xuning / sherpaonnx

转到一个项目
Toggle navigation Toggle navigation pinning
Fangjun Kuang
Committed by GitHub
537e163d 1 parent 1ef8a7a2

WebAssembly example for VAD + Non-streaming ASR (#1284)

内嵌 并排对比
正在显示 29 个修改的文件 包含 1281 行增加 和 70 行删除
... ... @@ -25,8 +25,12 @@ jobs:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install emsdk
uses: mymindstorm/setup-emsdk@v14
with:
version: 3.1.51
actions-cache-folder: 'emsdk-cache'
- name: View emsdk version
shell: bash
... ...
... ... @@ -27,6 +27,9 @@ jobs:
fetch-depth: 0
- name: Install emsdk
uses: mymindstorm/setup-emsdk@v14
with:
version: 3.1.51
actions-cache-folder: 'emsdk-cache'
- name: View emsdk version
shell: bash
... ...
... ... @@ -25,8 +25,12 @@ jobs:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install emsdk
uses: mymindstorm/setup-emsdk@v14
with:
version: 3.1.51
actions-cache-folder: 'emsdk-cache'
- name: View emsdk version
shell: bash
... ...
... ... @@ -25,6 +25,7 @@ jobs:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install emsdk
uses: mymindstorm/setup-emsdk@v14
with:
... ...
name: wasm-simd-hf-space-vad-asr
on:
push:
branches:
- wasm
tags:
- 'v[0-9]+.[0-9]+.[0-9]+*'
workflow_dispatch:
concurrency:
group: wasm-simd-hf-space-vad-asr${{ github.ref }}
cancel-in-progress: true
jobs:
wasm-simd-hf-space-vad-asr:
name: ${{ matrix.index }}/${{ matrix.total }}
runs-on: ${{ matrix.os }}
strategy:
fail-fast: false
matrix:
os: [ubuntu-latest]
total: ["8"]
index: ["0", "1", "2", "3", "4", "5", "6", "7"]
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install Python dependencies
shell: bash
run: |
python3 -m pip install --upgrade pip jinja2
- name: Install emsdk
uses: mymindstorm/setup-emsdk@v14
with:
version: 3.1.51
actions-cache-folder: 'emsdk-cache'
- name: View emsdk version
shell: bash
run: |
emcc -v
echo "--------------------"
emcc --check
- name: Generate build script
shell: bash
run: |
cd scripts/wasm
total=${{ matrix.total }}
index=${{ matrix.index }}
./generate-vad-asr.py --total $total --index $index
chmod +x run-vad-asr.sh
mv -v ./run-vad-asr.sh ../..
- name: Show build scripts
shell: bash
run: |
cat ./run-vad-asr.sh
- uses: actions/upload-artifact@v4
with:
name: run-vad-asr-${{ matrix.index }}
path: ./run-vad-asr.sh
- name: Build sherpa-onnx for WebAssembly
shell: bash
env:
MS_TOKEN: ${{ secrets.MODEL_SCOPE_GIT_TOKEN }}
HF_TOKEN: ${{ secrets.HF_TOKEN }}
run: |
./run-vad-asr.sh
- name: Release jar
if: (github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa') && github.event_name == 'push' && contains(github.ref, 'refs/tags/')
uses: svenstaro/upload-release-action@v2
with:
file_glob: true
overwrite: true
file: ./*.tar.bz2
- name: Upload wasm files
uses: actions/upload-artifact@v4
with:
name: sherpa-onnx-wasm-simd-vad-asr-${{ matrix.index }}
path: ./sherpa-onnx-wasm-simd-*.tar.bz2
... ...
... ... @@ -25,8 +25,12 @@ jobs:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install emsdk
uses: mymindstorm/setup-emsdk@v14
with:
version: 3.1.51
actions-cache-folder: 'emsdk-cache'
- name: View emsdk version
shell: bash
... ...
... ... @@ -25,8 +25,12 @@ jobs:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install emsdk
uses: mymindstorm/setup-emsdk@v14
with:
version: 3.1.51
actions-cache-folder: 'emsdk-cache'
- name: View emsdk version
shell: bash
... ...
... ... @@ -25,8 +25,12 @@ jobs:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install emsdk
uses: mymindstorm/setup-emsdk@v14
with:
version: 3.1.51
actions-cache-folder: 'emsdk-cache'
- name: View emsdk version
shell: bash
... ...
... ... @@ -36,6 +36,7 @@ option(SHERPA_ONNX_ENABLE_WASM_TTS "Whether to enable WASM for TTS" OFF)
option(SHERPA_ONNX_ENABLE_WASM_ASR "Whether to enable WASM for ASR" OFF)
option(SHERPA_ONNX_ENABLE_WASM_KWS "Whether to enable WASM for KWS" OFF)
option(SHERPA_ONNX_ENABLE_WASM_VAD "Whether to enable WASM for VAD" OFF)
option(SHERPA_ONNX_ENABLE_WASM_VAD_ASR "Whether to enable WASM for VAD+ASR" OFF)
option(SHERPA_ONNX_ENABLE_WASM_NODEJS "Whether to enable WASM for NodeJS" OFF)
option(SHERPA_ONNX_ENABLE_BINARY "Whether to build binaries" ON)
option(SHERPA_ONNX_ENABLE_TTS "Whether to build TTS related code" ON)
... ... @@ -137,6 +138,7 @@ message(STATUS "SHERPA_ONNX_ENABLE_WASM_TTS ${SHERPA_ONNX_ENABLE_WASM_TTS}")
message(STATUS "SHERPA_ONNX_ENABLE_WASM_ASR ${SHERPA_ONNX_ENABLE_WASM_ASR}")
message(STATUS "SHERPA_ONNX_ENABLE_WASM_KWS ${SHERPA_ONNX_ENABLE_WASM_KWS}")
message(STATUS "SHERPA_ONNX_ENABLE_WASM_VAD ${SHERPA_ONNX_ENABLE_WASM_VAD}")
message(STATUS "SHERPA_ONNX_ENABLE_WASM_VAD_ASR ${SHERPA_ONNX_ENABLE_WASM_VAD_ASR}")
message(STATUS "SHERPA_ONNX_ENABLE_WASM_NODEJS ${SHERPA_ONNX_ENABLE_WASM_NODEJS}")
message(STATUS "SHERPA_ONNX_ENABLE_BINARY ${SHERPA_ONNX_ENABLE_BINARY}")
message(STATUS "SHERPA_ONNX_ENABLE_TTS ${SHERPA_ONNX_ENABLE_TTS}")
... ... @@ -211,11 +213,22 @@ if(SHERPA_ONNX_ENABLE_WASM)
endif()
if(SHERPA_ONNX_ENABLE_WASM_KWS)
if(NOT SHERPA_ONNX_ENABLE_WASM)
message(FATAL_ERROR "Please set SHERPA_ONNX_ENABLE_WASM to ON if you enable WASM for KWS")
endif()
add_definitions(-DSHERPA_ONNX_ENABLE_WASM_KWS=1)
endif()
if(SHERPA_ONNX_ENABLE_WASM_VAD)
add_definitions(-DSHERPA_ONNX_ENABLE_WASM_VAD=1)
if(NOT SHERPA_ONNX_ENABLE_WASM)
message(FATAL_ERROR "Please set SHERPA_ONNX_ENABLE_WASM to ON if you enable WASM for VAD")
endif()
endif()
if(SHERPA_ONNX_ENABLE_WASM_VAD_ASR)
if(NOT SHERPA_ONNX_ENABLE_WASM)
message(FATAL_ERROR "Please set SHERPA_ONNX_ENABLE_WASM to ON if you enable WASM for VAD+ASR")
endif()
endif()
if(NOT CMAKE_CXX_STANDARD)
... ...
... ... @@ -14,13 +14,13 @@
### Supported platforms
|Architecture| Android | iOS | Windows | macOS | linux |
|------------|------------------|---------------|------------|-------|-------|
| x64 | ✔️ | | ✔️ | ✔️ | ✔️ |
| x86 | ✔️ | | ✔️ | | |
| arm64 | ✔️ | ✔️ | ✔️ | ✔️ | ✔️ |
| arm32 | ✔️ | | | | ✔️ |
| riscv64 | | | | | ✔️ |
|Architecture| Android | iOS | Windows | macOS | linux |
|------------|---------|---------|------------|-------|-------|
| x64 | ✔️ | | ✔️ | ✔️ | ✔️ |
| x86 | ✔️ | | ✔️ | | |
| arm64 | ✔️ | ✔️ | ✔️ | ✔️ | ✔️ |
| arm32 | ✔️ | | | | ✔️ |
| riscv64 | | | | | ✔️ |
### Supported programming languages
... ... @@ -37,7 +37,7 @@
|-------|----------|----------|------------|
| ✔️ | ✔️ | ✔️ | ✔️ |
For Rust support, please see https://github.com/thewh1teagle/sherpa-rs
For Rust support, please see [sherpa-rs][sherpa-rs]
It also supports WebAssembly.
