MySLR - fast, obedient and accurate!
Version 1.2 does not require VAE
This is a fusion of three of my favorite models. I have been mixing models in different proportions for a long time to achieve good results for me.
And yes, I'm happy) The model copes well with many areas of art, creates realistic characters in amazing detail. Perfectly generates hands, face, hair. Understands skin texture correctly.
Recommended settings:
Sampling Method: DDPM, Euler A Turbo, Euler A
Number of sampling steps: 10-12
CFG scale: 1-4 (lower, more realistic)
Use minimal negativity: ugly, deformed, noisy, blurry, distorted, out of focus
In most cases, the model does an excellent job without negativity.
No matter what you decide to create today, this model is very likely to make your wishes come true) I will be glad if this model helps you!
If you like my work, please subscribe to me and write a review. Thank you for appreciating my work.
Description
Baked VAE. The excess brightness has been reduced
FAQ
Comments (11)
So which models did you merge?
This model was mixed with three models that I had previously made by mixing other models.
@Abzaloff exact names please.... thanks in advanced.
@frikitin Unfortunately, I can no longer remember which exact models I mixed this one with. It's been over a year since then, and the source files weren’t preserved.
спасибо
can you send a 16bit smaller version. thanks
The FP16 model will be released soon. You can check out my profile for another new model that I posted today.
i just wish it was half the size. I never found any difference between large and small models other than it takes more bandwidth and space
The FP16 model will be released soon. You can check out my profile for another new model that I posted today.
there is a difference but for the average use or even most use cases FP32 isn't necessary, but i appreciate the full weights
@Chadxai the latest version of FP32. Here we were talking about another version