After some training and a lot of trial and error with merging a good model out of it, I can present you Version 2 of my Latex Model.
You can find:
- the raw version that was trained on SD 1.5
- a realistism version merged with Deliberate and Bondage
- an anime version merged with Abyss, HassanBlend and Bondage
For the preview images I used (mostly) no artist tags but I would assume that with artists in your prompt you can make both versions more cartoonish/drawn looking.
If you want hoods it is best to put both "latex mask, latex hood" into your prompt.
I recommend using the "vae-ft-mse-840000-ema-pruned" to correct the greyish color on the anime model (I haven't tried either model without because I just have it always active anyway)
For example:
The Deliberate version was merged like this:
A: Deliberate B: Bondage C: SD 1.5 - add difference 0.3
A: Output B: LatexV2_25k C: SD 1.5 - add difference 0.5
Description
A merge between Deliberate, Bondage and my new latex version
FAQ
Comments (7)
Unfortunately I can't get stable txt2img results with the latest V2 model. Do you use AUTOMATIC1111 toolkit?
Yes I use Automatic1111, but this model is made for SD 1.5
Fantastic work! Any chance to share a slim prunned safetensor version?
I actually made a LORA which is much more user friendly
Any chance of making a pruned version?
Mmm... You most certainly have good taste.
With models like this i have obtained the most picturesque images, the most successful landscapes and technical images of objects, buildings and cities. When you see NSFW for the first time you think that this model only serves to generate dirty images and in return you get the best quality art