... ... @@ -51,7 +51,7 @@ This repository supports running the following functions **locally**
- Speaker verification
- Spoken language identification
- Audio tagging
- VAD (e.g., [silero-vad](https://github.com/snakers4/silero-vad))
- VAD (e.g., [silero-vad][silero-vad])
- Keyword spotting
on the following platforms and operating systems:
... ... @@ -62,11 +62,12 @@ on the following platforms and operating systems:
- iOS
- NodeJS
- WebAssembly
- [Raspberry Pi](https://www.raspberrypi.com/)
- [RV1126](https://www.rock-chips.com/uploads/pdf/2022.8.26/191/RV1126%20Brief%20Datasheet.pdf)
- [LicheePi4A](https://sipeed.com/licheepi4a)
- [VisionFive 2](https://www.starfivetech.com/en/site/boards)
- [旭日X3派](https://developer.horizon.ai/api/v1/fileData/documents_pi/index.html)
- [Raspberry Pi][Raspberry Pi]
- [RV1126][RV1126]
- [LicheePi4A][LicheePi4A]
- [VisionFive 2][VisionFive 2]
- [旭日X3派][旭日X3派]
- [爱芯派][爱芯派]
- etc
with the following APIs
... ... @@ -81,59 +82,68 @@ with the following APIs
You can visit the following Huggingface spaces to try `sherpa-onnx` without
installing anything. All you need is a browser.
| Description | URL |
|---|---|
| Speech recognition | [Click me](https://huggingface.co/spaces/k2-fsa/automatic-speech-recognition)|
| Speech recognition with [Whisper](https://github.com/openai/whisper)| [Click me](https://huggingface.co/spaces/k2-fsa/automatic-speech-recognition-with-whisper)|
| Speech synthesis | [Click me](https://huggingface.co/spaces/k2-fsa/text-to-speech)|
| Generate subtitles| [Click me](https://huggingface.co/spaces/k2-fsa/generate-subtitles-for-videos)|
|Audio tagging| [Click me](https://huggingface.co/spaces/k2-fsa/audio-tagging)|
|Spoken language identification with [Whisper](https://github.com/openai/whisper)|[Click me](https://huggingface.co/spaces/k2-fsa/spoken-language-identification)|
| Description | URL |
|-------------------------------------------------------|------------------------------------|
| Speech recognition | [Click me][hf-space-asr] |
| Speech recognition with [Whisper][Whisper] | [Click me][hf-space-asr-whisper] |
| Speech synthesis | [Click me][hf-space-tts] |
| Generate subtitles | [Click me][hf-space-subtitle] |
| Audio tagging | [Click me][hf-space-audio-tagging] |
| Spoken language identification with [Whisper][Whisper]| [Click me][hf-space-slid-whisper] |
We also have spaces built using WebAssembly. The are listed below:
| Description | URL| Chinese users|
|---|---|---|
|Voice activity detection with [silero-vad](https://github.com/snakers4/silero-vad)| [Click me](https://huggingface.co/spaces/k2-fsa/web-assembly-vad-sherpa-onnx)|[地址](https://modelscope.cn/studios/csukuangfj/web-assembly-vad-sherpa-onnx)|
|Real-time speech recognition (Chinese + English) with Zipformer | [Click me](https://huggingface.co/spaces/k2-fsa/web-assembly-asr-sherpa-onnx-zh-en)|[地址](https://modelscope.cn/studios/k2-fsa/web-assembly-asr-sherpa-onnx-zh-en)|
|Real-time speech recognition (Chinese + English) with Paraformer|[Click me](https://huggingface.co/spaces/k2-fsa/web-assembly-asr-sherpa-onnx-zh-en-paraformer)| [地址](https://modelscope.cn/studios/k2-fsa/web-assembly-asr-sherpa-onnx-zh-en-paraformer)|
|Real-time speech recognition (Chinese + English + Cantonese) with Paraformer|[Click me](https://huggingface.co/spaces/k2-fsa/web-assembly-asr-sherpa-onnx-zh-cantonese-en-paraformer)| [地址](https://modelscope.cn/studios/k2-fsa/web-assembly-asr-sherpa-onnx-zh-cantonese-en-paraformer)|
|Real-time speech recognition (English) |[Click me](https://huggingface.co/spaces/k2-fsa/web-assembly-asr-sherpa-onnx-en)|[地址](https://modelscope.cn/studios/k2-fsa/web-assembly-asr-sherpa-onnx-en)|
|Speech synthesis (English) |[Click me](https://huggingface.co/spaces/k2-fsa/web-assembly-tts-sherpa-onnx-en)| [地址](https://modelscope.cn/studios/k2-fsa/web-assembly-tts-sherpa-onnx-en)|
|Speech synthesis (German)|[Click me](https://huggingface.co/spaces/k2-fsa/web-assembly-tts-sherpa-onnx-de)| [地址](https://modelscope.cn/studios/k2-fsa/web-assembly-tts-sherpa-onnx-de)|
| Description | Huggingface space| ModelScope space|
|------------------------------------------------------------------------------------------|------------------|-----------------|
|Voice activity detection with [silero-vad][silero-vad] | [Click me][wasm-hf-vad]|[地址][wasm-ms-vad]|
|Real-time speech recognition (Chinese + English) with Zipformer | [Click me][wasm-hf-streaming-asr-zh-en-zipformer]|[地址][wasm-hf-streaming-asr-zh-en-zipformer]|
|Real-time speech recognition (Chinese + English) with Paraformer |[Click me][wasm-hf-streaming-asr-zh-en-paraformer]| [地址][wasm-ms-streaming-asr-zh-en-paraformer]|
|Real-time speech recognition (Chinese + English + Cantonese) with [Paraformer-large][Paraformer-large]|[Click me][wasm-hf-streaming-asr-zh-en-yue-paraformer]| [地址][wasm-ms-streaming-asr-zh-en-yue-paraformer]|
|Real-time speech recognition (English) |[Click me][wasm-hf-streaming-asr-en-zipformer] |[地址][wasm-ms-streaming-asr-en-zipformer]|
|VAD + speech recognition (Chinese + English + Korean + Japanese + Cantonese) with [SenseVoice][SenseVoice]|[Click me][wasm-hf-vad-asr-zh-en-ko-ja-yue-sense-voice]| [地址][wasm-ms-vad-asr-zh-en-ko-ja-yue-sense-voice]|
|VAD + speech recognition (English) with [Whisper][Whisper] tiny.en|[Click me][wasm-hf-vad-asr-en-whisper-tiny-en]| [地址][wasm-ms-vad-asr-en-whisper-tiny-en]|
|VAD + speech recognition (English) with Zipformer trained with [GigaSpeech][GigaSpeech] |[Click me][wasm-hf-vad-asr-en-zipformer-gigaspeech]| [地址][wasm-ms-vad-asr-en-zipformer-gigaspeech]|
|VAD + speech recognition (Chinese) with Zipformer trained with [WenetSpeech][WenetSpeech] |[Click me][wasm-hf-vad-asr-zh-zipformer-wenetspeech]| [地址][wasm-ms-vad-asr-zh-zipformer-wenetspeech]|
|VAD + speech recognition (Japanese) with Zipformer trained with [ReazonSpeech][ReazonSpeech]|[Click me][wasm-hf-vad-asr-ja-zipformer-reazonspeech]| [地址][wasm-ms-vad-asr-ja-zipformer-reazonspeech]|
|VAD + speech recognition (Thai) with Zipformer trained with [GigaSpeech2][GigaSpeech2] |[Click me][wasm-hf-vad-asr-th-zipformer-gigaspeech2]| [地址][wasm-ms-vad-asr-th-zipformer-gigaspeech2]|
|VAD + speech recognition (Chinese 多种方言) with a [TeleSpeech-ASR][TeleSpeech-ASR] CTC model|[Click me][wasm-hf-vad-asr-zh-telespeech]| [地址][wasm-ms-vad-asr-zh-telespeech]|
|VAD + speech recognition (English + Chinese, 及多种中文方言) with Paraformer-large |[Click me][wasm-hf-vad-asr-zh-en-paraformer-large]| [地址][wasm-ms-vad-asr-zh-en-paraformer-large]|
|VAD + speech recognition (English + Chinese, 及多种中文方言) with Paraformer-small |[Click me][wasm-hf-vad-asr-zh-en-paraformer-small]| [地址][wasm-ms-vad-asr-zh-en-paraformer-small]|
|Speech synthesis (English) |[Click me][wasm-hf-tts-piper-en]| [地址][wasm-ms-tts-piper-en]|
|Speech synthesis (German) |[Click me][wasm-hf-tts-piper-de]| [地址][wasm-ms-tts-piper-de]|
### Links for pre-built Android APKs
| Description | URL | 中国用户 |
|--------------------------------|-----------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------|
| Streaming speech recognition | [Address](https://k2-fsa.github.io/sherpa/onnx/android/apk.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/android/apk-cn.html) |
| Text-to-speech | [Address](https://k2-fsa.github.io/sherpa/onnx/tts/apk-engine.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/tts/apk-engine-cn.html) |
