Trigger Word: south-indian-beauty

Ecosystem: Boogu (High-Fidelity Image Architecture)

Training Engine: AI-Toolkit

Dataset Configuration: 10 Epochs | 3,000 Precision Steps

Recommended Inference Settings

To get the most out of this symmetric Rank 32 Boogu LoRA, use these baseline configurations:

• LoRA Weight: 0.75 - 0.95 (The constant learning rate allows the model to scale smoothly without breaking the base style).

• Sampling Steps: 25 - 35

• CFG Scale: 3.5 - 5.0

• Sampler: Euler or DPM++ 2M

• Resolution: Highly optimized for 1024x1024 native or matching vertical crops.