It's a model that was merged using a supermerger
↓↓↓ fantasticmix2.5D ↓↓↓
https://civarchive.com/models/20632/fantasticmix25d
An example is using dynamic thresholding. Reduce the CFG scale to 7 or less when not in use
You must use VAE
Description
I merged several types of face lora into a little. and I adjusted the color a little bit
FAQ
Comments (8)
V1 Pruned PLS
Why V7 it's low contrast ? or something wrong on my pc ?
Please read the explanation in the text. Use the vae
Although it is now 7.0, 3.0 is the best effect. In my tests so far, I've found that it even outperforms all existing models, all of them. That's great. It works better with the eyes and teeth, and in the case of Tucson, it can more perfectly recognize the movement of closing the eyes. But why does the later version not have the stable quality of 3.0? I hope the author makes a better version based on 3.0. Mainly further optimization of teeth and eyes!
Thank you for your advice.
@michin Version 3.0 in Stable-diffusion-img2img.
It is possible to generate a very good effect in denoising strength, with a parameter of 0.4 or less.
The shape of the eyes, mouth, and hair can be closer to the original image, and the resulting image will look closer to the model.
It is possible to learn the shape of the original picture while making it exquisite.
So 3.0 is very good for animation. And it's the best model I've found so far.
I wonder if it can be further optimized based on this feature?
Translated with www.DeepL.com/Translator (free version)
@michin Version 3.0 at img2img is stunning!
I've done really great animations in 3.0.
While finding it to be the best so far.
But I think it should be able to be better.
I posted a video below under the 3.0 post.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.