|Voice activity detection (VAD) | [Address](https://k2-fsa.github.io/sherpa/onnx/vad/apk.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/vad/apk-cn.html)|
|VAD + non-streaming speech recognition| [Address](https://k2-fsa.github.io/sherpa/onnx/vad/apk-asr.html)| [点此](https://k2-fsa.github.io/sherpa/onnx/vad/apk-asr-cn.html)|
|Two-pass speech recognition| [Address](https://k2-fsa.github.io/sherpa/onnx/android/apk-2pass.html)| [点此](https://k2-fsa.github.io/sherpa/onnx/android/apk-2pass-cn.html)|
| Audio tagging | [Address](https://k2-fsa.github.io/sherpa/onnx/audio-tagging/apk.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/audio-tagging/apk-cn.html) |
| Audio tagging (WearOS) | [Address](https://k2-fsa.github.io/sherpa/onnx/audio-tagging/apk-wearos.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/audio-tagging/apk-wearos-cn.html) |
| Speaker identification | [Address](https://k2-fsa.github.io/sherpa/onnx/speaker-identification/apk.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/speaker-identification/apk-cn.html) |
| Spoken language identification | [Address](https://k2-fsa.github.io/sherpa/onnx/spoken-language-identification/apk.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/spoken-language-identification/apk-cn.html) |
|Keyword spotting| [Address](https://k2-fsa.github.io/sherpa/onnx/kws/apk.html)| [点此](https://k2-fsa.github.io/sherpa/onnx/kws/apk-cn.html)|
| Description | URL | 中国用户 |
|----------------------------------------|------------------------------|-----------------------------|
| Streaming speech recognition | [Address][apk-streaming-asr] | [点此][apk-streaming-asr-cn]|
| Text-to-speech | [Address][apk-tts] | [点此][apk-tts-cn] |
| Voice activity detection (VAD) | [Address][apk-vad] | [点此][apk-vad-cn] |
| VAD + non-streaming speech recognition | [Address][apk-vad-asr] | [点此][apk-vad-asr-cn] |
| Two-pass speech recognition | [Address][apk-2pass] | [点此][apk-2pass-cn] |
| Audio tagging | [Address][apk-at] | [点此][apk-at-cn] |
| Audio tagging (WearOS) | [Address][apk-at-wearos] | [点此][apk-at-wearos-cn] |
| Speaker identification | [Address][apk-sid] | [点此][apk-sid-cn] |
| Spoken language identification | [Address][apk-slid] | [点此][apk-slid-cn] |
| Keyword spotting | [Address][apk-kws] | [点此][apk-kws-cn] |
### Links for pre-built Flutter APPs
#### Real-time speech recognition
| Description | URL | 中国用户 |
|--------------------------------|---------------------------------------------------------------------|---------------------------------------------------------------------|
| Streaming speech recognition | [Address](https://k2-fsa.github.io/sherpa/onnx/flutter/asr/app.html)| [点此](https://k2-fsa.github.io/sherpa/onnx/flutter/asr/app-cn.html)|
| Description | URL | 中国用户 |
|--------------------------------|-------------------------------------|-------------------------------------|
| Streaming speech recognition | [Address][apk-flutter-streaming-asr]| [点此][apk-flutter-streaming-asr-cn]|
#### Text-to-speech
| Description | URL | 中国用户 |
|--------------------------------|--------------------------------------------------------------|-----------------------------------------------------------------------------|
| Android (arm64-v8a, armeabi-v7a, x86_64) | [Address](https://k2-fsa.github.io/sherpa/onnx/flutter/tts-android.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/flutter/tts-android-cn.html)|
| Linux (x64) | [Address](https://k2-fsa.github.io/sherpa/onnx/flutter/tts-linux.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/flutter/tts-linux-cn.html) |
| macOS (x64) | [Address](https://k2-fsa.github.io/sherpa/onnx/flutter/tts-macos-x64.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/flutter/tts-macos-x64-cn.html) |
| macOS (arm64) | [Address](https://k2-fsa.github.io/sherpa/onnx/flutter/tts-macos-arm64.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/flutter/tts-macos-arm64-cn.html)|
| Windows (x64) | [Address](https://k2-fsa.github.io/sherpa/onnx/flutter/tts-win.html) | [点此](https://k2-fsa.github.io/sherpa/onnx/flutter/tts-win-cn.html) |
| Description | URL | 中国用户 |
|------------------------------------------|------------------------------------|------------------------------------|
| Android (arm64-v8a, armeabi-v7a, x86_64) | [Address][flutter-tts-android] | [点此][flutter-tts-android-cn] |
| Linux (x64) | [Address][flutter-tts-linux] | [点此][flutter-tts-linux-cn] |
| macOS (x64) | [Address][flutter-tts-macos-x64] | [点此][flutter-tts-macos-arm64-cn] |
| macOS (arm64) | [Address][flutter-tts-macos-arm64] | [点此][flutter-tts-macos-x64-cn] |
| Windows (x64) | [Address][flutter-tts-win-x64] | [点此][flutter-tts-win-x64-cn] |
> Note: You need to build from source for iOS.
... ... @@ -141,23 +151,23 @@ We also have spaces built using WebAssembly. The are listed below:
#### Generating subtitles
| Description | URL | 中国用户 |
|--------------------------------|---------------------------------------------------------------------|---------------------------------------------------------------------|
| Generate subtitles (生成字幕) | [Address](https://k2-fsa.github.io/sherpa/onnx/lazarus/download-generated-subtitles.html)| [点此](https://k2-fsa.github.io/sherpa/onnx/lazarus/download-generated-subtitles-cn.html)|
| Description | URL | 中国用户 |
|--------------------------------|----------------------------|----------------------------|
| Generate subtitles (生成字幕) | [Address][lazarus-subtitle]| [点此][lazarus-subtitle-cn]|
### Links for pre-trained models
| Description | URL |
|--------------------------------|--------------------------------------------------------------------------------------------------------------------------------|
| Speech recognition (speech to text, ASR) | [Address](https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models) |
| Text-to-speech (TTS) | [Address](https://github.com/k2-fsa/sherpa-onnx/releases/tag/tts-models) |
| VAD | [Address](https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx)|
| Keyword spotting |[Address](https://github.com/k2-fsa/sherpa-onnx/releases/tag/kws-models)|
| Audio tagging | [Address](https://github.com/k2-fsa/sherpa-onnx/releases/tag/audio-tagging-models)|
| Speaker identification (Speaker ID) | [Address](https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-recongition-models)|
| Spoken language identification (Language ID) | See multi-lingual [Whisper](https://github.com/openai/whisper) ASR models from [Speech recognition](https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models) |
| Punctuation| [Address](https://github.com/k2-fsa/sherpa-onnx/releases/tag/punctuation-models)|
| Description | URL |
|---------------------------------------------|---------------------------------------------------------------------------------------|
| Speech recognition (speech to text, ASR) | [Address][asr-models] |
| Text-to-speech (TTS) | [Address][tts-models] |
| VAD | [Address][vad-models] |
| Keyword spotting | [Address][kws-models] |
| Audio tagging | [Address][at-models] |
| Speaker identification (Speaker ID) | [Address][sid-models] |
| Spoken language identification (Language ID)| See multi-lingual [Whisper][Whisper] ASR models from [Speech recognition][asr-models]|
| Punctuation | [Address][punct-models] |
### Useful links
... ... @@ -169,3 +179,100 @@ We also have spaces built using WebAssembly. The are listed below:
Please see
https://k2-fsa.github.io/sherpa/social-groups.html
for 新一代 Kaldi **微信交流群** and **QQ 交流群**.
