MeinaUnreal版本V1 ( beta ) ( beta )

MeinaUnreal objetive is to be able to do anime art with a 2.5d feeling.
I have a discord where you can share images, discuss prompt and ask for help.
https://discord.gg/meinaverse ヾ(≧▽≦*)o
我有个可以让你分享图片和参与讨论与询问问题的discord群,

I also have a ko-fi and Patreon page where you can support me or buy me a coffee <3 , it will be very much appreciated:
https://ko-fi.com/meina and https://www.patreon.com/MeinaMix

MeinaUnreal is officially hosted for online generation in:
- SeaArt
- Mage.space ( with animate feature )

Recommendations of use:
--------------------------------------------------------------------------------
Enable Quantization in K samplers.

Hires.fix is needed to drastically improve the quality of face and eyes!
All of the sample images were generated in 512x1024 and upscaled 2x twice, hires and extra.
--------------------------------------------------------------------------------
Recommendations:

Sampler: Euler a, 40 steps.
Sampler: DPM++ SDE Karras, 20 to 30 steps.
Sampler: DPM++ 2M Karras, 20 to 40 steps.
CFG Scale: 7.
Resolutions: 512x768, 512x1024 for Portrait!
Resolutions: 768x512, 1024x512, 1536x512 for Landscape!
Hires.fix: R-ESRGAN 4x , with 10 to 30 steps at 0.3 up to 0.5 denoising.
Clip Skip: 2.
Negatives: ' (worst quality, low quality:1.4), monochrome, zombie, (interlocked fingers), '
--------------------------------------------------------------------------------
This is an Attempt to update MeinaMix V5 into its own branch according to feedback. so in the merge list its MeinaMix V1~5.1, MeinaHentai V2.1.

描述:

If you have questions related to the model, samples, how to get started or just want to share your creations, feel free to join my discord: https://discord.gg/XC9nGZNDUd <3

V1 is a beta or Concept of update of MeinaMix V5, its unstable and very hit or miss, it absolutly hate hands but if you have the patience to generate enough images there can be some really cool images.

If you rather wait for a more stable and consistent version, please wait for V2! with feedback + proper weight changes it will be more stable and consistent.

( all of the images in the samples are upscaled by 2x using SDwebui extras tab, and the original size of them is 512x1024 which is made to represent worst case scenario, i generated 5 and picked the best between the 5 images.)

Recommendations of use:
---------------------------------------------
Enable Quantization in K samplers.

Avoid or Do not add ' Photorealistic ', ' 3d ' or ' realistic ' in the prompt. Those tend to make weird generations in V1.

Negatives: ' (worst quality, low quality:1.4), monochrome, zombie, '

The best samplers in most of the generations is Euler a - 25 steps OR Euler - 50 steps. ( Clip skip 2. )

Resolution: 512x768 , 512x1024 is unstable and others resolutions may work but they're not tested.

Upscaler in most of the scenarios is R-ESRGAN 4x or +Anime6B, with 15 steps at 0.1 up to 0.3 denoising.
-----------------------------------------------

训练词语: