Local Orpheus

A local text-to-speech system using the Orpheus AI model.

Installation

# Install all dependencies
pip install -r requirements.txt

Backend Options [lmstudio, ollama, local]

• LM Studio (default): Download here
• Ollama: Download here
• Local: Uses llama-cpp-python
  • Default runs on CPU
  • For CUDA: pip install llama-cpp-python --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu121
  • Windows/Cuda Issue: https://github.com/abetlen/llama-cpp-python/issues/2001#issuecomment-2801082928

Base Model

isaiahbjork/orpheus-3b-0.1-ft-Q4_K_M-GGUF

Usage

# Default CLI Loop (LM Studio)
python TTS.py

# Change backend
python TTS.py --backend local

# Specify text
python TTS.py --text "Hello, this is Orpheus."

# Change voice (options: tara, leah, jess, leo, dan, mia, zac, zoe)
python TTS.py --voice leo

# Save to file without playing
python TTS.py --text "Save this to a file" --output "my_speech.wav" --no-play

Output files are saved to the "outputs" directory by default.