[sherpa-rs]: https://github.com/thewh1teagle/sherpa-rs
[silero-vad]: https://github.com/snakers4/silero-vad
[Raspberry Pi]: https://www.raspberrypi.com/
[RV1126]: https://www.rock-chips.com/uploads/pdf/2022.8.26/191/RV1126%20Brief%20Datasheet.pdf
[LicheePi4A]: https://sipeed.com/licheepi4a
[VisionFive 2]: https://www.starfivetech.com/en/site/boards
[旭日X3派]: https://developer.horizon.ai/api/v1/fileData/documents_pi/index.html
[爱芯派]: https://wiki.sipeed.com/hardware/zh/maixIII/ax-pi/axpi.html
[hf-space-asr]: https://huggingface.co/spaces/k2-fsa/automatic-speech-recognition
[Whisper]: https://github.com/openai/whisper
[hf-space-asr-whisper]: https://huggingface.co/spaces/k2-fsa/automatic-speech-recognition-with-whisper
[hf-space-tts]: https://huggingface.co/spaces/k2-fsa/text-to-speech
[hf-space-subtitle]: https://huggingface.co/spaces/k2-fsa/generate-subtitles-for-videos
[hf-space-audio-tagging]: https://huggingface.co/spaces/k2-fsa/audio-tagging
[hf-space-slid-whisper]: https://huggingface.co/spaces/k2-fsa/spoken-language-identification
[wasm-hf-vad]: https://huggingface.co/spaces/k2-fsa/web-assembly-vad-sherpa-onnx
[wasm-ms-vad]: https://modelscope.cn/studios/csukuangfj/web-assembly-vad-sherpa-onnx
[wasm-hf-streaming-asr-zh-en-zipformer]: https://huggingface.co/spaces/k2-fsa/web-assembly-asr-sherpa-onnx-zh-en
[wasm-ms-streaming-asr-zh-en-zipformer]: https://modelscope.cn/studios/k2-fsa/web-assembly-asr-sherpa-onnx-zh-en
[wasm-hf-streaming-asr-zh-en-paraformer]: https://huggingface.co/spaces/k2-fsa/web-assembly-asr-sherpa-onnx-zh-en-paraformer
[wasm-ms-streaming-asr-zh-en-paraformer]: https://modelscope.cn/studios/k2-fsa/web-assembly-asr-sherpa-onnx-zh-en-paraformer
[Paraformer-large]: https://www.modelscope.cn/models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/summary
[wasm-hf-streaming-asr-zh-en-yue-paraformer]: https://huggingface.co/spaces/k2-fsa/web-assembly-asr-sherpa-onnx-zh-cantonese-en-paraformer
[wasm-ms-streaming-asr-zh-en-yue-paraformer]: https://modelscope.cn/studios/k2-fsa/web-assembly-asr-sherpa-onnx-zh-cantonese-en-paraformer
[wasm-hf-streaming-asr-en-zipformer]: https://huggingface.co/spaces/k2-fsa/web-assembly-asr-sherpa-onnx-en
[wasm-ms-streaming-asr-en-zipformer]: https://modelscope.cn/studios/k2-fsa/web-assembly-asr-sherpa-onnx-en
[SenseVoice]: https://github.com/FunAudioLLM/SenseVoice
[wasm-hf-vad-asr-zh-en-ko-ja-yue-sense-voice]: https://huggingface.co/spaces/k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-en-ja-ko-cantonese-sense-voice
[wasm-ms-vad-asr-zh-en-ko-ja-yue-sense-voice]: https://www.modelscope.cn/studios/csukuangfj/web-assembly-vad-asr-sherpa-onnx-zh-en-jp-ko-cantonese-sense-voice
[wasm-hf-vad-asr-en-whisper-tiny-en]: https://huggingface.co/spaces/k2-fsa/web-assembly-vad-asr-sherpa-onnx-en-whisper-tiny
[wasm-ms-vad-asr-en-whisper-tiny-en]: https://www.modelscope.cn/studios/csukuangfj/web-assembly-vad-asr-sherpa-onnx-en-whisper-tiny
[wasm-hf-vad-asr-en-zipformer-gigaspeech]: https://huggingface.co/spaces/k2-fsa/web-assembly-vad-asr-sherpa-onnx-en-zipformer-gigaspeech
[wasm-ms-vad-asr-en-zipformer-gigaspeech]: https://www.modelscope.cn/studios/k2-fsa/web-assembly-vad-asr-sherpa-onnx-en-zipformer-gigaspeech
[wasm-hf-vad-asr-zh-zipformer-wenetspeech]: https://huggingface.co/spaces/k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-zipformer-wenetspeech
[wasm-ms-vad-asr-zh-zipformer-wenetspeech]: https://www.modelscope.cn/studios/k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-zipformer-wenetspeech
[ReazonSpeech]: https://research.reazon.jp/_static/reazonspeech_nlp2023.pdf
[wasm-hf-vad-asr-ja-zipformer-reazonspeech]: https://huggingface.co/spaces/k2-fsa/web-assembly-vad-asr-sherpa-onnx-ja-zipformer
[wasm-ms-vad-asr-ja-zipformer-reazonspeech]: https://www.modelscope.cn/studios/csukuangfj/web-assembly-vad-asr-sherpa-onnx-ja-zipformer
[GigaSpeech2]: https://github.com/SpeechColab/GigaSpeech2
[wasm-hf-vad-asr-th-zipformer-gigaspeech2]: https://huggingface.co/spaces/k2-fsa/web-assembly-vad-asr-sherpa-onnx-th-zipformer
[wasm-ms-vad-asr-th-zipformer-gigaspeech2]: https://www.modelscope.cn/studios/csukuangfj/web-assembly-vad-asr-sherpa-onnx-th-zipformer
[TeleSpeech-ASR]: https://github.com/Tele-AI/TeleSpeech-ASR
[wasm-hf-vad-asr-zh-telespeech]: https://huggingface.co/spaces/k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-telespeech
[wasm-ms-vad-asr-zh-telespeech]: https://www.modelscope.cn/studios/k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-telespeech
[wasm-hf-vad-asr-zh-en-paraformer-large]: https://huggingface.co/spaces/k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-en-paraformer
[wasm-ms-vad-asr-zh-en-paraformer-large]: https://www.modelscope.cn/studios/k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-en-paraformer
[wasm-hf-vad-asr-zh-en-paraformer-small]: https://huggingface.co/spaces/k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-en-paraformer-small
[wasm-ms-vad-asr-zh-en-paraformer-small]: https://www.modelscope.cn/studios/k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-en-paraformer-small
[wasm-hf-tts-piper-en]: https://huggingface.co/spaces/k2-fsa/web-assembly-tts-sherpa-onnx-en
[wasm-ms-tts-piper-en]: https://modelscope.cn/studios/k2-fsa/web-assembly-tts-sherpa-onnx-en
[wasm-hf-tts-piper-de]: https://huggingface.co/spaces/k2-fsa/web-assembly-tts-sherpa-onnx-de
[wasm-ms-tts-piper-de]: https://modelscope.cn/studios/k2-fsa/web-assembly-tts-sherpa-onnx-de
[apk-streaming-asr]: https://k2-fsa.github.io/sherpa/onnx/android/apk.html
[apk-streaming-asr-cn]: https://k2-fsa.github.io/sherpa/onnx/android/apk-cn.html
[apk-tts]: https://k2-fsa.github.io/sherpa/onnx/tts/apk-engine.html
[apk-tts-cn]: https://k2-fsa.github.io/sherpa/onnx/tts/apk-engine-cn.html
[apk-vad]: https://k2-fsa.github.io/sherpa/onnx/vad/apk.html
[apk-vad-cn]: https://k2-fsa.github.io/sherpa/onnx/vad/apk-cn.html
[apk-vad-asr]: https://k2-fsa.github.io/sherpa/onnx/vad/apk-asr.html
[apk-vad-asr-cn]: https://k2-fsa.github.io/sherpa/onnx/vad/apk-asr-cn.html
[apk-2pass]: https://k2-fsa.github.io/sherpa/onnx/android/apk-2pass.html
[apk-2pass-cn]: https://k2-fsa.github.io/sherpa/onnx/android/apk-2pass-cn.html
[apk-at]: https://k2-fsa.github.io/sherpa/onnx/audio-tagging/apk.html
[apk-at-cn]: https://k2-fsa.github.io/sherpa/onnx/audio-tagging/apk-cn.html
[apk-at-wearos]: https://k2-fsa.github.io/sherpa/onnx/audio-tagging/apk-wearos.html
[apk-at-wearos-cn]: https://k2-fsa.github.io/sherpa/onnx/audio-tagging/apk-wearos-cn.html
[apk-sid]: https://k2-fsa.github.io/sherpa/onnx/speaker-identification/apk.html
[apk-sid-cn]: https://k2-fsa.github.io/sherpa/onnx/speaker-identification/apk-cn.html
[apk-slid]: https://k2-fsa.github.io/sherpa/onnx/spoken-language-identification/apk.html
[apk-slid-cn]: https://k2-fsa.github.io/sherpa/onnx/spoken-language-identification/apk-cn.html
[apk-kws]: https://k2-fsa.github.io/sherpa/onnx/kws/apk.html
[apk-kws-cn]: https://k2-fsa.github.io/sherpa/onnx/kws/apk-cn.html
[apk-flutter-streaming-asr]: https://k2-fsa.github.io/sherpa/onnx/flutter/asr/app.html
[apk-flutter-streaming-asr-cn]: https://k2-fsa.github.io/sherpa/onnx/flutter/asr/app-cn.html
[flutter-tts-android]: https://k2-fsa.github.io/sherpa/onnx/flutter/tts-android.html
[flutter-tts-android-cn]: https://k2-fsa.github.io/sherpa/onnx/flutter/tts-android-cn.html
[flutter-tts-linux]: https://k2-fsa.github.io/sherpa/onnx/flutter/tts-linux.html
[flutter-tts-linux-cn]: https://k2-fsa.github.io/sherpa/onnx/flutter/tts-linux-cn.html
[flutter-tts-macos-x64]: https://k2-fsa.github.io/sherpa/onnx/flutter/tts-macos-x64.html
[flutter-tts-macos-arm64-cn]: https://k2-fsa.github.io/sherpa/onnx/flutter/tts-macos-x64-cn.html
[flutter-tts-macos-arm64]: https://k2-fsa.github.io/sherpa/onnx/flutter/tts-macos-arm64.html
[flutter-tts-macos-x64-cn]: https://k2-fsa.github.io/sherpa/onnx/flutter/tts-macos-arm64-cn.html
[flutter-tts-win-x64]: https://k2-fsa.github.io/sherpa/onnx/flutter/tts-win.html
[flutter-tts-win-x64-cn]: https://k2-fsa.github.io/sherpa/onnx/flutter/tts-win-cn.html
[lazarus-subtitle]: https://k2-fsa.github.io/sherpa/onnx/lazarus/download-generated-subtitles.html
[lazarus-subtitle-cn]: https://k2-fsa.github.io/sherpa/onnx/lazarus/download-generated-subtitles-cn.html
[asr-models]: https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
[tts-models]: https://github.com/k2-fsa/sherpa-onnx/releases/tag/tts-models
[vad-models]: https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
[kws-models]: https://github.com/k2-fsa/sherpa-onnx/releases/tag/kws-models
[at-models]: https://github.com/k2-fsa/sherpa-onnx/releases/tag/audio-tagging-models
[sid-models]: https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-recongition-models
[slid-models]: https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-recongition-models
[punct-models]: https://github.com/k2-fsa/sherpa-onnx/releases/tag/punctuation-models
[GigaSpeech]: https://github.com/SpeechColab/GigaSpeech
[WenetSpeech]: https://github.com/wenet-e2e/WenetSpeech
... ...
