This model was trained on ~3000 NSFW photographs.
The data was collected by volunteers in the Unstable Diffusion community, and the model was trained by the Unstable Diffusion development team. You can join the community here: https://discord.gg/unstablediffusion
Description
FAQ
Comments (14)
how is this nsfw? Seeing no evidence of it
try "raw 4k professional photo of a beautiful nude woman, beautiful face, seductive gaze", and as negative tags "cropped, disfigured". that's one of my test prompts, it should work
@digburn can you upload the result?
It is fully NSFW and best from all models i've tried! Models are just amazing, will try to upload a result.
So far, I'm getting decent results with your model on my (slightly NSFW) TI embedding but I'm still curious about the details of your model. What types of images did you train it on?
it was trained on 3000 high quality photos, mostly women
This Model is made by the Unstable Diffusion community and part of the kickstarter they are running. Go checkout the Unstable Diffusion community!
I should have added a description or something but I am definitely the person who trained this model, @digburn in the Unstable Diffusion server, sorry for the confusion
then im sry. without any description it seemed like its just uploaded. i just wanted to giv credit where it belongs because the unstable community is awesom!
After a while using this one, it's obvious Hassan's stuff is far more coherent. There's a bunch of stuff on here more advanced than this.
Agreed - version 0.9 should be out soon and will be a lot better. Merges like Hassan are generally more coherent than individual models like this - I'd love to see if Hassan can take this model and stabilise it.
can we get safetensors?
Thanks for the model!
Was it trained on SD 1.5 or a mix?
You have mentioned a newer version, is it coming?