No trigger word
Helping words: alt girl, bad quality, grain, poor lighting
(They don't required, but with them output is better)
Recommended weight: 0.9-1
To my disappointment flux is very poor at generating realistic selfies of the girls, so this lora should fix it. didn't understand the settings, so there may be aggravation in the hands or prompt adherence
Trained on 50 handpicked images (only girls faces) with llava captions.
Maybe I'll try it soon with a different dataset (not just female faces) and different captions. Glad for advice and any help
Description
chroma version, trained on the same dataset as flux lora
works on default and flash as well
FAQ
Comments (7)
Hey, that’s a nice LoRA! It’d be great to see more LoRAs from you with that amateur 2000s photo aesthetic, like the ones by Danrisi. Also, you’ve got some really nice artistic LoRAs with a painterly style, it would be awesome to see them converted for Chroma too.
💕👍🏻
noodlewhore lora lmao
zit version pls
im waiting for the base version, it should be more trainable. zit already is good enough for these girl type tbh
and i tried civitai zit trainer 3 times, all of them are failed, 0 epochs, idk why
@Kellenok this just released apparently. dont know if its better https://huggingface.co/ostris/Z-Image-De-Turbo
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.