EngineX-Iluvatar/enginex_bi_series-sherpa-onnx
Archived
15
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Add Lazarus example for generating subtitles using Silero VAD with non-streaming ASR (#1251)

Browse Source
This commit is contained in:
Fangjun Kuang
2024-08-15 22:19:45 +08:00
committed by GitHub
parent 97a6a2a16a
commit fbe35ba736
32 changed files with 1697 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

369
.github/workflows/lazarus.yaml vendored Normal file
View File

@@ -0,0 +1,369 @@
name: lazarus
on:
push:
branches:
- master
paths:
- '.github/workflows/lazarus.yaml'
- 'CMakeLists.txt'
- 'cmake/**'
- 'lazarus-examples/**'
- 'sherpa-onnx/csrc/*'
- 'sherpa-onnx/c-api/*'
- 'sherpa-onnx/pascal-api/*'
pull_request:
branches:
- master
paths:
- '.github/workflows/lazarus.yaml'
- 'CMakeLists.txt'
- 'cmake/**'
- 'lazarus-examples/**'
- 'sherpa-onnx/csrc/*'
- 'sherpa-onnx/c-api/*'
- 'sherpa-onnx/pascal-api/*'
workflow_dispatch:
concurrency:
group: lazarus-${{ github.ref }}
cancel-in-progress: true
permissions:
contents: read
jobs:
build:
name: ${{ matrix.os }}
runs-on: ${{ matrix.os }}
strategy:
fail-fast: false
matrix:
os: [ubuntu-20.04, macos-latest, macos-13, windows-latest]
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: ccache
uses: hendrikmuhs/ccache-action@v1.2
with:
key: ${{ matrix.os }}
# See https://github.com/gcarreno/setup-lazarus
- uses: gcarreno/setup-lazarus@v3
with:
lazarus-version: "stable"
with-cache: true
- name: Lazarus info
shell: bash
run: |
which lazbuild
lazbuild --help
- name: FPC info
shell: bash
run: |
which fpc
fpc -i
- name: OS info
shell: bash
run: |
uname -a
- name: Install patchelf for ubuntu
if: matrix.os == 'ubuntu-20.04'
shell: bash
run: |
sudo apt-get update -q
sudo apt-get install -q -y patchelf
- name: Show Patchelf version (ubuntu)
if: matrix.os == 'ubuntu-20.04'
shell: bash
run: |
patchelf --version
patchelf --help
which patchelf
- name: Configure CMake
shell: bash
run: |
export CMAKE_CXX_COMPILER_LAUNCHER=ccache
export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
cmake --version
mkdir build
cd build
os=${{ matrix.os }}
if [[ $os == 'windows-latest' || $os == 'ubuntu-20.04' ]]; then
BUILD_SHARED_LIBS=ON
else
BUILD_SHARED_LIBS=OFF
fi
cmake \
-DCMAKE_INSTALL_PREFIX=./install \
-D BUILD_SHARED_LIBS=$BUILD_SHARED_LIBS \
-D SHERPA_ONNX_ENABLE_BINARY=OFF \
-D CMAKE_BUILD_TYPE=Release \
..
- name: Build sherpa-onnx
shell: bash
run: |
export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
cd build
cmake --build . --target install --config Release -j 2
ls -lh install/lib/
cp -v install/lib/*.dll ../lazarus-examples/generate_subtitles/ || true
cp -v install/lib/*.so* ../lazarus-examples/generate_subtitles/ || true
- name: Build generating subtitles
shell: bash
run: |
