SonicVox-Multilingual

🎙️ SonicVox-Multilingual is a multilingual Text-to-Speech (TTS) model built on top of Chatterbox Multilingual by Resemble AI.

It is designed to generate natural, expressive speech across multiple languages with support for zero-shot voice cloning using a reference audio sample.

✨ Features

• 🌍 Multilingual Text-to-Speech
• 🎙️ Natural and expressive speech generation
• 🗣️ Zero-shot voice cloning
• 🔊 High-quality speech synthesis
• 🎚️ Voice conditioning using reference audio
• ⚡ GPU-accelerated inference
• 🧩 Based on Chatterbox Multilingual V3

🌎 Supported Languages

SonicVox-Multilingual is based on the multilingual Chatterbox model.

Supported languages depend on the underlying checkpoint and include multiple languages such as:

• English
• Spanish
• French
• German
• Italian
• Portuguese
• Hindi
• Telugu
• Tamil
• Kannada
• Malayalam
• Chinese
• Japanese
• Korean
• Arabic
• Russian
• Turkish
• Polish

Always verify the supported language IDs with the specific Chatterbox checkpoint being used.

🎤 Voice Cloning

SonicVox-Multilingual supports voice cloning using a reference audio file.

Example:

import torchaudio as ta
from chatterbox.mtl_tts import ChatterboxMultilingualTTS

model = ChatterboxMultilingualTTS.from_pretrained(device="cuda")

text = "Welcome to SonicVox Multilingual."

wav = model.generate(
    text,
    language_id="en",
    audio_prompt_path="reference.wav"
)

ta.save("output.wav", wav, model.sr)