VoiceHub: A Unified Inference Interface for TTS Models

🛠️ Installation

uv venv --python 3.12
source .venv/bin/activate
uv pip install voicehub

📚 Usage

VoiceHub provides a simple, unified interface for working with various Text-to-Speech (TTS) models. Below are examples showing how to use different supported TTS models with the same consistent approach.

OrpheusTTS Model

from voicehub.automodel import AutoInferenceModel

model = AutoInferenceModel.from_pretrained(
    model_type="orpheustts",  # or "dia" or "vui"
    model_path="canopylabs/orpheus-3b-0.1-ft",
    device="cuda",
)

output = model(
    text="Hey, here is some random stuff, the text the less likely the model can cope!",
    voice="tara",
    output_file="output.wav",
)

DiaTTS Model

from voicehub.automodel import AutoInferenceModel

model = AutoInferenceModel.from_pretrained(
    model_type="dia",  # or "dia" or "vui"
    model_path="dia/dia-100m-base.pt",
    device="cuda",
)

output = model(
    text="Hey, here is some random stuff, the text the less likely the model can cope!",
    output_file="output.wav",
)

VuiTTS Model

from voicehub.automodel import AutoInferenceModel

model = AutoInferenceModel.from_pretrained(
    model_type="vui",  # or "dia" or "vui"
    model_path="fluxions/vui",
    device="cuda",
)

output = model(
    text="Hey, here is some random stuff, the text the less likely the model can cope!",
    output_file="output.wav",
)

Llasa Model

from voicehub.automodel import AutoInferenceModel

model = AutoInferenceModel.from_pretrained(
    model_type="llasa",  # or "dia" or "vui"
    model_path="HKUSTAudio/Llasa-1B-Multilingual",
    device="cuda",
)

output = model(
    text="Hey, here is some random stuff, the text the less likely the model can cope!",
    output_file="output.wav",
)

Llasa Voice Clone Model

from voicehub.automodel import AutoInferenceModel

model = AutoInferenceModel.from_pretrained(
    model_type="llasa_voice_clone",  # or "dia" or "vui"
    model_path="HKUSTAudio/Llasa-1B-Multilingual",
    device="cuda",
)

output = model(
    text="Hey, here is some random stuff, the text the less likely the model can cope!",
    output_file="output.wav",
)

Chatterbox Model

from voicehub.automodel import AutoInferenceModel

model = AutoInferenceModel.from_pretrained(
    model_type="chatterbox",  # or "dia" or "vui"
    model_path="ResembleAI/chatterbox",
    device="cuda",
)

output = model(
    text="Hey, here is some random stuff, the text the less likely the model can cope!",
    output_file="output.wav",
)

🤗 Contributing

uv pip install pre-commit
pre-commit install
pre-commit run --all-files

📝 Acknowledgments

• Orpheus-TTS
• Dia
• Vui
• Llasa
• XCodec2
• Chatterbox