Maya Voice AI

A demonstration of the Maya1 voice AI model, which generates realistic voice audio from text input with emotional and descriptive control.

Description

This project showcases the Maya1 model, an open-source voice AI that can synthesize speech with specified voice characteristics and emotions. The demo script generates audio from a text prompt using a voice description, producing high-quality voice output.

Features

• Text-to-speech synthesis with voice descriptions
• Emotional and stylistic voice control
• High-quality audio output at 24kHz
• Uses advanced language models and audio codecs

Installation

1. Ensure you have Python 3.8 or higher installed.

2. Install the required dependencies:

   pip install torch transformers snac soundfile

3. The models will be automatically downloaded when you run the script:

   • Maya1 model: maya-research/maya1
   • SNAC codec: hubertsiuzdak/snac_24khz

Usage

Run the demo script to generate voice audio:

python maya1_demo.py

The script will:

• Load the Maya1 model and SNAC codec
• Generate voice based on the predefined description and text
• Save the output as output.wav

You can modify the description and text variables in maya1_demo.py to customize the voice generation.

Requirements

• Python 3.8+
• PyTorch
• Transformers library
• SNAC audio codec
• Soundfile for audio I/O

Output

The generated audio is saved as output.wav in the project directory. The audio is encoded at 24kHz sample rate.

License

This project uses open-source models and libraries. Please refer to the individual model licenses for usage terms.