Plan to made ~all characters from Astarotte no Omocha anime in next week
other characters from anime:
Works fine with all sd 1.5 models. Trained on anything V5 [Prt-RE]
strength 0.6-0.8, hiresfix or img2img for best results(it's works for every LoRa).
Train dataset is big(like 600 pictures) -> you need to use prompt for choise front/back/side view,
and negative prompt for non cropping if you want fine artistic results. But ofcourse easier
to copipast from preview images if you don't deep into neuro art.
Train models isn't easy, crypto donations on some coffee are welcome:
Boosty, subscribtion with additional content:
Description
additional trigger words:
devil school uniform gown - tryed to dress school uniform from anime
child - small dataset where Astarotta a more young
FAQ
Comments (3)
Yes plz
Where may I find this model? My results are absolutely nothing like yours. I have tried two separate models so far. This model is what I need: 7thHentaiHeaven_randomizer89
But Even If I didn't have that model, I have questions. why did you train this lora on images of Astarotte where she is half out of the frame? And why are there other characters in this same lora? I love the idea of these, and I had very high hopes, but I am pretty sad so far. BlueMix checkpoint got decent results for the colors and style to mix with this lora, but otherwise this hasn't been a great experience so far.
How many images do you usually train on to achieve the anime style?