My very first Lora. Results may look very novice mind that. But I trained on images of her from direct screencaps from the show and nude model sheet I made using her proportions. Didn't come out too bad.
You may want to use "green pants, apron, green collar shirt, red hair, yellow gloves, green shoes" if you want her original look.
This Lora was trained on the anythingv3 model, but these images are from the anything 4.0 model. I may in the future try to make a better version once i get more experience in making them.
Update: I added more images trained on her to the new model. You might want to use the new one with the anything.v4.5 model for better results.
I added a nude database also. You can trigger an accurate representation of her by adding "(on-model). Or you can check the prompts in the database provided.
Edit: for the weight, use 0.7-0.9
Description
The new version is trained on 3092 steps with more images added
FAQ
Comments (9)
Man of culture
I didn't know I needed this until now
Face differs too much.
Ive noticed this too. I'm coming up with a new way to train loras
@Geno yayy thanks!
i think i fixed that now
@Geno wow cool. ill try it
She's got it goin' on.
Can't load this Lora
raise AssertionError(f"Bad Lora layer name: {key_diffusers} - must end in lora_up.weight, lora_down.weight or alpha")
AssertionError: Bad Lora layer name: lora_unet_down_blocks_0_downsamplers_0_conv.lora_mid.weight - must end in lora_up.weight, lora_down.weight or alpha
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.