sherpa-onnx-asr-models

ZhaoChaoqun/sherpa-onnx-asr-models

PublisherZhaoChaoqun

Download access requirementsModelScope source

🎙️ Sherpa-ONNX ASR Models Mirror GitHub 模型国内镜像 China Mirror for sherpa-onnx ASR Models [](https://opensource.org/licenses/Apache-2.0)

Speech & Audio

View all download instructions Mirror repository documentation and license ↗

Model details

模型介绍

中文整理

🎙️ Sherpa-ONNX ASR Models Mirror

GitHub 模型国内镜像 China Mirror for sherpa-onnx ASR Models**

[](https://opensource.org/licenses/Apache-2.0)

[](https://onnxruntime.ai/)

[](#模型列表)

[](#支持平台)

English 中文

⚡ 国内高速下载 🔒 完全离线运行 📱 全平台支持**

Chinese

📖 简介

本仓库是 k2-fsa/sherpa-onnx 官方 ASR 模型的**国内镜像**，解决 GitHub Releases 下载缓慢的问题。

sherpa-onnx** 是新一代 Kaldi 团队开发的跨平台语音识别推理框架，基于 ONNX Runtime，支持在各种设备上进行高效的**离线语音识别**，无需网络连接。

🌟 特色

🚀 **国内高速下载**：托管于 ModelScope，国内用户下载速度可达 10MB/s+

📦 **完整镜像**：包含 160+ 个官方模型，覆盖主流语音识别架构

🔄 **持续同步**：定期从 GitHub 同步最新模型

📝 **详细文档**：提供完整的使用示例和模型选择指南

📋 模型列表

🇨🇳 中文模型（推荐）

模型类型大小特点推荐场景

----- ------ ------ ------ ----------

sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17 SenseVoice ~230MB 多语言、情感识别通用场景

sherpa-onnx-sense-voice-zh-en-ja-ko-yue-int8-2024-07-17 SenseVoice ~60MB INT8 量化版移动端/边缘设备

