Model Introduction

Supercha is a high-quality 2D anime model.

Come To Support Me

If you think I have done a good job or have been helpful to you, please buy me a cup of coffee to support me! Plant a seed of kindness and you'll reap a harvest of love, and this will become my motivation for work! Please buy me a coffee.

Sneap Peak

Supercha_v2 is a model that can achieve rich scenes with simple tags. I won't include comparison images here.


1girl,
Negative prompt: supercha_negv2_p1,supercha_negv2_p2
Steps: 20, Sampler: Euler a, CFG scale: 3, Seed: 3921166414, Size: 512x704,
Denoising strength: 0.7, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent


1girl,
Negative prompt: supercha_negv2_p1,supercha_negv2_p2
Steps: 20, Sampler: Euler a, CFG scale: 2.5, Seed: 1774299098, Size: 512x704, Variation seed: 4282404084,
Variation seed strength: 0.7, Seed resize from: 2048x2048,Denoising strength: 0.7, Clip skip: 2, ENSD: 31337, Hires upscale: 2,
Hires upscaler: Latent


1girl,
Negative prompt: supercha_negv2_p1,supercha_negv2_p2
Steps: 20, Sampler: Euler a, CFG scale: 2.5, Seed: 1497493303, Size: 832x448,
Denoising strength: 0.7, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent


1girl,
Steps: 20, Sampler: Euler a, CFG scale: 2.5, Seed: 4161025517, Size: 832x448,
Denoising strength: 0.7, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent

1girl,
Steps: 20, Sampler: Euler a, CFG scale: 3, Seed: 3929693217, Size: 832x448, Variation seed: 433383484,
Variation seed strength: 0.7,Seed resize from: 2048x2048, Denoising strength: 0.7, Clip skip: 2, ENSD: 31337, Hires upscale: 2,
Hires upscaler: Latent

More Details [optional]

If you're using this model for the first time, i recommend downloading the corresponding version of Supercha's embedding and VAE for a significant improvement in performance.

Here are the recommended parameters:

Clip Skip: 2

CFG Scale: 2.5-3.5

In addition, I also recommend using the ControlNet plugin for coloring, which may produce better results than directly using the generated output.