#!/usr/bin/env bash
# Copyright (c) 2024 Xiaomi Corporation
#
# This script is to build sherpa-onnx for WebAssembly (VAD+ASR)
# Note: ASR here means non-streaming ASR
set -ex
if [ x"$EMSCRIPTEN" == x"" ]; then
if ! command -v emcc &> /dev/null; then
echo "Please install emscripten first"
echo ""
echo "You can use the following commands to install it:"
echo ""
echo "git clone https://github.com/emscripten-core/emsdk.git"
echo "cd emsdk"
echo "git pull"
echo "./emsdk install latest"
echo "./emsdk activate latest"
echo "source ./emsdk_env.sh"
exit 1
else
EMSCRIPTEN=$(dirname $(realpath $(which emcc)))
fi
fi
export EMSCRIPTEN=$EMSCRIPTEN
echo "EMSCRIPTEN: $EMSCRIPTEN"
if [ ! -f $EMSCRIPTEN/cmake/Modules/Platform/Emscripten.cmake ]; then
echo "Cannot find $EMSCRIPTEN/cmake/Modules/Platform/Emscripten.cmake"
echo "Please make sure you have installed emsdk correctly"
exit 1
fi
mkdir -p build-wasm-simd-vad-asr
pushd build-wasm-simd-vad-asr
export SHERPA_ONNX_IS_USING_BUILD_WASM_SH=ON
cmake \
-DCMAKE_INSTALL_PREFIX=./install \
-DCMAKE_BUILD_TYPE=Release \
-DCMAKE_TOOLCHAIN_FILE=$EMSCRIPTEN/cmake/Modules/Platform/Emscripten.cmake \
\
-DSHERPA_ONNX_ENABLE_PYTHON=OFF \
-DSHERPA_ONNX_ENABLE_TESTS=OFF \
-DSHERPA_ONNX_ENABLE_CHECK=OFF \
-DBUILD_SHARED_LIBS=OFF \
-DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
-DSHERPA_ONNX_ENABLE_JNI=OFF \
-DSHERPA_ONNX_ENABLE_TTS=OFF \
-DSHERPA_ONNX_ENABLE_C_API=ON \
-DSHERPA_ONNX_ENABLE_WEBSOCKET=OFF \
-DSHERPA_ONNX_ENABLE_GPU=OFF \
-DSHERPA_ONNX_ENABLE_WASM=ON \
-DSHERPA_ONNX_ENABLE_WASM_VAD_ASR=ON \
-DSHERPA_ONNX_ENABLE_BINARY=OFF \
-DSHERPA_ONNX_LINK_LIBSTDCPP_STATICALLY=OFF \
..
make -j2
make install
echo "pwd: $PWD"
cp -fv ../wasm/vad/sherpa-onnx-vad.js ./install/bin/wasm/vad-asr/
cp -fv ../wasm/asr/sherpa-onnx-asr.js ./install/bin/wasm/vad-asr/
ls -lh install/bin/wasm/vad-asr
... ...
#!/usr/bin/env python3
import argparse
from dataclasses import dataclass
from typing import List, Optional
import jinja2
def get_args():
parser = argparse.ArgumentParser()
parser.add_argument(
"--total",
type=int,
default=1,
help="Number of runners",
)
parser.add_argument(
"--index",
type=int,
default=0,
help="Index of the current runner",
)
return parser.parse_args()
@dataclass
class Model:
model_name: str
hf: str # huggingface space name
ms: str # modelscope space name
short_name: str
cmd: str = ""
def get_models():
models = [
Model(
model_name="sherpa-onnx-whisper-tiny.en",
hf="k2-fsa/web-assembly-vad-asr-sherpa-onnx-en-whisper-tiny",
ms="csukuangfj/web-assembly-vad-asr-sherpa-onnx-en-whisper-tiny",
short_name="vad-asr-en-whisper_tiny",
cmd="""
pushd $model_name
mv -v tiny.en-encoder.int8.onnx ../whisper-encoder.onnx
mv -v tiny.en-decoder.int8.onnx ../whisper-decoder.onnx
mv -v tiny.en-tokens.txt ../tokens.txt
popd
rm -rf $model_name
sed -i.bak 's/Zipformer/Whisper tiny.en supporting English 英文/g' ../index.html
git diff
""",
),
Model(
model_name="sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17",
hf="k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-en-ja-ko-cantonese-sense-voice",
ms="csukuangfj/web-assembly-vad-asr-sherpa-onnx-zh-en-jp-ko-cantonese-sense-voice",
short_name="vad-asr-zh_en_ja_ko_cantonese-sense_voice_small",
cmd="""
pushd $model_name
mv -v model.int8.onnx ../sense-voice.onnx
mv -v tokens.txt ../
popd
rm -rf $model_name
sed -i.bak 's/Zipformer/SenseVoice Small supporting English, Chinese, Japanese, Korean, Cantonese 中英日韩粤/g' ../index.html
git diff
""",
),
Model(
model_name="sherpa-onnx-paraformer-zh-2023-09-14",
hf="k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-en-paraformer",
ms="k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-en-paraformer",
short_name="vad-asr-zh_en-paraformer_large",
cmd="""
pushd $model_name
mv -v model.int8.onnx ../paraformer.onnx
mv -v tokens.txt ../
popd
rm -rf $model_name
sed -i.bak 's/Zipformer/Paraformer supporting Chinese, English 中英/g' ../index.html
git diff
""",
),
Model(
model_name="sherpa-onnx-paraformer-zh-small-2024-03-09",
hf="k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-en-paraformer-small",
ms="k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-en-paraformer-small",
short_name="vad-asr-zh_en-paraformer_small",
cmd="""
pushd $model_name
mv -v model.int8.onnx ../paraformer.onnx
mv -v tokens.txt ../
popd
rm -rf $model_name
sed -i.bak 's/Zipformer/Paraformer-small supporting Chinese, English 中英文/g' ../index.html
git diff
""",
),
Model(
model_name="sherpa-onnx-zipformer-gigaspeech-2023-12-12",
hf="k2-fsa/web-assembly-vad-asr-sherpa-onnx-en-zipformer-gigaspeech",
ms="k2-fsa/web-assembly-vad-asr-sherpa-onnx-en-zipformer-gigaspeech",
short_name="vad-asr-en-zipformer_gigaspeech",
cmd="""
pushd $model_name
mv encoder-epoch-30-avg-1.int8.onnx ../transducer-encoder.onnx
mv decoder-epoch-30-avg-1.onnx ../transducer-decoder.onnx
mv joiner-epoch-30-avg-1.int8.onnx ../transducer-joiner.onnx
mv tokens.txt ../
popd
rm -rf $model_name
sed -i.bak 's/Zipformer/Zipformer supporting English 英语/g' ../index.html
git diff
""",
),
Model(
model_name="icefall-asr-zipformer-wenetspeech-20230615",
hf="k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-zipformer-wenetspeech",
ms="k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-zipformer-wenetspeech",
short_name="vad-asr-zh-zipformer_wenetspeech",
cmd="""
pushd $model_name
mv -v data/lang_char/tokens.txt ../
mv -v exp/encoder-epoch-12-avg-4.int8.onnx ../transducer-encoder.onnx
mv -v exp/decoder-epoch-12-avg-4.onnx ../transducer-decoder.onnx
mv -v exp/joiner-epoch-12-avg-4.int8.onnx ../transducer-joiner.onnx
popd
rm -rf $model_name
sed -i.bak 's/Zipformer/Zipformer supporting Chinese 中文/g' ../index.html
git diff
""",
),
Model(
model_name="sherpa-onnx-zipformer-ja-reazonspeech-2024-08-01",
hf="k2-fsa/web-assembly-vad-asr-sherpa-onnx-ja-zipformer",
ms="csukuangfj/web-assembly-vad-asr-sherpa-onnx-ja-zipformer",
short_name="vad-asr-ja-zipformer_reazonspeech",
cmd="""
pushd $model_name
mv encoder-epoch-99-avg-1.int8.onnx ../transducer-encoder.onnx
mv decoder-epoch-99-avg-1.onnx ../transducer-decoder.onnx
mv joiner-epoch-99-avg-1.int8.onnx ../transducer-joiner.onnx
mv tokens.txt ../
popd
rm -rf $model_name
sed -i.bak 's/Zipformer/Zipformer supporting Japanese 日语/g' ../index.html
git diff
""",
),
Model(
model_name="sherpa-onnx-zipformer-thai-2024-06-20",
hf="k2-fsa/web-assembly-vad-asr-sherpa-onnx-th-zipformer",
ms="csukuangfj/web-assembly-vad-asr-sherpa-onnx-th-zipformer",
short_name="vad-asr-th-zipformer_gigaspeech2",
cmd="""
pushd $model_name
mv encoder-epoch-12-avg-5.int8.onnx ../transducer-encoder.onnx
mv decoder-epoch-12-avg-5.onnx ../transducer-decoder.onnx
mv joiner-epoch-12-avg-5.int8.onnx ../transducer-joiner.onnx
mv tokens.txt ../
popd
rm -rf $model_name
sed -i.bak 's/Zipformer/Zipformer supporting Thai 泰语/g' ../index.html
git diff
""",
),
Model(
model_name="sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04",
hf="k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-telespeech",
ms="k2-fsa/web-assembly-vad-asr-sherpa-onnx-zh-telespeech",
short_name="vad-asr-zh-telespeech",
cmd="""
pushd $model_name
mv model.int8.onnx ../telespeech.onnx
mv tokens.txt ../
popd
rm -rf $model_name
sed -i.bak 's/Zipformer/TeleSpeech-ASR supporting Chinese 多种中文方言/g' ../index.html
git diff
""",
),
]
return models
def main():
args = get_args()
index = args.index
total = args.total
assert 0 <= index < total, (index, total)
all_model_list = get_models()
num_models = len(all_model_list)
num_per_runner = num_models // total
if num_per_runner <= 0:
raise ValueError(f"num_models: {num_models}, num_runners: {total}")
start = index * num_per_runner
end = start + num_per_runner
remaining = num_models - args.total * num_per_runner
print(f"{index}/{total}: {start}-{end}/{num_models}")
d = dict()
d["model_list"] = all_model_list[start:end]
if index < remaining:
s = args.total * num_per_runner + index
d["model_list"].append(all_model_list[s])
print(f"{s}/{num_models}")
filename_list = [
"./run-vad-asr.sh",
]
for filename in filename_list:
environment = jinja2.Environment()
with open(f"{filename}.in") as f:
s = f.read()
template = environment.from_string(s)
s = template.render(**d)
with open(filename, "w") as f:
print(s, file=f)
if __name__ == "__main__":
main()
... ...
