fp32 Iwaku-Real: I'm not the owner, reuploaded from https://huggingface.co/Iwaku-Real/Qwen3-0.6B-Base-heretic-test

feel like it have more variety in generation with same prompt, but have way more artifact.

fp16 DavidAU: I'm not the owner, reuploaded from https://huggingface.co/DavidAU/Qwen3-0.6B-heretic-abliterated-uncensored

edit: my old comic images (the one I uploaded) I am struggling to recreate them again. I am trying find out what cause that. the other last images (yae miko) those I am able to recreate

use this text encoder with the anima model

This text encoder behave different than the text encoder that many use with anima model. one of them is that it follow your prompt way closer. the quality of image (style) is way worse, but if you care about a text encoder that follow your prompt way closer than it is worth it IMO.