LongCat 6B Model - Emoji Lora For Test
This repository contains two specialized LoRA adapters trained on the LongCat Image Dev base model and LongCat Image using the t2itrainer framework. Both models were fine-tuned on a custom Emoji dataset to generate stylized 3D anime characters. The adapters are designed to work with Stable Diffusion 1.5-based pipelines.
Model Variants
longcat_dev-9-450.safetensors
Based on LongCat Image Dev checkpoint.
✅ Closer adherence to training set characteristics
✅ Better preservation of emoji-specific featureslongcat_image-9-450.safetensors
Based on LongCat Image checkpoint.
✅ Superior aesthetic quality
✅ Improved limb coherence and anatomy
✅ More polished visual output
Trigger Word
a 3d anime character
Key Differences
| Model Variant | Best For | Visual Characteristics |
|---|---|---|
longcat_dev-9-450 |
Training set accuracy | Faithful to emoji semantics |
longcat_image-9-450 |
Production-quality outputs | Balanced composition, natural limbs |