#!/usr/bin/env bash
#
# Build WebAssembly APPs for huggingface spaces and modelscope spaces
set -ex
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
}
SHERPA_ONNX_VERSION=$(grep "SHERPA_ONNX_VERSION" ./CMakeLists.txt | cut -d " " -f 2 | cut -d '"' -f 2)
{% for model in model_list %}
model_name={{ model.model_name }}
short_name={{ model.short_name }}
hf_name={{ model.hf }}
ms_name={{ model.ms }}
pushd wasm/vad-asr
git checkout .
rm -rf assets
mkdir assets
cd assets
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/${model_name}.tar.bz2
tar xvf ${model_name}.tar.bz2
rm ${model_name}.tar.bz2
{{ model.cmd }}
popd
ls -lh wasm/vad-asr/assets
rm -rf build-wasm-simd-vad-asr/install
rm -rf build-wasm-simd-vad-asr/wasm
./build-wasm-simd-vad-asr.sh
dst=sherpa-onnx-wasm-simd-${SHERPA_ONNX_VERSION}-${short_name}
mv build-wasm-simd-vad-asr/install/bin/wasm/vad-asr $dst
ls -lh $dst
tar cjfv $dst.tar.bz2 ./$dst
ls -lh *.tar.bz2
git config --global user.email "csukuangfj@gmail.com"
git config --global user.name "Fangjun Kuang"
export GIT_LFS_SKIP_SMUDGE=1
export GIT_CLONE_PROTECTION_ACTIVE=false
rm -rf ms
git clone https://www.modelscope.cn/studios/$ms_name.git ms
cd ms
cp -v ../$dst/* .
git status
git lfs track "*.data"
git lfs track "*.wasm"
ls -lh
git add .
git commit -m "update model"
git push https://oauth2:${MS_TOKEN}@www.modelscope.cn/studios/$ms_name.git
cd ..
rm -rf ms
rm -rf huggingface
git clone https://huggingface.co/spaces/$hf_name huggingface
cd huggingface
cp -v ../$dst/* .
git status
git lfs track "*.data"
git lfs track "*.wasm"
ls -lh
git add .
git commit -m "update model"
git push https://csukuangfj:$HF_TOKEN@huggingface.co/spaces/$hf_name main
cd ..
rm -rf huggingface
rm -rf $dst
ls -lh *.tar.bz2
{% endfor %}
... ...
... ... @@ -13,6 +13,7 @@
#include "sherpa-onnx/csrc/audio-tagging.h"
#include "sherpa-onnx/csrc/circular-buffer.h"
#include "sherpa-onnx/csrc/display.h"
#include "sherpa-onnx/csrc/file-utils.h"
#include "sherpa-onnx/csrc/keyword-spotter.h"
#include "sherpa-onnx/csrc/macros.h"
#include "sherpa-onnx/csrc/offline-punctuation.h"
... ... @@ -1638,3 +1639,7 @@ int32_t SherpaOnnxLinearResamplerResampleGetOutputSampleRate(
void SherpaOnnxLinearResamplerReset(SherpaOnnxLinearResampler *p) {
p->impl->Reset();
}
int32_t SherpaOnnxFileExists(const char *filename) {
return sherpa_onnx::FileExists(filename);
}
... ...
... ... @@ -1361,6 +1361,9 @@ SHERPA_ONNX_API int32_t SherpaOnnxLinearResamplerResampleGetInputSampleRate(
SHERPA_ONNX_API int32_t SherpaOnnxLinearResamplerResampleGetOutputSampleRate(
const SherpaOnnxLinearResampler *p);
// Return 1 if the file exists; return 0 if the file does not exist.
SHERPA_ONNX_API int32_t SherpaOnnxFileExists(const char *filename);
#if defined(__GNUC__)
#pragma GCC diagnostic pop
#endif
... ...
... ... @@ -14,6 +14,10 @@ if(SHERPA_ONNX_ENABLE_WASM_VAD)
add_subdirectory(vad)
endif()
if(SHERPA_ONNX_ENABLE_WASM_VAD_ASR)
add_subdirectory(vad-asr)
endif()
if(SHERPA_ONNX_ENABLE_WASM_NODEJS)
add_subdirectory(nodejs)
endif()
... ...
... ... @@ -80,3 +80,10 @@ assets fangjun$ tree -L 1
0 directories, 4 files
```
You can find example build scripts at:
- Streaming Zipformer (English + Chinese): https://github.com/k2-fsa/sherpa-onnx/blob/master/.github/workflows/ wasm-simd-hf-space-zh-en-asr-zipformer.yaml
- Streaming Zipformer (English): https://github.com/k2-fsa/sherpa-onnx/blob/master/.github/workflows/wasm-simd-hf-space-en-asr-zipformer.yaml
- Streaming Paraformer (English + Chinese): https://github.com/k2-fsa/sherpa-onnx/blob/master/.github/workflows/wasm-simd-hf-space-zh-en-asr-paraformer.yaml
- Streaming Paraformer (English + Chinese + Cantonese): https://github.com/k2-fsa/sherpa-onnx/blob/master/.github/workflows/wasm-simd-hf-space-zh-cantonese-en-asr-paraformer.yaml
... ...
... ... @@ -3,7 +3,7 @@
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width" />
<title>Next-gen Kaldi WebAssembly with sherpa-onnx for Text-to-speech</title>
<title>Next-gen Kaldi WebAssembly with sherpa-onnx for ASR</title>
<style>
h1,div {
text-align: center;
... ...
... ... @@ -30,3 +30,8 @@ assets fangjun$ tree -L 1
1 directory, 3 files
```
You can find example build scripts at:
- English TTS: https://github.com/k2-fsa/sherpa-onnx/blob/master/.github/workflows/wasm-simd-hf-space-en-tts.yaml
- German TTS: https://github.com/k2-fsa/sherpa-onnx/blob/master/.github/workflows/wasm-simd-hf-space-de-tts.yaml
... ...
