
Time to tackle the lack of anime degeneracy made for the new and awesome Flux.1!
v0.1
Make your own fun and cute anime images!
Only girls for now and no horny *bonk*
(doesn't do NSFW, might add some later)
Note: This is very much experimental, unexpected results may occur.
Usage: no trigger words needed!
-
use a combination of flux-ish, natural language and danbooru tags, because this LoRA is trained with the latter
-
use short and simple prompts
-
complex prompts sometimes create photorealistic images
-
example: a young woman with [hair color] hair and [eye color] wearing [clothes], 1girl, [basic facial expressions, eg. smile] (for more examples see images)
-
I recommend weights of >1 for the best effect
-
recommended versions (best results): v0.1a and v0.1c
Lore: After a rather long break I decided to start making and modifying models again. Pony Diffusion got rather boring, so Flux came as a nice opportunity to tinker a bit. I used 75 training images, most of which sourced from gelbooru since danbooru didn't feel like giving me any more bandwidth than some kbits per sec. Fun aside, the last 75 of the 300 steps were rather useless making little difference in style and quality of version v0.1d compared to v0.1c. Base model was vanilla Flux.1-dev. Version b tends to do some funky stuff, idk why. Ver d might be overtrained. Feel free to try all of them.
Have fun!
License: none other than all the flux stuff
描述:
The other one I recommend. Seems to generate images with more details but less saturated colors.
训练词语:
名称: aniflux_experimental_v0.1c.safetensors
大小 (KB): 149708
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success