cd lazarus-examples/generate_subtitles
os=${{ matrix.os }}
if [[ $os == macos-13 ]]; then
lazbuild --verbose --build-mode=Release --widgetset=cocoa ./generate_subtitles.lpi
elif [[ $os == macos-latest ]]; then
lazbuild --verbose --build-mode=Release --widgetset=cocoa --cpu=aarch64 ./generate_subtitles.lpi
elif [[ $os == 'ubuntu-20.04' ]]; then
lazbuild --verbose --build-mode=Release-Linux ./generate_subtitles.lpi
else
lazbuild --verbose --build-mode=Release ./generate_subtitles.lpi
fi
- name: Display generating subtitles
shell: bash
run: |
cd lazarus-examples/generate_subtitles
ls -lh
- name: Collect generating subtitles (Ubuntu)
if: matrix.os == 'ubuntu-20.04'
shell: bash
run: |
SHERPA_ONNX_VERSION=$(grep "SHERPA_ONNX_VERSION" ./CMakeLists.txt | cut -d " " -f 2 | cut -d '"' -f 2)
cd lazarus-examples/generate_subtitles
ls -lh
readelf -d ./generate_subtitles
echo '----------'
ldd ./generate_subtitles
d=generate_subtitles-linux-x64-$SHERPA_ONNX_VERSION
echo "---before running patchelf---"
readelf -d ./generate_subtitles
patchelf --set-rpath '$ORIGIN' ./generate_subtitles
echo "---after running patchelf---"
readelf -d ./generate_subtitles
mkdir -p $d
cp -v ./generate_subtitles $d/
cp -v *.so $d/
mv -v $d /tmp/linux-x64
ls -lh /tmp/linux-x64
- name: Collect generating subtitles (windows)
if: matrix.os == 'windows-latest'
shell: bash
run: |
SHERPA_ONNX_VERSION=$(grep "SHERPA_ONNX_VERSION" ./CMakeLists.txt | cut -d " " -f 2 | cut -d '"' -f 2)
cd lazarus-examples/generate_subtitles
ls -lh
d=generate-subtitles-windows-x64-$SHERPA_ONNX_VERSION
mkdir -p $d
cp -v ./generate_subtitles.exe $d/
cp -v onnxruntime.dll $d/
cp -v sherpa-onnx-c-api.dll $d/
mv $d ../../windows-x64
cd ../..
ls -lh windows-x64
- name: Collect generating subtitles (macos)
if: matrix.os == 'macos-13' || matrix.os == 'macos-latest'
shell: bash
run: |
SHERPA_ONNX_VERSION=$(grep "SHERPA_ONNX_VERSION" ./CMakeLists.txt | cut -d " " -f 2 | cut -d '"' -f 2)
cd lazarus-examples/generate_subtitles
ls -lh
file ./generate_subtitles
echo '----------'
otool -L ./generate_subtitles
rm -v generate_subtitles.app/Contents/MacOS/generate_subtitles
cp -v ./generate_subtitles generate_subtitles.app/Contents/MacOS/generate_subtitles
chmod +x generate_subtitles.app/Contents/MacOS/generate_subtitles
if [[ ${{ matrix.os }} == 'macos-latest' ]]; then
mv generate_subtitles.app /tmp/macos-arm64
else
mv generate_subtitles.app /tmp/macos-x64
d=generate-subtitles-macos-x64-$SHERPA_ONNX_VERSION.app
fi
ls -lh /tmp
echo "---"
ls -lh /tmp/macos-*
- uses: actions/upload-artifact@v4
if: matrix.os == 'ubuntu-20.04'
with:
name: linux-x64
path: /tmp/linux-x64
- uses: actions/upload-artifact@v4
if: matrix.os == 'macos-latest'
with:
name: macos-arm64
path: /tmp/macos-arm64
- uses: actions/upload-artifact@v4