sherpa-onnx-paraformer-zh-2024-03-09 Paraformer ~220MB 中文识别最佳纯中文场景

sherpa-onnx-paraformer-zh-small-2024-03-09 Paraformer ~30MB 轻量版资源受限设备

sherpa-onnx-streaming-zipformer-zh-2025-06-30 Zipformer ~80MB 流式识别实时转录

sherpa-onnx-streaming-paraformer-bilingual-zh-en Paraformer ~220MB 中英双语流式中英混合实时

🌍 多语言模型

模型支持语言大小特点

----- ---------- ------ ------

sherpa-onnx-whisper-large-v3 99+ 语言 ~3GB 最高准确率

sherpa-onnx-whisper-turbo 99+ 语言 ~1.5GB 速度优化版

sherpa-onnx-whisper-medium 99+ 语言 ~1.5GB 平衡之选

sherpa-onnx-whisper-small 99+ 语言 ~460MB 轻量版

sherpa-onnx-whisper-tiny 99+ 语言 ~75MB 最小版本

🇬🇧 英文模型

模型类型大小特点

----- ------ ------ ------

sherpa-onnx-zipformer-en-2023-06-26 Zipformer ~70MB 英文通用

sherpa-onnx-streaming-zipformer-en-2023-06-26 Zipformer ~20MB 流式英文

sherpa-onnx-conformer-en-2023-03-18 Conformer ~120MB 高准确率

sherpa-onnx-nemo-fast-conformer-ctc-en-24500 NeMo ~120MB NVIDIA NeMo

🌏 其他语言

语言推荐模型大小

----- ---------- ------

日语 sherpa-onnx-zipformer-ja-reazonspeech-2024-08-01 ~70MB

韩语 sherpa-onnx-zipformer-korean-2024-06-24 ~70MB

粤语 sherpa-onnx-zipformer-cantonese-2024-03-13 ~70MB

俄语 sherpa-onnx-zipformer-ru-2024-09-18 ~70MB

法语 sherpa-onnx-streaming-zipformer-fr-2023-04-14 ~20MB

德语 sherpa-onnx-streaming-zipformer-de-kroko-2025-08-06 ~20MB

西班牙语 sherpa-onnx-streaming-zipformer-es-kroko-2025-08-06 ~20MB

越南语 sherpa-onnx-zipformer-vi-2025-04-20 ~70MB

泰语 sherpa-onnx-zipformer-thai-2024-06-20 ~70MB

📱 移动端优化模型

模型大小特点

----- ------ ------

sherpa-onnx-streaming-zipformer-zh-14M-2023-02-23-mobile ~14MB 中文流式，超轻量

sherpa-onnx-streaming-zipformer-en-20M-2023-02-17-mobile ~20MB 英文流式，超轻量

sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20-mobile ~30MB 中英双语流式

sherpa-onnx-moonshine-tiny-en-int8 ~30MB 英文，极致压缩

🚀 快速开始

方式一：直接下载（推荐）

方式二：ModelScope SDK

方式三：Git Clone（下载全部）

💻 使用示例

Python

流式识别

Swift (iOS/macOS)

Android (Kotlin)

📊 模型架构说明

架构类型特点适用场景

----- ------ ------ ----------
*SenseVoice** 非流式多语言、情感识别、高准确率通用语音识别
*Paraformer** 非流式中文优化、速度快中文语音识别
*Whisper** 非流式 99+ 语言、最广泛支持多语言场景
*Zipformer** 流式/非流式轻量、低延迟实时转录
*Conformer** 流式/非流式高准确率对准确率要求高
*NeMo** 非流式 NVIDIA 优化 GPU 部署

🖥️ 支持平台

平台架构状态备注

----- ------ ------ ------
*macOS** arm64 (M1/M2/M3/M4) ✅ 原生支持
*macOS** x86 64 (Intel) ✅ 原生支持
*iOS** arm64 ✅ 支持 iPhone/iPad
*Android** arm64-v8a ✅ 支持大部分设备
*Android** armeabi-v7a ✅ 支持旧设备
*Linux** x86 64 ✅ 原生支持
*Linux** aarch64 ✅ 支持树莓派等
*Windows** x64 ✅ 原生支持
*Windows** x86 ✅ 32 位支持
*WebAssembly** wasm32 ✅ 浏览器运行

📈 性能参考

在 Apple M1 MacBook Pro 上的测试结果：

模型加载时间 RTF 内存占用

----- ---------- ----- ----------

SenseVoice INT8 ~1s 0.05 ~200MB

Paraformer ~2s 0.08 ~300MB

Whisper Small ~3s 0.15 ~500MB

Streaming Zipformer ~0.5s 实时 ~100MB

RTF (Real-Time Factor): < 1.0 表示比实时更快

🔗 相关资源

*sherpa-onnx 官方仓库**: https://github.com/k2-fsa/sherpa-onnx
*官方文档**: https://k2-fsa.github.io/sherpa/onnx/
*原始模型下载**: https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
*FunASR**: https://github.com/alibaba-damo-academy/FunASR
*SenseVoice**: https://github.com/FunAudioLLM/SenseVoice
*Whisper**: https://github.com/openai/whisper

📜 许可证

本仓库中的模型遵循各自的开源协议，大部分为 Apache License 2.0。

English

📖 Introduction

This repository is a **China mirror** of the official ASR models from k2-fsa/sherpa-onnx, providing faster download speeds for users in China.

sherpa-onnx** is a cross-platform speech recognition inference framework developed by the Next-gen Kaldi team, based on ONNX Runtime. It enables efficient **offline speech recognition** on various devices without network connectivity.

🌟 Features

🚀 **Fast Download in China**: Hosted on ModelScope, download speeds can reach 10MB/s+

📦 **Complete Mirror**: Contains 160+ official models covering mainstream ASR architectures

🔄 **Regular Sync**: Periodically synchronized from GitHub

📝 **Detailed Documentation**: Complete usage examples and model selection guide

📋 Model Categories

Recommended Models

Model Type Size Features

------ ------ ------ ----------

sherpa-onnx-sense-voice-zh-en-ja-ko-yue-int8-* SenseVoice ~60MB Multi-language, INT8 quantized

sherpa-onnx-whisper-turbo Whisper ~1.5GB 99+ languages, speed optimized

sherpa-onnx-streaming-zipformer-* Zipformer ~20MB Real-time streaming

By Language

*Chinese**: Paraformer, SenseVoice, Streaming Zipformer
*English**: Zipformer, Conformer, NeMo, Whisper
*Japanese**: Zipformer ReazonSpeech
*Korean**: Zipformer
*Multi-language**: Whisper, SenseVoice

🚀 Quick Start

💻 Usage Example

📜 License

Models in this repository follow their respective open-source licenses, mostly Apache License 2.0.

