What do you mean uncensored?
This model, just like the other Hellcoders I have uploaded, are different from normal 2.1 because its text encoder do EVERYTHING (and more) than a 1.5 model can, while 2.0/2.1 can not.
What's different from other the Hellcoders you posted?
This model features a fine-tuned negative prompt that has been 'programmed' to force the offset noise on txt2img while also improving the quality of each generation with almost zero effort (just copypaste and adjust with some other keyword).
Why?
This allows you to use the model also in Diffusers, since the original model needed the negative embeddings that can not be used in the Diffusers Format.
Why would I use Diffusers?
You can use Stable Diffusion on Colab (legally) using the diffusers format, since Google has banned ONLY the WebUIs (Automatic1111, VladDiffusion &TC), this is VERY good if your computer is not good enough or it's even better if you're about to go on vacation (almost summer holidays time) and you have only a meh laptop.
How do I use Diffusers on Colab?
Try the Deforum notebook.
Description
It is fine-tuned on quite a lot of negative prompts, try combining them how you want, use all these trigger words only in the negative prompt.
FAQ
Comments (8)
All the images generated took some short prompts like "SFW portrait photo of character/person", the Marilyn photo required 'black and white' in the negative but that's not overbearing.
All the images were generated on Deforum 0.7, the negative prompt used was '#spoonbanal##rotten#output##ammazza#che#cagata#'
I read above:
"This model, just like the other Hellcoders I have uploaded, are different from normal 2.1 because its text encoder do EVERYTHING (and more) than a 1.5 model can"
You state "uncensored"
I can't wait to see actual non-lora-filled pics which provide proof of this! :)
Check the other models I uploaded https://civitai.com/models/106489/hellcoder-trying-to-uncensor-stable-diffusion-21
The pictures I uploaded in this model, as written in the description that you should have read, were made with Deforum colab notebook and this means that there is no metadata available because it is NOT Automatic1111.
Anyone mind explaining the gibberish in the triggers section?
As in the description of the model, you should use those on the negative prompt. They are fine-tuned negative prompts for the usage in Diffusers format, since it does not support embeddings or negative embeddings.
It gave this error: Prompt execution failed
Prompt outputs failed validation:
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.