if: matrix.os == 'macos-13'
with:
name: macos-x64
path: /tmp/macos-x64
- uses: actions/upload-artifact@v4
if: matrix.os == 'windows-latest'
with:
name: windows-x64
path: ./windows-x64
release:
runs-on: ${{ matrix.os }}
needs: [build]
strategy:
fail-fast: false
matrix:
os: [ubuntu-latest]
total: ["2"]
index: ["0", "1"]
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Retrieve artifact from windows x64
uses: actions/download-artifact@v4
with:
name: windows-x64
path: /tmp/windows-x64
- name: Retrieve artifact from linux x64
uses: actions/download-artifact@v4
with:
name: linux-x64
path: /tmp/linux-x64
- name: Retrieve artifact from macos x64
uses: actions/download-artifact@v4
with:
name: macos-x64
path: /tmp/macos-x64
- name: Retrieve artifact from macos arm64
uses: actions/download-artifact@v4
with:
name: macos-arm64
path: /tmp/macos-arm64
- name: Display build files
shell: bash
run: |
ls -lh /tmp
echo "---linux-x64---"
ls -lh /tmp/linux-x64/
readelf -d /tmp/linux-x64/generate_subtitles
echo "---"
ldd /tmp/linux-x64/generate_subtitles
echo "---macos-x64---"
ls -lh /tmp/macos-x64/
mkdir -p /tmp/macos-x64/Contents/Resources
chmod +x /tmp/macos-x64/Contents/MacOS/generate_subtitles
echo "---macos-arm64---"
ls -lh /tmp/macos-arm64/
mkdir -p /tmp/macos-arm64/Contents/Resources
chmod +x /tmp/macos-arm64/Contents/MacOS/generate_subtitles
echo "---windows-x64---"
ls -lh /tmp/windows-x64/
- name: Install Python dependencies
shell: bash
run: |
python3 -m pip install --upgrade pip jinja2
- name: Generate build script
shell: bash
run: |
cd scripts/lazarus
total=${{ matrix.total }}
index=${{ matrix.index }}
./generate-subtitles.py --total $total --index $index
chmod +x build-generate-subtitles.sh
mv -v ./build-generate-subtitles.sh ../..
- name: Generate tar files
shell: bash
run: |
./build-generate-subtitles.sh
- name: Display tar files
shell: bash
run: |
ls -lh /tmp/out
- name: Publish to huggingface
env:
HF_TOKEN: ${{ secrets.HF_TOKEN }}
uses: nick-fields/retry@v3
with:
max_attempts: 20
timeout_seconds: 200
shell: bash
command: |
git config --global user.email "csukuangfj@gmail.com"
git config --global user.name "Fangjun Kuang"
rm -rf huggingface
export GIT_LFS_SKIP_SMUDGE=1
export GIT_CLONE_PROTECTION_ACTIVE=false
SHERPA_ONNX_VERSION=$(grep "SHERPA_ONNX_VERSION" ./CMakeLists.txt | cut -d " " -f 2 | cut -d '"' -f 2)
echo "SHERPA_ONNX_VERSION $SHERPA_ONNX_VERSION"
git clone https://huggingface.co/csukuangfj/sherpa-onnx-bin huggingface
cd huggingface
git fetch
git pull
git merge -m "merge remote" --ff origin main
d=generate-subtitles/$SHERPA_ONNX_VERSION
mkdir -p $d
cp -v /tmp/out/*.tar.bz2 $d/
git status
git lfs track "*.tar.bz2"
git add .
git commit -m "add more files"
git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/sherpa-onnx-bin main

