FunAudioLLM/CosyVoice2-0.5B文本转语音

正式环境

https://api.302.ai

POST

/siliconflow/v1/audio/speech

CosyVoice 2 is a streaming speech synthesis model based on large language models, featuring a unified framework design for both streaming and non-streaming modes. The model enhances the encoding efficiency of speech tokens through Finite Scalar Quantization (FSQ), simplifies the language model architecture for text-to-speech, and introduces a chunk-aware causal flow matching model to support various synthesis scenarios. In streaming mode, the model achieves ultra-low latency of 150ms while maintaining nearly the same synthesis quality as the non-streaming mode.
Supported model: FunAudioLLM/CosyVoice2-0.5B
Paid Input: 0.007PTC/1000Token

请求参数

Header 参数

Body 参数application/json

示例

{
  "model": "FunAudioLLM/CosyVoice2-0.5B",
  "input": "你能带着幸福的情绪说出来吗？<|endofprompt|>今天真是开心，春节快到了！太开心了，春节快到了！[笑声][呼吸]。",
  "voice": "FunAudioLLM/CosyVoice2-0.5B:diana",
  "response_format": "mp3",
  "sample_rate": 44100,
  "stream": true,
  "speed": 1,
  "gain": 0
}

示例代码

Shell

JavaScript

Java

Swift

PHP

Python

HTTP

Objective-C

Ruby

OCaml

Dart

curl --location --request POST 'https://api.302.ai/siliconflow/v1/audio/speech' \
--header 'Content-Type: application/json' \
--data-raw '{
  "model": "FunAudioLLM/CosyVoice2-0.5B",
  "input": "你能带着幸福的情绪说出来吗？<|endofprompt|>今天真是开心，春节快到了！太开心了，春节快到了！[笑声][呼吸]。",
  "voice": "FunAudioLLM/CosyVoice2-0.5B:diana",
  "response_format": "mp3",
  "sample_rate": 44100,
  "stream": true,
  "speed": 1,
  "gain": 0
}'

返回响应

🟢200成功

application/json

Body

示例

{
    "url": "https://file.302.ai/gpt/imgs/20250804/5c49c8dc28cf4965b35bf22c93881885.mp3"
}

修改于 2025-08-13 06:39:52

Create text-to-speech request

fnlp/MOSS-TTSD-v0.5文本转语音