Merge of the Puffy and Guofeng models. I did not make either of the source models or alter the source models with additional training data. These models are harsh, rough, and unrefined. These models are posted to show variations in different merge techniques and more will be added. If these models could be utilized in any way I would be happy, but these may also be useless. Please leave any feedback.
Please give your support to the original models found here:
Puffy v2.0
https://civarchive.com/models/31262/puffy
Guofeng v3.4
https://civarchive.com/models/10415/3-guofeng3
Check model details for each version and the results I produced with these. Some mix's are better than others. All of them produce very beautiful women.
Description
Merge with Merge Block Pattern; Puffy <-> Guofeng Pattern: GRAD A
Results ----- Puffy Guofeng GRAD A
------------------------------
girl in a forest with floating flowers, river, puffy, puffy cheeks, cute,
-5/5 Decent Image's (hands, fingers)
boy skydiving, clouds, wind, muscular,
-5/5 Decent Image's (anatomy)
futuristic mecha in a destroyed city, scifi, mechanical, gears, transparent glass, crystal,
-4.5/5 Decent Image's (anatomy)
horse, flowing mane, striped, multicolor,
-4.5/5 Decent Image's (hoofs, anatomy)
goodra from pokemon, wet, slippery,
-0/5 Decent Image's (not goodra, anatomy))
eastern dragon flying in the sky, clouds,
-2/5 Decent Image's (not an eastern dragon, hands, anatomy)
fox, fluffy, cute, tongue out, puffy,
-5/5 Decent Image's (not a fox)
FAQ
Comments (6)
Just wanted to re-clarify. These are mainly for the purposes of learning the different merge types and what changes between them. Personal fav at this time Is SmoothStep
What tool are you using to merge these?
@kajuberdut Merge Block Weighted (GitHub - bbc-mc/sdweb-merge-block-weighted-gui: Merge models with separate rate for each 25 U-Net block (input, middle, output). Extension for Stable Diffusion UI by AUTOMATIC1111) for Automatic1111
@kajuberdut The idea I would like to see is what method does what and how each affects each other. In retrospect; I should have used 2 conflicting models of great contrast than 2 that are pretty similar. My hope is I can Identify what parts each of the sets affect what and allow another level of control over my merge's
@Delsigina This is a great experiment. I don't have the bandwidth to upload many models but perhaps I'll make some comparison X/Y charts soon by merging some of the models I already have on my disk.
@kajuberdut same lmao. over at the fiancee's. so when I can ill publish the models as they upload.