🎙️ 让语音识别更简单 Making Speech Recognition Easier**

如有问题或建议，欢迎提交 Issue For issues or suggestions, please submit an Issue

[](https://github.com/k2-fsa/sherpa-onnx)

[](https://modelscope.cn/models/zhaochaoqun/sherpa-onnx-asr-models)

优先采用原作者提供的中文模型卡片；仅有英文说明时由闲社AI翻译整理。重要参数、许可证和商用范围请以原文为准。

你将获得什么

完整仓库方式包含该固定版本中的权重、配置、分词器及说明文件；指定版本方式只下载所选文件或整组分片。多模态模型的投影文件、配置等依赖，请按作者说明配齐。

完整仓库可能包含多个权重格式，因此需要的磁盘空间可能大于单一格式的模型大小。下载文件不等于完成模型部署。

使用前须知

运行环境

运行模型需要的设备、工具与依赖，以原作者说明为准。

授权范围

是否允许商用、修改和分发，请查看模型许可证。

闲社提供资料与镜像下载方法，不代表已运行模型或完成安全审计。下载的代码应先检查，再执行。

Problem with the information or download?

Pinned version · Multiple download options

sherpa-onnx-asr-models model download

Download the full repository or choose a version or file.

Download tools · More than GGUF

Use the Hugging Face download tool to obtain ZhaoChaoqun/sherpa-onnx-asr-models. Download Safetensors, GGUF, ONNX, configuration and other repository files. Downloading does not mean the tool can run this model.

Pinned-version file information is being completed. Existing downloads remain available, or load the file manifest below.

Choose a mirror or official source, your operating system and download scope below, then copy the command. Install the download tool first if needed.

Pinned version: 0e7f9669b448c4b5561cf518785ac82c48901464. Downloads files only; does not start the model.

Choose a download source

Pinned ModelScope version: 0e7f9669b448c4b5561cf518785ac82c48901464. Version identifiers can differ across sites. The alternative source was provided after checking weights and runtime configuration; this is not proof of a successful model run.

A checked, pinned ModelScope source is available for this model. Use the ModelScope download tool.

Choose what to download

Need just one GGUF precision? Load the file manifest and choose a version. Direct file links do not require Python; the command below can download multiple shards together.

Quantization / file

This reads the pinned version's file manifest; no model weights are downloaded.

Choose your operating system

For command-line downloads, follow these steps to get the full repository or selected files. Direct file links do not need these steps.

Open a terminal window

Create and open a folder for the model. In the folder's address bar, enter powershell and press Enter.

If Python is not installed, get it from the official Python website. In the terminal, enter python --version; continue when a version number appears.

First time? Install the download tool

Copy this command, paste it into the terminal, press Enter and wait for installation to finish.

python -m pip install --upgrade huggingface_hub

Already installed? Continue to step 3.

Run the download command

Select the button below, paste into the same terminal and press Enter. The tool downloads all files included in the current command.

Copying does not start a download. Paste and run the command on your device.

Where are the downloaded files?

Open the folder you chose. Look for the models folder. When progress stops and the terminal returns to its prompt without errors, check that all files are present. 'Copied' does not mean 'downloaded'.

No response or an error?

python not found: Install Python and reopen the terminal. On Windows, follow the installer's instructions to enable command-line access.

hf not found: Confirm that step 2 succeeded, then reopen the terminal. If needed, add Python's Scripts folder to your user PATH. Do not repeatedly download scripts.

Network interrupted: Stay in the same folder and run the same command again. The tool checks existing files.

Permission or disk error: Choose a writable folder with enough free space. Do not disable system security protection.

View the full download command

modelscope download --model "ZhaoChaoqun/sherpa-onnx-asr-models" --revision "0e7f9669b448c4b5561cf518785ac82c48901464" --local_dir "./models/ZhaoChaoqun--sherpa-onnx-asr-models--ms-0e7f9669b448"

Files transfer directly from the source you choose, not through Xianshe servers.

The version is pinned. Repeating the same command checks existing files. Change the source manually if it is slow or unavailable.

Public models do not require submitting keys to Xianshe. Restricted models still require the author's authorization. On Linux / macOS, use python3 instead of python when appropriate.