if(NOT $ENV{SHERPA_ONNX_IS_USING_BUILD_WASM_SH})
message(FATAL_ERROR "Please use ./build-wasm-simd-vad.sh to build for wasm VAD")
endif()
if(NOT EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/assets/silero_vad.onnx" OR NOT EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/assets/tokens.txt")
message(FATAL_ERROR "Please read ${CMAKE_CURRENT_SOURCE_DIR}/assets/README.md before you continue")
endif()
set(exported_functions
# VAD
SherpaOnnxCreateCircularBuffer
SherpaOnnxDestroyCircularBuffer
SherpaOnnxCircularBufferPush
SherpaOnnxCircularBufferGet
SherpaOnnxCircularBufferFree
SherpaOnnxCircularBufferPop
SherpaOnnxCircularBufferSize
SherpaOnnxCircularBufferHead
SherpaOnnxCircularBufferReset
SherpaOnnxCreateVoiceActivityDetector
SherpaOnnxDestroyVoiceActivityDetector
SherpaOnnxVoiceActivityDetectorAcceptWaveform
SherpaOnnxVoiceActivityDetectorEmpty
SherpaOnnxVoiceActivityDetectorDetected
SherpaOnnxVoiceActivityDetectorPop
SherpaOnnxVoiceActivityDetectorClear
SherpaOnnxVoiceActivityDetectorFront
SherpaOnnxDestroySpeechSegment
SherpaOnnxVoiceActivityDetectorReset
SherpaOnnxVoiceActivityDetectorFlush
# non-streaming ASR
SherpaOnnxAcceptWaveformOffline
SherpaOnnxCreateOfflineRecognizer
SherpaOnnxCreateOfflineStream
SherpaOnnxDecodeMultipleOfflineStreams
SherpaOnnxDecodeOfflineStream
SherpaOnnxDestroyOfflineRecognizer
SherpaOnnxDestroyOfflineRecognizerResult
SherpaOnnxDestroyOfflineStream
SherpaOnnxDestroyOfflineStreamResultJson
SherpaOnnxGetOfflineStreamResult
SherpaOnnxGetOfflineStreamResultAsJson
#
SherpaOnnxFileExists
)
set(mangled_exported_functions)
foreach(x IN LISTS exported_functions)
list(APPEND mangled_exported_functions "_${x}")
endforeach()
list(JOIN mangled_exported_functions "," all_exported_functions)
include_directories(${CMAKE_SOURCE_DIR})
set(MY_FLAGS " -s FORCE_FILESYSTEM=1 -s INITIAL_MEMORY=512MB -s ALLOW_MEMORY_GROWTH=1")
string(APPEND MY_FLAGS " -sSTACK_SIZE=10485760 ") # 10MB
string(APPEND MY_FLAGS " -sEXPORTED_FUNCTIONS=[_CopyHeap,_malloc,_free,${all_exported_functions}] ")
string(APPEND MY_FLAGS "--preload-file ${CMAKE_CURRENT_SOURCE_DIR}/assets@. ")
string(APPEND MY_FLAGS " -sEXPORTED_RUNTIME_METHODS=['ccall','stringToUTF8','setValue','getValue','lengthBytesUTF8','UTF8ToString'] ")
message(STATUS "MY_FLAGS: ${MY_FLAGS}")
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${MY_FLAGS}")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${MY_FLAGS}")
set(CMAKE_EXECUTBLE_LINKER_FLAGS "${CMAKE_EXECUTBLE_LINKER_FLAGS} ${MY_FLAGS}")
if (NOT CMAKE_EXECUTABLE_SUFFIX STREQUAL ".js")
message(FATAL_ERROR "The default suffix for building executables should be .js!")
endif()
# set(CMAKE_EXECUTABLE_SUFFIX ".html")
add_executable(sherpa-onnx-wasm-main-vad-asr sherpa-onnx-wasm-main-vad-asr.cc)
target_link_libraries(sherpa-onnx-wasm-main-vad-asr sherpa-onnx-c-api)
install(TARGETS sherpa-onnx-wasm-main-vad-asr DESTINATION bin/wasm/vad-asr)
install(
FILES
"$<TARGET_FILE_DIR:sherpa-onnx-wasm-main-vad-asr>/sherpa-onnx-wasm-main-vad-asr.js"
"index.html"
"app-vad-asr.js"
"$<TARGET_FILE_DIR:sherpa-onnx-wasm-main-vad-asr>/sherpa-onnx-wasm-main-vad-asr.wasm"
"$<TARGET_FILE_DIR:sherpa-onnx-wasm-main-vad-asr>/sherpa-onnx-wasm-main-vad-asr.data"
DESTINATION
bin/wasm/vad-asr
)
... ...
// This file copies and modifies code
// from https://mdn.github.io/web-dictaphone/scripts/app.js
// and https://gist.github.com/meziantou/edb7217fddfbb70e899e
const startBtn = document.getElementById('startBtn');
const stopBtn = document.getElementById('stopBtn');
const clearBtn = document.getElementById('clearBtn');
const hint = document.getElementById('hint');
const soundClips = document.getElementById('sound-clips');
let textArea = document.getElementById('results');
let lastResult = '';
let resultList = [];
clearBtn.onclick = function() {
resultList = [];
textArea.value = getDisplayResult();
textArea.scrollTop = textArea.scrollHeight; // auto scroll
};
function getDisplayResult() {
let i = 0;
let ans = '';
for (let s in resultList) {
if (resultList[s] == '') {
continue;
}
if (resultList[s] == 'Speech detected') {
ans += '' + i + ': ' + resultList[s];
i += 1;
} else {
ans += ', ' + resultList[s] + '\n';
}
}
if (lastResult.length > 0) {
ans += '' + i + ': ' + lastResult + '\n';
}
return ans;
}
Module = {};
let audioCtx;
let mediaStream;
let expectedSampleRate = 16000;
let recordSampleRate; // the sampleRate of the microphone
let recorder = null; // the microphone
let leftchannel = []; // TODO: Use a single channel
let recordingLength = 0; // number of samples so far
let vad = null;
let buffer = null;
let recognizer = null;
let printed = false;
function fileExists(filename) {
const filenameLen = Module.lengthBytesUTF8(filename) + 1;
const buffer = Module._malloc(filenameLen);
Module.stringToUTF8(filename, buffer, filenameLen);
let exists = Module._SherpaOnnxFileExists(buffer);
Module._free(buffer);
return exists;
}
function createOfflineRecognizerSenseVoice() {}
function initOfflineRecognizer() {
let config = {
modelConfig: {
debug: 1,
tokens: './tokens.txt',
},
};
if (fileExists('sense-voice.onnx') == 1) {
config.modelConfig.senseVoice = {
model: './sense-voice.onnx',
useInverseTextNormalization: 1,
};
} else if (fileExists('whisper-encoder.onnx')) {
config.modelConfig.whisper = {
encoder: './whisper-encoder.onnx',
decoder: './whisper-decoder.onnx',
};
} else if (fileExists('transducer-encoder.onnx')) {
config.modelConfig.transducer = {
encoder: './transducer-encoder.onnx',
decoder: './transducer-decoder.onnx',
joiner: './transducer-joiner.onnx',
};
config.modelConfig.modelType = 'transducer';
} else if (fileExists('nemo-transducer-encoder.onnx')) {
config.modelConfig.transducer = {
encoder: './nemo-transducer-encoder.onnx',
decoder: './nemo-transducer-decoder.onnx',
joiner: './nemo-transducer-joiner.onnx',
};
config.modelConfig.modelType = 'nemo_transducer';
} else if (fileExists('paraformer.onnx')) {
config.modelConfig.paraformer = {
model: './paraformer.onnx',
};
} else if (fileExists('telespeech.onnx')) {
config.modelConfig.telespeechCtc = './telespeech.onnx';
} else {
console.log('Please specify a model.');
alert('Please specify a model.');
}
recognizer = new OfflineRecognizer(config, Module);
}
Module.onRuntimeInitialized = function() {
console.log('inited!');
hint.innerText = 'Model loaded! Please click start';
startBtn.disabled = false;
vad = createVad(Module);
console.log('vad is created!', vad);
buffer = new CircularBuffer(30 * 16000, Module);
console.log('CircularBuffer is created!', buffer);
initOfflineRecognizer();
};
if (navigator.mediaDevices.getUserMedia) {
console.log('getUserMedia supported.');
// see https://w3c.github.io/mediacapture-main/#dom-mediadevices-getusermedia
const constraints = {audio: true};
let onSuccess = function(stream) {
if (!audioCtx) {
audioCtx = new AudioContext({sampleRate: expectedSampleRate});
}
console.log(audioCtx);
recordSampleRate = audioCtx.sampleRate;
console.log('sample rate ' + recordSampleRate);
// creates an audio node from the microphone incoming stream
mediaStream = audioCtx.createMediaStreamSource(stream);
console.log('media stream', mediaStream);
// https://developer.mozilla.org/en-US/docs/Web/API/AudioContext/createScriptProcessor
// bufferSize: the onaudioprocess event is called when the buffer is full
var bufferSize = 4096;
var numberOfInputChannels = 1;
var numberOfOutputChannels = 2;
if (audioCtx.createScriptProcessor) {
recorder = audioCtx.createScriptProcessor(
bufferSize, numberOfInputChannels, numberOfOutputChannels);
} else {
recorder = audioCtx.createJavaScriptNode(
bufferSize, numberOfInputChannels, numberOfOutputChannels);
}
console.log('recorder', recorder);
recorder.onaudioprocess = function(e) {
let samples = new Float32Array(e.inputBuffer.getChannelData(0))
samples = downsampleBuffer(samples, expectedSampleRate);
buffer.push(samples);
while (buffer.size() > vad.config.sileroVad.windowSize) {
const s = buffer.get(buffer.head(), vad.config.sileroVad.windowSize);
vad.acceptWaveform(s);
buffer.pop(vad.config.sileroVad.windowSize);
if (vad.isDetected() && !printed) {
printed = true;
lastResult = 'Speech detected';
}
if (!vad.isDetected()) {
printed = false;
if (lastResult != '') {
resultList.push(lastResult);
}
lastResult = '';
}
while (!vad.isEmpty()) {
const segment = vad.front();
