A billie lora trained on a single event with various images. Captures likeness, hair, and expression well.
Examples
example output
example output
example output
example output
example output
example output
Training Details
Field
Value
Base model
Tongyi-MAI/Z-Image-Turbo
Training tool
ai-toolkit
Steps
3000
Resolution
1024×1024
Precision
qfloat8
Trigger word
billie eilish
Usage
Recommended lora strength: 0.8 – 0.9.
Known limitations
Trained on a single event dataset, so background environments may drift toward the training context without explicit prompting. The dataset images are all from a single event, so clothing/background leans heavily towards that one outfit. Prompting for clothing however makes it easy to adjust this.
This is my first attempt at training a lora, so expect some jank.