30
lazarus-examples/.gitignore vendored Normal file
View File

@@ -0,0 +1,30 @@
# Lazarus compiler-generated binaries (safe to delete)
*.exe
*.dll
*.so
*.dylib
*.lrs
*.res
*.compiled
*.dbg
*.ppu
*.o
*.or
*.a
# Lazarus autogenerated files (duplicated info)
*.rst
*.rsj
*.lrt
# Lazarus local files (user-specific info)
*.lps
# Lazarus backups and unit output folders.
# These can be changed by user in Lazarus/project options.
backup/
*.bak
lib/
# Application bundle for Mac OS
*.app/

3
lazarus-examples/generate_subtitles/.gitignore vendored Normal file
View File

@@ -0,0 +1,3 @@
generate_subtitles.app
generate_subtitles
generate_subtitles.dSYM

BIN
lazarus-examples/generate_subtitles/generate_subtitles.ico Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 130 KiB

208
lazarus-examples/generate_subtitles/generate_subtitles.lpi Normal file
View File

@@ -0,0 +1,208 @@
<?xml version="1.0" encoding="UTF-8"?>
<CONFIG>
<ProjectOptions>
<Version Value="12"/>
<PathDelim Value="\"/>
<General>
<SessionStorage Value="InProjectDir"/>
<Title Value="generate_subtitles"/>
<Scaled Value="True"/>
<ResourceType Value="res"/>
<UseXPManifest Value="True"/>
<XPManifest>
<DpiAware Value="True"/>
</XPManifest>
<Icon Value="0"/>
</General>
<BuildModes>
<Item Name="Default" Default="True"/>
<Item Name="Debug">
<CompilerOptions>
<Version Value="11"/>
<PathDelim Value="\"/>
<Target>
<Filename Value="generate_subtitles"/>
</Target>
<SearchPaths>
<IncludeFiles Value="$(ProjOutDir)"/>
<Libraries Value="..\..\build-static\install\lib;..\..\build\install\lib"/>
<OtherUnitFiles Value="..\..\sherpa-onnx\pascal-api"/>
<UnitOutputDirectory Value="lib\$(TargetCPU)-$(TargetOS)"/>
</SearchPaths>
<Parsing>
<SyntaxOptions>
<IncludeAssertionCode Value="True"/>
</SyntaxOptions>
</Parsing>
<CodeGeneration>
<Checks>
<IOChecks Value="True"/>
<RangeChecks Value="True"/>
<OverflowChecks Value="True"/>
<StackChecks Value="True"/>
</Checks>
<VerifyObjMethodCallValidity Value="True"/>
</CodeGeneration>
<Linking>
<Debugging>
<DebugInfoType Value="dsDwarf3"/>
<UseHeaptrc Value="True"/>
<TrashVariables Value="True"/>
<StripSymbols Value="True"/>
<UseExternalDbgSyms Value="True"/>
</Debugging>
<Options>
<Win32>
<GraphicApplication Value="True"/>
</Win32>
</Options>
</Linking>
</CompilerOptions>
</Item>
<Item Name="Release">
<CompilerOptions>
<Version Value="11"/>
<PathDelim Value="\"/>
<Target>
<Filename Value="generate_subtitles"/>
</Target>
<SearchPaths>
<IncludeFiles Value="$(ProjOutDir)"/>
<Libraries Value="..\..\build-static\install\lib;..\..\build\install\lib"/>
<OtherUnitFiles Value="..\..\sherpa-onnx\pascal-api"/>
<UnitOutputDirectory Value="lib\$(TargetCPU)-$(TargetOS)"/>
</SearchPaths>
<CodeGeneration>
<SmartLinkUnit Value="True"/>
<Optimizations>
