VERSION 2.0!
A bit of an improvement on the dataset and captioning in v2.0, should have better prompt adherence overall, especially around poses and facial expressions!
This new version is also using a higher overall step count, since so much of training dataset was nudes, her nipples very often stick out of whatever she's wearing (especially with sulphur/10eros base models) - You can avoid this with lower LoRA strength (0.5-0.8) if you want her clothed without nipples!
My first attempt at a custom original character LoRA! A sexy redhead with green eyes named "Leena Cindertrap". Training dataset created with strict Z-Image-Turbo prompt and cfg to generate consistent base images, upscaled 4k, then resized/cropped to 2048x2048. The generated dataset images contained variety of poses, including full nude images, so basic nude generations should work decently well on regular LTX2.3 base without any other LoRAs.
Trigger word: l33naFor i2v:
Works well generating starting images with [ZIT] Leena Cindertrap (OC) Character LoRA for Z-Image-Turbo
Prompt Example
l33na is wearing a sexy blue bikini, standing outside next to a pool, looking at the camera with a shy seductive expression she says "Do you want to get wet with me?"NSFW Example
l33na is completely nude sitting outside on a park bench with her hands on her knees with her legs spread apart and open wide, her pussy is visibly wet and drips a subtle few small drops of clear liquid, she says "I'm wet and ready to go... get your dick out and fuck me already"Disclaimer
This model depicts an original fictional character and was created solely for artistic and creative purposes. The character is not intended to represent, reproduce, or imitate any real person, public figure, celebrity, or private individual. No real person's name, image, likeness, identity, or biometric data was used in the creation or training of this model. Any resemblance to actual persons, living or deceased, is entirely coincidental and unintentional.
The creator makes no claim that the generated outputs depict any real individual. Users are solely responsible for their use of this model and any content generated with it, including compliance with applicable laws, regulations, platform policies, and third-party rights.
Description
Release v1.0!
FAQ
Comments (3)
You trained this Lora on just images? I didn't know you could do that? Did you use the LTX trainer or AI toolkit?
Ya! Using AI Toolkit, mostly default settings.
Cindertrap 🤭