Character LoRA for just1n, trained on SDXL 1.0.
This model is intended for generating a masculine / bara-style male character with strong body shape consistency, chest hair, glasses, underwear/pants styling, and nipple piercing details depending on prompt strength.
Trigger word:
just1n
Recommended prompt starters:
just1n, 1boy, male focus, bara
Example prompts:
just1n, 1boy, male focus, bara, underwear, chest hair
just1n, male focus, pants, underwear, bara, nipple piercing, chest hair
just1n, 1boy, male focus, bara, nipples pierced by thick gold rings
Recommended negative prompt:
bad quality, low quality, worst quality, jpeg artifacts, blurry, pixelated, out of focus, watermark, text, signature
Training details:
Base model: SDXL 1.0
Type: Character LoRA
Resolution: 1024
Training images: 114
Labels/captions: 114
Epochs: 20
Steps: 10000
Network dim: 32
Network alpha: 32
Optimizer: Adafactor
LR scheduler: Cosine
Notes:
This is the SDXL V1 version of the LoRA. It may behave differently from the Pony, Illustrious, Custom, and Anima versions, so use it primarily with SDXL-compatible checkpoints.
For best results, keep just1n near the start of the prompt and add body/style descriptors after it.
Description
V2 — Pony / PDXL version
Added a Pony / PDXL-compatible version of jul4-just1n. This is a separately trained Pony version, not just the SDXL V1 file reused.
Changes from V1:
* New Pony-trained LoRA variant.
* Intended specifically for Pony Diffusion V6 XL and Pony-derived checkpoints.
* Keeps the same character/concept direction as the SDXL version, but adapted for Pony model behavior and prompting syntax.
* Includes newly generated Pony workflow preview samples.
* The final recommended LoRA is epoch 20.
Note: Use this version with Pony/PDXL models. Use V1 for standard SDXL checkpoints.