A spin off from Level4. Built to produce high quality photos. Sometimes photos will come out as uncanny as they are on the edge of realism.
Description
FAQ
Comments (11)
Nice model, will compare realism to my model. Looks awesome so far.
Clip needs to be fixed: Contains no junk data. CLIP had incorrect positions, fixed: 7, 14, 19, 23, 28, 33, 38, 41, 43, 46, 51, 56, 61, 66, 71, 76.
Have a look at the Toolkit Extension: https://github.com/arenatemp/stable-diffusion-webui-model-toolkit
Thank you for the toolkit. I have been looking for new tools to help build models. Only have used the default automatic1111's model merge so far. A while back I saw an image of some tool that allowed you to adjust the mix ratio of every layer. Would you happen to know what this tool is?
I see that I can extract the different components of the model but how can I add components like the missing clip and controlnet layers?
@AreThoseLevel4Plates Don’t know, maybe head to the discord. There are many helpful people who know more about checkpoint merging :)
@AreThoseLevel4Plates Hi I believe you are asking about the block weighted merge option. It is available as an extension, under the extension options in A1111. It is called: sdweb-merge-block-weighted-gui https://github.com/bbc-mc/sdweb-merge-block-weighted-gui.git
Is this fundamentally different from Level4? Or just a continuation?
The base was Level4_v5.0,
Custom training and merging was used to diversify the model. This model will be continuously updated to move more towards realism. My first model (Level4) will focus more on semi-realistic/digital art style.
This is really good.
Hoping for another realism update.
This is my favorite realistic model on the Citadel.
Thank you Commander Shephard