Home › Audio Models › coqui/XTTS-v2

coqui/XTTS-v2 - AI Audio Models Tool

Overview

coqui/XTTS-v2 is a high-quality, few‑shot text‑to‑speech (TTS) model published on Hugging Face that focuses on fast voice cloning and cross‑language synthesis. The model can clone a speaker’s voice from very short reference audio (advertised as a 6‑second sample) and synthesize speech in a different target language while preserving speaker identity and prosody. XTTS‑v2 emphasizes improved speaker conditioning, stability, and style/emotion transfer compared with its predecessor. XTTS‑v2 supports multi‑language output (17 languages listed on the model page) and is supplied as a ready-to-use text‑to‑speech pipeline on Hugging Face, allowing developers to prototype voice cloning and multilingual TTS quickly. According to the Hugging Face model page, the model has been widely adopted (several million downloads) and is commonly used for research, prototyping voice assistants, and creating style‑aware synthetic speech. For technical details, usage notes, and the latest changes, refer to the model card on Hugging Face.

Model Statistics

  • Downloads: 4,680,558
  • Likes: 3294
  • Pipeline: text-to-speech

License: other

Model Details

Architecture and components: XTTS‑v2 is a neural text‑to‑speech model that combines speaker conditioning and style/emotion control to produce natural-sounding synthetic voices. The model is presented on Hugging Face as a text-to-speech pipeline; the model card highlights improved speaker conditioning and stability relative to XTTS v1. It is designed for few‑shot (6‑second) speaker cloning and for cross‑lingual synthesis where a speaker sample in one language is used to generate speech in another. Capabilities: XTTS‑v2 supports voice cloning from short reference audio, emotion and style transfer, speaker identity preservation across target languages, and multi‑language output (17 languages supported). The model is intended for prototyping and research; exact training data, parameter counts, and low‑level architecture (e.g., specific vocoder or backbone) are not published on the Hugging Face model card and therefore should be checked in the model repository or official documentation if required. Deployment: XTTS‑v2 is available through the Hugging Face model hub as a text-to-speech pipeline (see model page). Developers can run inference via the transformers pipeline or through Hugging Face inference endpoints; verify model-card usage examples for recommended input parameters for reference audio, style tokens, and sampling settings.

Key Features

  • Few‑shot voice cloning from a 6‑second reference audio sample
  • Cross‑language synthesis: preserve speaker identity across languages
  • Emotion and style transfer to change expressive characteristics
  • Improved speaker conditioning and stability over XTTS v1
  • Supports 17 output languages (listed on the model page)
  • Ready-to-run Hugging Face text-to-speech pipeline for fast prototyping
  • Widely used community model with millions of downloads

Example Usage

Example (python):

from transformers import pipeline

# Basic example: synthesize plain text using the Hugging Face pipeline.
# For voice cloning (reference audio), consult the model card for exact parameter names.

tts = pipeline("text-to-speech", model="coqui/XTTS-v2")

# Synthesize a short phrase
result = tts("Hello — this is a quick test of XTTS-v2.")

# The pipeline returns audio bytes or a dict depending on the environment; save to a file.
# If result is a dict with 'wav', write result['wav']. If it's bytes, write result directly.
output_path = "xtts_v2_output.wav"

if isinstance(result, dict) and "wav" in result:
    audio_bytes = result["wav"]
else:
    audio_bytes = result

with open(output_path, "wb") as f:
    f.write(audio_bytes)

print(f"Saved synthesized audio to {output_path}. See the model card for reference-audio cloning examples.")

Benchmarks

Hugging Face downloads: 4,680,558 (Source: https://huggingface.co/coqui/XTTS-v2)

Hugging Face likes: 3,294 (Source: https://huggingface.co/coqui/XTTS-v2)

Pipeline: text-to-speech (Source: https://huggingface.co/coqui/XTTS-v2)

Last Refreshed: 2026-01-09

Key Information

  • Category: Audio Models
  • Type: AI Audio Models Tool
Visit Official Website

Related Tools

OpenVoice

OpenVoice is a versatile instant voice cloning framework that allows users to generate speech in multiple languages using only a short audio clip from a reference speaker. The tool provides granular control over voice styles, such as emotion, accent, rhythm, pauses, and intonation, and supports zero-shot cross-lingual voice cloning, enabling users to clone voices across different languages without needing training data for those languages.
Parler-TTS

A text-to-speech inference and training library for generating high-fidelity speech from text, offering an open-source solution for TTS applications.
SpeechBrain

An all-in-one open-source conversational AI toolkit based on PyTorch offering speech recognition, text-to-speech, speaker recognition, and more.
Whisper Large

A robust speech recognition model based on a Transformer architecture that supports multilingual transcription, speech translation, and language identification.
openai/whisper-large-v3-turbo

A finetuned, pruned version of Whisper large-v3 for automatic speech recognition and speech translation. This model reduces the number of decoding layers from 32 to 4 to achieve much faster inference, with only a minor quality trade-off. It supports 99 languages and integrates with Hugging Face Transformers for efficient transcription and translation.
OpenVoice V2

OpenVoice V2 is an advanced text-to-speech model that provides instant voice cloning with accurate tone color reproduction and flexible voice style control. It supports zero-shot cross-lingual synthesis in multiple languages and has improved audio quality over its previous version. Released under the MIT License, it is geared towards both research and commercial use.