler: dpmpp_sde_gpu

Scheduler: karras

Denoise: 1.0

Noise mode: GPU

NL version should be fine around 20 steps and 6.5CFG at the same settings.

Negatives:

Whatever you need, here is an artsy one.

~

embedding:unaestheticXL_hk1

, hat, (low quality, worst quality:1.2), (grayscale, grainy, border), (watermark:1.2)

on-site generation is available here.

Tip Jar

Discord: https://discord.gg/btCfTh4jgt

描述:

v1:

DarkPizza Origin v3.0, DuplicationXL V1.0, VooDoo, aamXLAnimeMix v1.0, Lightning, XL_VAE_C A little extra training to offset some styles, merged a couple of trained loras, WIP

训练词语: