ComfyUI-Orpheus-TTS

ComfyUI-Orpheus-TTS

This project adds high-quality Text-to-Speech capabilities to ComfyUI using the Orpheus TTS model. Create natural-sounding voices with emotional expressions, multilingual support, and audio effects.

image

Features

• 🎙️ High-quality, natural-sounding speech synthesis
• 🎭 Support for emotional expressions and paralinguistic elements
• 👥 Multiple voice options (tara, leah, jess, leo, dan, mia, zac, zoe, etc.)
• 📝 Long text handling with automatic chunking for consistent output
• 🎛️ Professional audio effects:
  • Pitch shifting (-12 to +12 semitones)
  • Speed adjustment (0.5x to 2.0x speed)
  • Volume control with anti-clipping protection
  • Audio normalization option
  • Reverb with adjustable room size and amount
  • Echo with configurable delay and decay
• 🌐 Optional support for private Hugging Face models
• 💻 Cross-platform: Works on Windows, Linux/WSL, and macOS

Installation

1. Install the Extension

Clone this repository into your ComfyUI's custom_nodes directory:

cd ComfyUI/custom_nodes
git clone https://github.com/ShmuelRonen/ComfyUI-Orpheus-TTS.git

2. Install Required Python Dependencies

pip install torch numpy soundfile transformers huggingface_hub nltk snac

For WSL 2, you may need to install directly from the GitHub repository:

pip install git+https://github.com/hubertsiuzdak/snac.git

3. Install SoX (Required for Audio Effects)

Windows

1. Download SoX for Windows from the official SourceForge page

   • Download the .exe installer (e.g., sox-14.4.2-win32.exe)

2. Run the installer:

   • Follow the installation prompts
   • Important: Note the installation directory (default is usually C:\Program Files (x86)\sox-14-4-2\)

3. No need to add to PATH - the extension uses the direct path to SoX

WSL 2 (Ubuntu)

sudo apt-get update
sudo apt-get install sox

macOS

brew install sox

4. Restart ComfyUI

After installing all required components, restart ComfyUI to load the extension.

Configuration

Hugging Face Authentication (Optional)

To access private models on Hugging Face, create a file named hf_config.json in the extension directory and insert your HF Token KEY:

{
  "token": "YOUR_HUGGING_FACE_TOKEN_HERE"
}

3. Save the file and restart ComfyUI

Your token will be used to authenticate with Hugging Face when downloading models. This is only required if you're using private models or if you need higher rate limits.

Nodes

Orpheus TTS Model Loader

Loads the required models for Orpheus TTS.

Inputs:

• snac_model_path (optional): Path to SNAC model (default: "hubertsiuzdak/snac_24khz")
• orpheus_model_path (optional): Path to Orpheus model (default: "canopylabs/orpheus-3b-0.1-ft")

Outputs:

• model: Model reference to be passed to the generate node

Orpheus TTS Generate

Generates speech from text input.

Inputs:

• model: Model reference from the loader node
• text: The text to convert to speech
• voice: Voice style to use (tara, leah, jess, leo, dan, mia, zac, zoe, etc.)
• language (optional): Language for multilingual output (en, fr, es, etc.)
• max_chunk_size (optional): Maximum chunk size for long text processing

Outputs:

• audio: Audio data to be passed to preview or effects nodes

Orpheus Audio Effects

Applies high-quality audio processing to the generated speech.

Inputs:

• audio: Audio data from the generate node
• pitch_shift: Semitone adjustment (-12 to +12)
• speed_factor: Playback speed modifier (0.5x to 2.0x)
• sox_path (optional): Custom path to SoX executable
• gain_db (optional): Volume adjustment in decibels
• use_limiter (optional): Enable/disable limiter for positive gain
• normalize_audio (optional): Enable/disable