🇨🇳中文 | 🌐English | 📖文档/Docs | 🤖模型/Models

Online Demo

Parrots: ASR and TTS toolkit

Introduction

Parrots, Automatic Speech Recognition(ASR), Text-To-Speech(TTS) toolkit, support Chinese, English, Japanese, etc.

parrots实现了语音识别和语音合成模型一键调用，开箱即用，支持中英文。

Features

ASR: Automatic Speech Recognition based on distilwhisper, supporting Chinese, English, and other languages
TTS: Text-to-Speech based on GPT-SoVITS, supporting Chinese, English, Japanese, and other languages
IndexTTS2: Integrated IndexTTS2 model for emotionally expressive and duration-controlled zero-shot speech synthesis
- Precise speech duration control
- Emotion and speaker identity disentanglement for independent control
- Multiple emotion control methods: audio reference, emotion vectors, text descriptions
- Highly expressive emotional speech synthesis
Streaming TTS: Support for streaming speech synthesis with low latency for real-time audio output

Install

pip install torch # or conda install pytorch
pip install -r requirements.txt
pip install parrots

pip install torch # or conda install pytorch
git clone https://github.com/shibing624/parrots.git
cd parrots
python setup.py install

Demo

HuggingFace Demo: https://huggingface.co/spaces/shibing624/parrots

run example: examples/tts_gradio_demo.py to see the demo:

python examples/tts_gradio_demo.py

Usage

ASR(Speech Recognition)

example: examples/demo_asr.py

import os
import sys

sys.path.append('..')
from parrots import SpeechRecognition

pwd_path = os.path.abspath(os.path.dirname(__file__))

if __name__ == '__main__':
    m = SpeechRecognition()
    r = m.recognize_speech_from_file(os.path.join(pwd_path, 'tushuguan.wav'))
    print('[提示] 语音识别结果：', r)

output:

{'text': '北京图书馆'}

TTS(Speech Synthesis)

GPT-SoVITS Basic Usage

example: examples/demo_tts.py

from parrots import TextToSpeech

# Initialize TTS model (no manual path configuration needed)
m = TextToSpeech(
    speaker_model_path="shibing624/parrots-gpt-sovits-speaker-maimai",
    speaker_name="MaiMai",
    device="cpu",  # or "cuda" for GPU
    half=False     # Set to True for half precision acceleration
)

# Generate speech
m.predict(
    text="Hello, welcome to Beijing. This is a demo of synthesized audio. Welcome to Beijing!",
    text_language="auto",  # Auto-detect language, or specify "zh", "en", "ja"
    output_path="output_audio.wav"
)

output:

Save audio to output_audio.wav

Streaming TTS (Low Latency)

Support for streaming speech synthesis, suitable for real-time conversation scenarios:

from parrots import TextToSpeech
import soundfile as sf
import numpy as np

m = TextToSpeech(
    speaker_model_path="shibing624/parrots-gpt-sovits-speaker-maimai",
    speaker_name="MaiMai",
)

# Stream generate speech
audio_chunks = []
for audio_chunk in m.predict_stream(
    text="This is a longer text that will be synthesized into speech in a streaming manner.",
    text_language="en",
    stream_chunk_size=20  # Control latency, smaller = lower latency
):
    audio_chunks.append(audio_chunk)
    # You can play audio_chunk in real-time here

# Save complete audio
full_audio = np.concatenate(audio_chunks)
sf.write("streaming_output.wav", full_audio, m.sampling_rate)

Log Management

Control log output level:

from parrots import TextToSpeech
from parrots.log import set_log_level, logger

# Set log level
set_log_level("INFO")  # Options: DEBUG, INFO, WARNING, ERROR

m = TextToSpeech(
    speaker_model_path="shibing624/parrots-gpt-sovits-speaker-maimai",
    speaker_name="MaiMai",
)

# Use logger
logger.info("Starting speech synthesis...")
m.predict(
    text="Hello, world!",
    text_language="en",
    output_path="output.wav"
)

IndexTTS2 Advanced Usage

IndexTTS2 is a breakthrough model for emotionally expressive and duration-controlled autoregressive zero-shot speech synthesis.

example: examples/demo_indextts.py

1. Basic Voice Cloning (Single Reference Audio)

from parrots.indextts import IndexTTS2

tts = IndexTTS2()
text = "Hello, welcome to Beijing. This is a demo of synthesized audio."
tts.infer(text=text, output_path="gen.wav", verbose=True)