<OptimizationLevel Value="3"/>
</Optimizations>
</CodeGeneration>
<Linking>
<Debugging>
<GenerateDebugInfo Value="False"/>
<RunWithoutDebug Value="True"/>
<StripSymbols Value="True"/>
</Debugging>
<LinkSmart Value="True"/>
<Options>
<Win32>
<GraphicApplication Value="True"/>
</Win32>
</Options>
</Linking>
</CompilerOptions>
</Item>
<Item Name="Release-Linux">
<CompilerOptions>
<Version Value="11"/>
<PathDelim Value="\"/>
<Target>
<Filename Value="generate_subtitles"/>
</Target>
<SearchPaths>
<IncludeFiles Value="$(ProjOutDir)"/>
<Libraries Value="..\..\build-static\install\lib;..\..\build\install\lib"/>
<OtherUnitFiles Value="..\..\sherpa-onnx\pascal-api"/>
<UnitOutputDirectory Value="lib\$(TargetCPU)-$(TargetOS)"/>
</SearchPaths>
<CodeGeneration>
<SmartLinkUnit Value="True"/>
<Optimizations>
<OptimizationLevel Value="3"/>
</Optimizations>
</CodeGeneration>
<Linking>
<Debugging>
<GenerateDebugInfo Value="False"/>
<RunWithoutDebug Value="True"/>
<StripSymbols Value="True"/>
</Debugging>
<LinkSmart Value="True"/>
<Options>
<Win32>
<GraphicApplication Value="True"/>
</Win32>
</Options>
</Linking>
<Other>
<CustomOptions Value="-dSHERPA_ONNX_USE_SHARED_LIBS"/>
</Other>
</CompilerOptions>
</Item>
</BuildModes>
<PublishOptions>
<Version Value="2"/>
<UseFileFilters Value="True"/>
</PublishOptions>
<RunParams>
<FormatVersion Value="2"/>
</RunParams>
<RequiredPackages>
<Item>
<PackageName Value="LCL"/>
</Item>
</RequiredPackages>
<Units>
<Unit>
<Filename Value="generate_subtitles.lpr"/>
<IsPartOfProject Value="True"/>
</Unit>
<Unit>
<Filename Value="unit1.pas"/>
<IsPartOfProject Value="True"/>
<ComponentName Value="Form1"/>
<HasResources Value="True"/>
<ResourceBaseClass Value="Form"/>
<UnitName Value="Unit1"/>
</Unit>
<Unit>
<Filename Value="my_worker.pas"/>
<IsPartOfProject Value="True"/>
</Unit>
</Units>
</ProjectOptions>
<CompilerOptions>
<Version Value="11"/>
<PathDelim Value="\"/>
<Target>
<Filename Value="generate_subtitles"/>
</Target>
<SearchPaths>
<IncludeFiles Value="$(ProjOutDir)"/>
<Libraries Value="..\..\build-static\install\lib;..\..\build\install\lib"/>
<OtherUnitFiles Value="..\..\sherpa-onnx\pascal-api"/>
<UnitOutputDirectory Value="lib\$(TargetCPU)-$(TargetOS)"/>
</SearchPaths>
<CodeGeneration>
<Optimizations>
<OptimizationLevel Value="2"/>
</Optimizations>
</CodeGeneration>
<Linking>
<Debugging>
<GenerateDebugInfo Value="False"/>
<DebugInfoType Value="dsDwarf3"/>
<StripSymbols Value="True"/>
</Debugging>
<Options>
<Win32>
<GraphicApplication Value="True"/>
</Win32>
</Options>
</Linking>
</CompilerOptions>
<Debugging>
<Exceptions>
<Item>
<Name Value="EAbort"/>
</Item>
<Item>
<Name Value="ECodetoolError"/>
</Item>
<Item>
<Name Value="EFOpenError"/>
</Item>
</Exceptions>
</Debugging>
</CONFIG>