const duration = segment.samples.length / expectedSampleRate;
let durationStr = `Duration: ${duration.toFixed(3)} seconds`;
vad.pop();
// non-streaming asr
const stream = recognizer.createStream();
stream.acceptWaveform(expectedSampleRate, segment.samples);
recognizer.decode(stream);
let recognitionResult = recognizer.getResult(stream);
console.log(recognitionResult);
let text = recognitionResult.text;
stream.free();
console.log(text);
if (text != '') {
durationStr += `. Result: ${text}`;
}
resultList.push(durationStr);
// now save the segment to a wav file
let buf = new Int16Array(segment.samples.length);
for (var i = 0; i < segment.samples.length; ++i) {
let s = segment.samples[i];
if (s >= 1)
s = 1;
else if (s <= -1)
s = -1;
buf[i] = s * 32767;
}
let clipName = new Date().toISOString() + '--' + durationStr;
const clipContainer = document.createElement('article');
const clipLabel = document.createElement('p');
const audio = document.createElement('audio');
const deleteButton = document.createElement('button');
clipContainer.classList.add('clip');
audio.setAttribute('controls', '');
deleteButton.textContent = 'Delete';
deleteButton.className = 'delete';
clipLabel.textContent = clipName;
clipContainer.appendChild(audio);
clipContainer.appendChild(clipLabel);
clipContainer.appendChild(deleteButton);
soundClips.appendChild(clipContainer);
audio.controls = true;
const blob = toWav(buf);
leftchannel = [];
const audioURL = window.URL.createObjectURL(blob);
audio.src = audioURL;
deleteButton.onclick = function(e) {
let evtTgt = e.target;
evtTgt.parentNode.parentNode.removeChild(evtTgt.parentNode);
};
clipLabel.onclick = function() {
const existingName = clipLabel.textContent;
const newClipName = prompt('Enter a new name for your sound clip?');
if (newClipName === null) {
clipLabel.textContent = existingName;
} else {
clipLabel.textContent = newClipName;
}
};
}
}
textArea.value = getDisplayResult();
textArea.scrollTop = textArea.scrollHeight; // auto scroll
};
startBtn.onclick = function() {
mediaStream.connect(recorder);
recorder.connect(audioCtx.destination);
console.log('recorder started');
stopBtn.disabled = false;
startBtn.disabled = true;
};
stopBtn.onclick = function() {
vad.reset();
buffer.reset();
console.log('recorder stopped');
// stopBtn recording
recorder.disconnect(audioCtx.destination);
mediaStream.disconnect(recorder);
startBtn.style.background = '';
startBtn.style.color = '';
// mediaRecorder.requestData();
stopBtn.disabled = true;
startBtn.disabled = false;
};
};
let onError = function(err) {
console.log('The following error occured: ' + err);
};
navigator.mediaDevices.getUserMedia(constraints).then(onSuccess, onError);
} else {
console.log('getUserMedia not supported on your browser!');
alert('getUserMedia not supported on your browser!');
}
// this function is copied/modified from
// https://gist.github.com/meziantou/edb7217fddfbb70e899e
function flatten(listOfSamples) {
let n = 0;
for (let i = 0; i < listOfSamples.length; ++i) {
n += listOfSamples[i].length;
}
let ans = new Int16Array(n);
let offset = 0;
for (let i = 0; i < listOfSamples.length; ++i) {
ans.set(listOfSamples[i], offset);
offset += listOfSamples[i].length;
}
return ans;
}
// this function is copied/modified from
// https://gist.github.com/meziantou/edb7217fddfbb70e899e
function toWav(samples) {
let buf = new ArrayBuffer(44 + samples.length * 2);
var view = new DataView(buf);
// http://soundfile.sapp.org/doc/WaveFormat/
// F F I R
view.setUint32(0, 0x46464952, true); // chunkID
view.setUint32(4, 36 + samples.length * 2, true); // chunkSize
// E V A W
view.setUint32(8, 0x45564157, true); // format
//
// t m f
view.setUint32(12, 0x20746d66, true); // subchunk1ID
view.setUint32(16, 16, true); // subchunk1Size, 16 for PCM
view.setUint32(20, 1, true); // audioFormat, 1 for PCM
view.setUint16(22, 1, true); // numChannels: 1 channel
view.setUint32(24, expectedSampleRate, true); // sampleRate
view.setUint32(28, expectedSampleRate * 2, true); // byteRate
view.setUint16(32, 2, true); // blockAlign
view.setUint16(34, 16, true); // bitsPerSample
view.setUint32(36, 0x61746164, true); // Subchunk2ID
view.setUint32(40, samples.length * 2, true); // subchunk2Size
let offset = 44;
for (let i = 0; i < samples.length; ++i) {
view.setInt16(offset, samples[i], true);
offset += 2;
}
return new Blob([view], {type: 'audio/wav'});
}
// this function is copied from
// https://github.com/awslabs/aws-lex-browser-audio-capture/blob/master/lib/worker.js#L46
function downsampleBuffer(buffer, exportSampleRate) {
if (exportSampleRate === recordSampleRate) {
return buffer;
}
var sampleRateRatio = recordSampleRate / exportSampleRate;
var newLength = Math.round(buffer.length / sampleRateRatio);
var result = new Float32Array(newLength);
var offsetResult = 0;
var offsetBuffer = 0;
while (offsetResult < result.length) {
var nextOffsetBuffer = Math.round((offsetResult + 1) * sampleRateRatio);
var accum = 0, count = 0;
for (var i = offsetBuffer; i < nextOffsetBuffer && i < buffer.length; i++) {
accum += buffer[i];
count++;
}
result[offsetResult] = accum / count;
offsetResult++;
offsetBuffer = nextOffsetBuffer;
}
return result;
};
... ...
# Introduction
## Download VAD models
Please download
https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
and put `silero_vad.onnx` into the current directory, i.e., `wasm/vad/assets`.
## Download non-streaming ASR models
Please refer to
https://k2-fsa.github.io/sherpa/onnx/pretrained_models/index.html
to download a non-streaming ASR model, i.e., an offline ASR model.
After downloading, you should rename the model files.
Please refer to
https://k2-fsa.github.io/sherpa/onnx/lazarus/generate-subtitles.html#download-a-speech-recognition-model
for how to rename.
You can find example build scripts at the following address:
https://github.com/k2-fsa/sherpa-onnx/blob/master/.github/workflows/wasm-simd-hf-space-vad-asr.yaml
... ...
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width" />
<title>Next-gen Kaldi WebAssembly with sherpa-onnx for VAD + ASR</title>
<style>
h1,div {
text-align: center;
}
textarea {
width:100%;
}
</style>
</head>
<body>
<h1>
Next-gen Kaldi + WebAssembly<br/>
VAD+ASR Demo with <a href="https://github.com/k2-fsa/sherpa-onnx">sherpa-onnx</a><br/>
(with Zipformer)
</h1>
<div>
<span id="hint">Loading model ... ...</span>
<br/>
<br/>
<button id="startBtn" disabled>Start</button>
<button id="stopBtn" disabled>Stop</button>
<button id="clearBtn">Clear</button>
<br/>
<br/>
<textarea id="results" rows="10" readonly></textarea>
</div>
<section flex="1" overflow="auto" id="sound-clips">
</section>
<script src="sherpa-onnx-asr.js"></script>
<script src="sherpa-onnx-vad.js"></script>
<script src="app-vad-asr.js"></script>
<script src="sherpa-onnx-wasm-main-vad-asr.js"></script>
</body>
... ...
../asr/sherpa-onnx-asr.js
\ No newline at end of file
... ...
../vad/sherpa-onnx-vad.js
\ No newline at end of file
... ...
// wasm/sherpa-onnx-wasm-main-vad-asr.cc
//
// Copyright (c) 2024 Xiaomi Corporation
#include <stdio.h>
#include <algorithm>
#include <memory>
#include "sherpa-onnx/c-api/c-api.h"
// see also
// https://emscripten.org/docs/porting/connecting_cpp_and_javascript/Interacting-with-code.html
extern "C" {
void CopyHeap(const char *src, int32_t num_bytes, char *dst) {
std::copy(src, src + num_bytes, dst);
}
}
... ...
... ... @@ -3,3 +3,6 @@
Please download
https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
and put `silero_vad.onnx` into the current directory, i.e., `wasm/vad/assets`.
You can find example build script at
https://github.com/k2-fsa/sherpa-onnx/blob/master/.github/workflows/wasm-simd-hf-space-silero-vad.yaml
... ...
... ... @@ -3,7 +3,7 @@
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width" />
<title>Next-gen Kaldi WebAssembly with sherpa-onnx for Text-to-speech</title>
<title>Next-gen Kaldi WebAssembly with sherpa-onnx for VAD</title>
<style>
h1,div {
text-align: center;
... ...
... ... @@ -172,7 +172,6 @@ class Vad {
constructor(configObj, Module) {
this.config = configObj;
const config = initSherpaOnnxVadModelConfig(configObj, Module);
Module._MyPrint(config.ptr);
const handle = Module._SherpaOnnxCreateVoiceActivityDetector(
config.ptr, configObj.bufferSizeInSeconds || 30);
freeConfig(config, Module);
... ...