Universal Text-to-Speech Testing and Comparison

The UTTS (Universal Text-to-Speech) project provides a comprehensive interface designed to seamlessly test and compare various text-to-speech models. This robust framework currently supports multiple leading TTS services, enhancing accessibility and flexibility for developers.

Supported Text-to-Speech Models

The UTTS interface supports the following text-to-speech services:

• OpenAI TTS
• ElevenLabs
• Kokoro
• Orpheus
• Zyphra/Zonos
• Hume AI
• Cartesia

Quick Start

To quickly experiment with TTS models, users can initiate the project using Google Colab. A sample notebook is available for demonstration:

Open In Colab

For Jupyter Notebook users, the following Python snippet demonstrates how to generate audio output:

from IPython.display import Audio
import utts

client = utts.UTTSClient(elevenlabs_api_key="your_elevenlabs_api_key")
audio = client.elevenlabs.generate('Hello, world!')
Audio(audio)

For batch generation capabilities, the utts.utils.batch_generate can be utilized.

Development Information

For developers intending to contribute or customize the UTTS project, the following prerequisites must be met:

• Python 3.11.12 or higher
• uv - A modern installer for Python packages
• Make - Required for executing development commands

Setup Instructions

To set up the UTTS environment for development, execute the following commands:

git clone https://github.com/arch1baald/utts.git
cd utts
make install-dev

Development Commands

Access to all available commands can be achieved via:

make help

To ensure code quality, linting and type checking can be performed with:

make lint

The UTTS project simplifies the evaluation and comparison of TTS models, making it easier for researchers and developers to incorporate text-to-speech functionalities into their applications.