shouhsu/orpheus-3b-0.1-ft_elon_v1

Orpheus-3B-0.1-ft-elon-v1

This is an improved fine-tuned version of Canopy Labs' Orpheus-TTS model, further optimized for Elon Musk's voice characteristics.

Model Description

• Base Model: canopylabs/orpheus-tts-0.1-pretrained
• Fine-tuning Target: Optimized for Elon Musk's voice characteristics (improved version)
• Model Architecture: 3B parameter Transformer model
• Sample Rate: 24kHz

Training Details

• Training Method: LoRA fine-tuning
• LoRA Parameters:
  • r: 32
  • alpha: 64
  • dropout: 0.15
• Target Modules: ["q_proj", "v_proj", "k_proj", "o_proj", "gate_proj", "up_proj", "down_proj"]
• Learning Rate: 0.0001
• Batch Size: 2
• Gradient Accumulation Steps: 8

Key Improvements from v0

• Enhanced speech naturalness and fluency
• Improved voice characteristic accuracy
• Better long-sentence generation capability
• Enhanced emotional expression in speech

Usage