OneLab TTS Library

OneLab is a powerful Python library for Text-to-Speech (TTS) and Podcast generation, built on top of ChatterboxTurboTTS. It allows you to easily convert text to speech using various voices and create multi-speaker podcasts with custom logging and progress tracking.

Features

• Text-to-Speech Conversion: Convert individual text segments to audio using specific voices.
• Podcast Generation: Create full conversations with multiple speakers.
• Dynamic Voice Loading: Automatically detects available voices from the package's sample directory.
• Custom Logging: Detailed logging and progress tracking during generation.
• Modular Design: Clean and extensible architecture.

Installation

You can install the package using pip:

pip install onelab

Or install directly from the source:

git clone https://github.com/satadeep3927/onelab.git
cd onelab
pip install .

Usage

Initialization

To use the library, simply initialize the OneLab class. The library automatically locates the bundled voice samples.

from onelab import OneLab

# Initialize OneLab
onelab = OneLab(device="cpu") 

Listing Available Voices

You can list all voices detected in the sample directory:

voices = onelab.tts.list_voices()
print("Available voices:", voices)

Converting Text to Speech

To convert a single line of text:

wav = onelab.tts.convert("Hello, world!", voice="charlie")
# Save the output
import torchaudio
torchaudio.save("output.wav", wav, onelab.tts.model.sr)

Creating a Podcast

To create a multi-speaker podcast, define your conversation segments and use the podcast.create method:

from onelab import ConversationInput

data: ConversationInput = {
    "segments": [
        {"voice": "charlie", "text": "Hello Emilia!"},
        {"voice": "emilia", "text": "Hi Charlie! How are you?"},
    ]
}

# Generate the podcast
podcast_wav = onelab.tts.podcast.create(data)

# Save the podcast
torchaudio.save("podcast.wav", podcast_wav, onelab.tts.model.sr)

API Reference

OneLab

The main entry point for the library.

• __init__(device: str = "cpu"): Initializes the library.

TextToSpeech (accessed via onelab.tts)

Handles TTS operations.

• list_voices() -> List[str]: Returns a list of available voice names.
• convert(text: str, voice: str) -> torch.Tensor: Generates audio for the given text and voice.

Podcast (accessed via onelab.tts.podcast)

Handles podcast generation.

• create(args: ConversationInput) -> torch.Tensor: Generates a concatenated audio waveform for the entire conversation.

License

MIT