2. Emotional Speech Synthesis (With Emotion Reference Audio)

Use a separate emotional reference audio to control the emotional expression:

from parrots.indextts import IndexTTS2

tts = IndexTTS2()
text = "The tavern is unconscionable, starting to auction rooms, ah, a bunch of fools."
tts.infer(
   speak_reference_audio_path_or_name='examples/voice_07.wav',  # Speaker timbre reference
   text=text,
   output_path="gen.wav",
   emo_reference_audio_path="examples/emo_sad.wav",  # Emotion reference audio
   verbose=True
)

3. Adjust Emotion Intensity

Control emotion influence with emo_alpha parameter (range 0.0-1.0):

from parrots.indextts import IndexTTS2

tts = IndexTTS2()
text = "The tavern is unconscionable, starting to auction rooms, ah, a bunch of fools."
tts.infer(
   speak_reference_audio_path_or_name='examples/voice_07.wav',
   text=text,
   output_path="gen.wav",
   emo_reference_audio_path="examples/emo_sad.wav",
   emo_alpha=0.6,  # 60% emotion intensity
   verbose=True
)

4. Emotion Vector Control

Directly provide an 8-dimensional emotion vector for precise control, in order: [happy, angry, sad, afraid, disgusted, melancholic, surprised, calm]

from parrots.indextts import IndexTTS2

tts = IndexTTS2()
text = "Wow! This drop rate is so high! I'm blessed by luck!"
tts.infer(
   speak_reference_audio_path_or_name='examples/voice_10.wav',
   text=text,
   output_path="gen.wav",
   emo_vector=[0, 0, 0, 0, 0, 0, 0.45, 0],  # Surprised emotion
   use_random=False,
   verbose=True
)

5. Text-Based Emotion Control

Enable use_emo_text to automatically infer emotions from text content:

from parrots.indextts import IndexTTS2

tts = IndexTTS2()
text = "Hide quickly! He's coming! He's coming to catch us!"
tts.infer(
   speak_reference_audio_path_or_name='examples/voice_12.wav',
   text=text,
   output_path="gen.wav",
   emo_alpha=0.6,
   use_emo_text=True,  # Enable text emotion analysis
   use_random=False,
   verbose=True
)

6. Independent Emotion Text Description

Provide a separate emotion description via emo_text parameter:

from parrots.indextts import IndexTTS2

tts = IndexTTS2()
text = "Hide quickly! He's coming! He's coming to catch us!"
emo_text = "You scared me to death! Are you a ghost?"  # Independent emotion description
tts.infer(
   speak_reference_audio_path_or_name='examples/voice_12.wav',
   text=text,
   output_path="gen.wav",
   emo_alpha=0.6,
   use_emo_text=True,
   emo_text=emo_text,
   use_random=False,
   verbose=True
)

Pinyin Control Notes:

IndexTTS2 supports mixed modeling of Chinese characters and Pinyin. When precise pronunciation control is needed, provide text with specific Pinyin annotations. Note: Pinyin control only supports valid Chinese Pinyin combinations.

Example:

text = "之前你做DE5很好，所以这一次也DEI3做DE2很好才XING2，如果这次目标完成得不错的话，我们就直接打DI1去银行取钱。"

命令行模式（CLI）

支持通过命令行方式执行ARS和TTS任务，代码：cli.py

> parrots -h                                    

NAME
    parrots

SYNOPSIS
    parrots COMMAND

COMMANDS
    COMMAND is one of the following:

     asr
       Entry point of asr, recognize speech from file

     tts
       Entry point of tts, generate speech audio from text

run：

pip install parrots -U
# asr example
parrots asr -h
parrots asr examples/tushuguan.wav

# tts example
parrots tts -h
parrots tts "你好，欢迎来北京。welcome to the city." output_audio.wav

asr、tts是二级命令，asr是语音识别，tts是语音合成，默认使用的模型是中文模型
各二级命令使用方法见parrots asr -h
上面示例中examples/tushuguan.wav是asr方法的audio_file_path参数，输入的音频文件（required）

Release Models

ASR

BELLE-2/Belle-distilwhisper-large-v2-zh

IndexTTS2

IndexTeam/IndexTTS-2 - Latest emotion expression and duration control model
IndexTeam/IndexTTS-1.5 - Improved stability and English performance
IndexTeam/Index-TTS - Initial version

