I AM NOT THE ORIGINAL CREATOR OF THIS EMBEDDING
HE IS -
https://huggingface.co/nick-x-hacker/bad-artist
The images above were generated with only "solo" in the positive prompt, and "sketch by bad-artist" (this embedding) in the negative.
The embedding uses only 2 tokens.
The images above were generated with only "solo" in the positive prompt, and "sketch by bad-artist" (this embedding) in the negative.
The embedding uses only 2 tokens.
There are currently 2 version:
'bad-artist': Not as strong, but produces pretty unique images (recommended)
'bad-artist-anime': More generic anime style (this was the first version uploaded)
I recommend using with 'by', so for example "sketch by bad-artist", or "painting by bad-artist", or "photograph by bad-artist", etc.
Trained with 2 vectors per token for 15000 (1850x8) steps, at 500x500, on an Anything-v3-based model.
Full generation parameters for images above (using the 'bad-artist' version, not the 'bad-artist-anime' version):
solo
Negative prompt: sketch by bad-artist
Steps: 15, Sampler: DPM++ 2M Karras, CFG scale: 4, Seed: 1476197242, Size: 512x640, Clip skip: 2