CosyVoice2

About CosyVoice2

CosyVoice2 is a next-generation speech synthesis model from FunAudioLLM (Alibaba). It delivers natural-sounding zero-shot voice cloning across multiple languages with streaming capability for low-latency applications. Built on a finite scalar quantization approach, it achieves excellent voice similarity with just a few seconds of reference audio.

Key Features

Use Cases

How to Use CosyVoice2

1. 1

   Sign up and claim free credits

   Create a free TextToSpeechAI account to claim your starter credits, or try the demo first. No GPU or local CosyVoice2 install is needed - everything runs on our infrastructure.

2. 2

   Select CosyVoice2 and add a reference clip

   Choose CosyVoice2 as your engine, then upload a clean 3-10 second reference recording of the voice you want to clone. CosyVoice2 will extract the speaker characteristics for zero-shot multilingual cloning.

3. 3

   Enter your text in any supported language

   Type or paste your script in Chinese, English, Japanese, Korean, or Cantonese. CosyVoice2 supports cross-lingual synthesis, so the cloned voice can speak a language different from the reference clip.

4. 4

   Generate the speech

   Click generate and CosyVoice2 synthesizes natural, multilingual speech in the cloned voice, usually within seconds for short text. Premium-tier usage costs 25 credits per 1,000 characters.

5. 5

   Download or use the API

   Download the finished audio as MP3 or WAV from your history, or automate CosyVoice2 voice cloning at scale through the TextToSpeechAI REST API.

CosyVoice2 API

Generate speech programmatically using the TextToSpeechAI REST API.

curl -X POST "https://api.texttospeechai.com/v1/generate/" \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "CosyVoice2 delivers natural multilingual speech with zero\u002Dshot voice cloning capability.",
    "voice": "en_US-lessac-medium"
  }'

Frequently Asked Questions