GPT-SoVITS TTS

shibing624/parrots-gpt-sovits-speaker

speaker name	说话人名	character	角色特点	language	语言
KuileBlanc	葵·勒布朗	lady	标准美式女声	en	英
LongShouRen	龙守仁	gentleman	标准美式男声	en	英
MaiMai	卖卖	singing female anchor	唱歌女主播声	zh	中
XingTong	星瞳	singing ai girl	活泼女声	zh	中
XuanShen	炫神	game male anchor	游戏男主播声	zh	中
KusanagiNene	草薙寧々	loli	萝莉女学生声	ja	日

shibing624/parrots-gpt-sovits-speaker-maimai

speaker name	说话人名	character	角色特点	language	语言
MaiMai	卖卖	singing female anchor	唱歌女主播声	zh	中

Contact

Issue(建议)：
邮件我：xuming: xuming624@qq.com
微信我：加我微信号：xuming624, 进Python-NLP交流群，备注：姓名-公司名-NLP

Citation

如果你在研究中使用了parrots，请按如下格式引用：

@misc{parrots,
  title={parrots: ASR and TTS Tool},
  author={Ming Xu},
  year={2024},
  howpublished={\url{https://github.com/shibing624/parrots}},
}

License

授权协议为 The Apache License 2.0，可免费用做商业用途。请在产品说明中附加parrots的链接和授权协议。

Contribute

项目代码还很粗糙，如果大家对代码有所改进，欢迎提交回本项目，在提交之前，注意以下两点：

在tests添加相应的单元测试
使用python -m pytest来运行所有单元测试，确保所有单测都是通过的

之后即可提交PR。

Reference

ASR(Speech Recognition)

TTS(Speech Synthesis)

IndexTeam/IndexTTS - IndexTTS2 emotion expression and duration control
coqui-ai/TTS
keonlee9420/Expressive-FastSpeech2
TensorSpeech/TensorflowTTS
RVC-Boss/GPT-SoVITS

Changelog

v0.3.0 (2025-11)

🔥 Integrated IndexTTS2 model for emotionally expressive and duration-controlled zero-shot speech synthesis
✨ Support multiple emotion control methods: audio reference, emotion vectors, text descriptions
✨ Implemented emotion and speaker identity disentanglement for independent control
✨ Support Pinyin mixed modeling for precise pronunciation control
🐛 Fixed transformers 4.50+ compatibility issues
🐛 Fixed dictionary parameter access errors
📝 Added IndexTTS2 usage examples and documentation

v0.2.0 (2025-10)

✨ Added streaming TTS feature with low-latency real-time speech synthesis
✨ Added unified logging system (based on loguru)
🐛 Fixed PyTorch 2.0+ weight_norm deprecation warning
🐛 Fixed torch.stft return_complex=False deprecation warning
🐛 Fixed librosa resample and time_stretch warnings
🔧 Optimized model loading mechanism, no need to manually add sys.path
📝 Improved documentation and example code

v0.1.0 (2024-12)

🎉 Initial release
✨ Support for ASR (Automatic Speech Recognition)
✨ Support for TTS (Text-to-Speech)
✨ Support for Chinese, English, and Japanese

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Parrots: ASR and TTS toolkit

Introduction

Features

Install

Demo

Usage

ASR(Speech Recognition)

TTS(Speech Synthesis)

GPT-SoVITS Basic Usage

Streaming TTS (Low Latency)

Log Management

IndexTTS2 Advanced Usage

命令行模式（CLI）

Release Models

ASR

IndexTTS2

GPT-SoVITS TTS

Contact

Citation

License

Contribute

Reference

ASR(Speech Recognition)

TTS(Speech Synthesis)

Changelog

v0.3.0 (2025-11)

v0.2.0 (2025-10)

v0.1.0 (2024-12)

FilesExpand file tree

README_EN.md

Latest commit

History

README_EN.md

File metadata and controls

Parrots: ASR and TTS toolkit

Introduction

Features

Install

Demo

Usage

ASR(Speech Recognition)

TTS(Speech Synthesis)

GPT-SoVITS Basic Usage

Streaming TTS (Low Latency)

Log Management

IndexTTS2 Advanced Usage

命令行模式（CLI）

Release Models

ASR

IndexTTS2

GPT-SoVITS TTS

Contact

Citation

License

Contribute

Reference

ASR(Speech Recognition)

TTS(Speech Synthesis)

Changelog

v0.3.0 (2025-11)

v0.2.0 (2025-10)

v0.1.0 (2024-12)