GraphCanon
Home/Inference & Serving/PPASR
PPASR logo

PPASR

yeyupiaoling/PPASR

PaddlePaddle-based end-to-end Chinese speech recognition framework

GitHub

GraphCanon updated 3w · GitHub synced 3w

872 stars129 forksLast push 8mo Python Apache-2.0

Decision brief

PPASR is an end-to-end Chinese speech recognition framework powered by PaddlePaddle, offering extensive support for advanced models like DeepSpeech2, Conformer and SqueezeFormer.

Good fit when

  • When you are working with Chinese language data as PPASR specifically supports a wide range of Chinese-based models and decoding methods
  • In scenarios requiring both stream and non-streaming recognition abilities since PPASR configures support for various models and decoders suitable for different recognition needs

Avoid when

  • When the project does not involve working with Python, as PPASR runs on a Python environment and is not compatible with other languages
  • For applications that need immediate access to pre-trained models without requiring an additional engagement like joining a knowledge community for download access

Observed Jul 17, 2026 · Source: enrich:decision_facts

PPASR alternativesTrust report

Verify the decision

Maintenance and security

Full trust report
Maintenance
Slowing (225d since push)
As of 3w
Provenance
Not a fork · Personal account
As of 3w
Security (OSV)
6 low (6 low)
As of 1mo

Public GitHub metadata and optional OSV scans. Signals, not a guarantee. Trust methodology.

Install

pip install PPASR
PyPI

Similar tools

Same-category neighbours. No typed graph edges are catalogued for this tool yet.

Evidence and technical details

Sourced facts, taxonomy, compatibility claims, README excerpt, and machine-readable endpoints.

Overview

PPASR is an automatic speech recognition framework based on PaddlePaddle that supports model training and deployment for both stream and non-streaming scenarios, with a focus on DeepSpeech2, Conformer, and Squeezeformer models.

Capability facts

Languages
python

Source: github.language · Jul 31, 2026

Categories

Inference & ServingModel Training

Compatibility

Sourced claims from the README excerpt - not unsourced marketing copy.

Python runtimePython

Source: README excerpt (regex_v1, Jul 31, 2026)

- Python 3.11
Source link

Tags

asrchineseconformerdeep-learningdeepspeech2paddlepaddlespeechspeech-recognition

README

PPASR流式与非流式语音识别项目

PPASR是一款基于PaddlePaddle实现的自动语音识别框架，PPASR中文名称PaddlePaddle中文语音识别（PaddlePaddle Automatic Speech Recognition），当前为V3版本，与V2版本不兼容，如果想使用V2版本，请在这个分支V2。PPASR致力于简单，实用的语音识别项目。可部署在服务器，Nvidia Jetson设备，未来还计划支持Android等移动设备。别忘了star

欢迎大家扫码入知识星球或者QQ群讨论，知识星球里面提供项目的模型文件和博主其他相关项目的模型文件，也包括其他一些资源。

知识星球 QQ群

本项目使用的环境：

  • Anaconda 3
  • Python 3.11
  • PaddlePaddle 2.6.1
  • Windows 11 or Ubuntu 22.04

在线试用

网页版： 在线试用地址

微信小程序
微信小程序

项目特点

  1. 支持多个语音识别模型，包含deepspeech2、conformer、squeezeformer、efficient_conformer等，每个模型都支持流式识别和非流式识别，在配置文件中streaming参数设置。
  2. 支持多种解码器，包含ctc_greedy_search、ctc_prefix_beam_search、attention_rescoring、ctc_beam_search等。
  3. 支持多种预处理方法，包含fbank、mfcc等。
  4. 支持多种数据增强方法，包含噪声增强、混响增强、语速增强、音量增强、重采样增强、位移增强、SpecAugmentor、SpecSubAugmentor等。
  5. 支持多种推理方法，包含短音频推理、长音频推理、流式推理、说话人分离推理等。
  6. 更多特点等待你发现。

与V2版本的区别

  1. 项目结构的优化，大幅度降低的使用难度。
  2. 更换预处理的库，改用kaldi_native_fbank，在提高数据预处理的速度，同时也支持多平台。
  3. 修改token的方法，使用sentencepiece制作token，这个框架极大的降低了多种语言的处理难度，同时还使中英文混合训练成为可能。

更新记录

  • 2025.03: 正式发布最终级的V3版本。

视频讲解

模型下载

  1. WenetSpeech (10000小时，普通话) 的预训练模型列表，错误率类型为字错率（CER）：
使用模型是否为流式预处理方式解码方式test_nettest_meetingaishell_test下载地址
ConformerTruefbankctc_greedy_search0.147580.195620.06925加入知识星球获取
ConformerTruefbankctc_prefix_beam_search0.146890.193230.06930加入知识星球获取
ConformerTruefbankattention_rescoring0.137860.189220.06028加入知识星球获取
ConformerTruefbankctc_beam_search0.206600.298350.05336加入知识星球获取
DeepSpeech2Truefbankctc_greedy_search加入知识星球获取
DeepSpeech2Truefbankctc_prefix_beam_search加入知识星球获取
DeepSpeech2Truefbankctc_beam_search加入知识星球获取
  1. AIShell (179小时，普通话) 的预训练模型列表，错误率类型为字错率（CER）：
使用模型是否为流式预处理方式解码方式自带的测试集下载地址
ConformerTruefbankctc_greedy_search0.06110加入知识星球获取
ConformerTruefbankctc_prefix_beam_search0.06114加入知识星球获取
ConformerTruefbankattention_rescoring0.05412加入知识星球获取
ConformerTruefbankctc_beam_search0.04468加入知识星球获取
DeepSpeech2Truefbankctc_greedy_search0.14134加入知识星球获取
DeepSpeech2Truefbankctc_prefix_beam_search0.14132加入知识星球获取
DeepSpeech2Truefbankctc_beam_search0.10598加入知识星球获取
  1. Librispeech (960小时，英语) 的预训练模型列表，错误率类型为词错率（WER）：
使用模型是否为流式预处理方式解码方式自带的测试集下载地址
ConformerTruefbankctc_greedy_search0.07562加入知识星球获取
ConformerTruefbankctc_prefix_beam_search0.07518加入知识星球获取
ConformerTruefbankattention_rescoring0.06669加入知识星球获取
ConformerTruefbankctc_beam_search

For agents

This page has a .md twin and JSON over the API.

/tools/yeyupiaoling-ppasr.mdGET /api/graphcanon/tools/yeyupiaoling-ppasrGET /api/graphcanon/graph?tool=yeyupiaoling-ppasr

Was this helpful?

Anonymous feedback helps us improve pages and translations.