>>49231966>Hard to tell with just 3 picsI've now generated ~300 pictures, with 1/3rd each using hll-s-2, hll-s-2 with badhandv4 negative embedding, and a third as a baseline using the same model without hll-s-2 merged into it.
>hll-s-2https://files.catbox.moe/jqxbcm.jpg>hll-s-2 with badhandv4https://files.catbox.moe/trfp36.jpg>no hll-s-2https://files.catbox.moe/zogcn3.jpg