Type | |
Stats | 1,630 |
Reviews | |
Published | Feb 12, 2023 |
Base Model | |
Hash | AutoV2 357BE8BDC7 |
If you like this model,Want to learn more and add ❤️ to receive the latest information.
(The difference between the two versions is: V2 is more stable, V1 is more luck-dependent but it is suitable as a merge material)
I performed a deep U-Net change merge on top of the DiscoMix_v2 model, replacing its Text Encoder and Auto Encoder, with a better Text Encoder and Auto Encoder.
The result of the test is that the model knows more than the original, it is easier to generate the desired effect, the abstract background image and the appearance of thickly painted color blocks are reduced, the picture is more detailed, and the hands are optimized a lot.
recommand :
Steps: 20+
hires.fix with denoising strength 0.5-0.6,
Upscale by : 1.25，1.5，1.75
Size: 640x960，960x640
Clip skip: 1，2，3
CFG scale: 7.5，9，11
Don't forget VAE
The DiscoMix_v2 model allows you to generate anime images without the difficulty of choosing from a wide range of other models, it will become your default choice :) , delete delete delete delete those models that take up your precious space, the DiscoMix_v2 model will be your only choice. It can satisfy all your imagination about anime characters (probably), as long as you know its tags (probably).
:)
if you would like to support me:https://www.buymeacoffee.com/EEEMax