26
lazarus-examples/generate_subtitles/generate_subtitles.lpr Normal file
View File

@@ -0,0 +1,26 @@
program generate_subtitles;
{$mode objfpc}{$H+}
uses
{$IFDEF UNIX}
cthreads,
cmem,
{$ENDIF}
{$IFDEF HASAMIGA}
athreads,
{$ENDIF}
Interfaces, // this includes the LCL widgetset
Forms, unit1, my_worker
{ you can add units after this };
{$R *.res}
begin
RequireDerivedFormResource:=True;
Application.Scaled:=True;
Application.Initialize;
Application.CreateForm(TForm1, Form1);
Application.Run;
end.

160
lazarus-examples/generate_subtitles/my_worker.pas Normal file
View File

@@ -0,0 +1,160 @@
unit my_worker;
{$mode ObjFPC}{$H+}
{
See
https://wiki.lazarus.freepascal.org/Multithreaded_Application_Tutorial
https://www.freepascal.org/docs-html/rtl/classes/tthread.html
}
interface
uses
{$IFDEF UNIX}
cthreads,
cmem,
{$ENDIF}
{$IFDEF HASAMIGA}
athreads,
{$ENDIF}
Classes, SysUtils;
type
TMyWorkerThread = class(TThread)
private
Status: AnsiString;
StartTime: Single;
StopTime: Single;
TotalDuration: Single;
procedure ShowStatus;
procedure ShowProgress;
protected
procedure Execute; override;
public
WaveFilename: AnsiString;
Constructor Create(CreateSuspended : boolean; Filename: AnsiString);
end;
var
MyWorkerThread: TMyWorkerThread;
implementation
uses
unit1, sherpa_onnx;
constructor TMyWorkerThread.Create(CreateSuspended : boolean; Filename: AnsiString);
begin
inherited Create(CreateSuspended);
WaveFilename := Filename;
FreeOnTerminate := True;
end;
procedure TMyWorkerThread.ShowStatus;
begin
Form1.UpdateResult(Status, StartTime, StopTime, TotalDuration);
end;
procedure TMyWorkerThread.ShowProgress;
begin
Form1.UpdateProgress(StopTime, TotalDuration);
end;
procedure TMyWorkerThread.Execute;
var
Wave: TSherpaOnnxWave;
WindowSize: Integer;
Offset: Integer;
SpeechSegment: TSherpaOnnxSpeechSegment;
Duration: Single;
Stream: TSherpaOnnxOfflineStream;
RecognitionResult: TSherpaOnnxOfflineRecognizerResult;
begin
Wave := SherpaOnnxReadWave(WaveFilename);
TotalDuration := 0;
StartTime := 0;
StopTime := 0;
if (Wave.Samples = nil) or (Length(Wave.Samples) = 0) then
begin
Status := Format('Failed to read %s. We only support 1 channel, 16000Hz, 16-bit encoded wave files',
[Wavefilename]);
Synchronize(@ShowStatus);
Exit;
end;
if Wave.SampleRate <> 16000 then
begin
Status := Format('Expected sample rate 16000. Given %d. Please select a new file', [Wave.SampleRate]);
Synchronize(@ShowStatus);
Exit;
end;
TotalDuration := Length(Wave.Samples) / Wave.SampleRate;
WindowSize := Form1.Vad.Config.SileroVad.WindowSize;
Offset := 0;
Form1.Vad.Reset;
while not Terminated and (Offset + WindowSize <= Length(Wave.Samples)) do
begin
Form1.Vad.AcceptWaveform(Wave.Samples, Offset, WindowSize);
Offset += WindowSize;
StopTime := Offset / Wave.SampleRate;
if (Offset mod 20480) = 0 then
Synchronize(@ShowProgress);
while not Terminated and not Form1.Vad.IsEmpty do
begin
SpeechSegment := Form1.Vad.Front;
Form1.Vad.Pop;
Stream := Form1.OfflineRecognizer.CreateStream;
Stream.AcceptWaveform(SpeechSegment.Samples, Wave.SampleRate);
Form1.OfflineRecognizer.Decode(Stream);
RecognitionResult := Form1.OfflineRecognizer.GetResult(Stream);
StartTime := SpeechSegment.Start / Wave.SampleRate;
Duration := Length(SpeechSegment.Samples) / Wave.SampleRate;
StopTime := StartTime + Duration;
Status := RecognitionResult.Text;
Synchronize(@ShowStatus);
FreeAndNil(Stream);
end;
end;
Form1.Vad.Flush;
while not Terminated and not Form1.Vad.IsEmpty do
begin
SpeechSegment := Form1.Vad.Front;
Form1.Vad.Pop;
Stream := Form1.OfflineRecognizer.CreateStream;
Stream.AcceptWaveform(SpeechSegment.Samples, Wave.SampleRate);
Form1.OfflineRecognizer.Decode(Stream);
RecognitionResult := Form1.OfflineRecognizer.GetResult(Stream);
StartTime := SpeechSegment.Start / Wave.SampleRate;
Duration := Length(SpeechSegment.Samples) / Wave.SampleRate;
StopTime := StartTime + Duration;
Status := RecognitionResult.Text;
Synchronize(@ShowStatus);
FreeAndNil(Stream);
end;
if Terminated then
Status := 'Cancelled!'
else
Status := 'DONE!';
Synchronize(@ShowStatus);
end;
end.

74
lazarus-examples/generate_subtitles/unit1.lfm Normal file
View File

@@ -0,0 +1,74 @@
object Form1: TForm1
Left = 366
Height = 623
Top = 117
Width = 852
Caption = 'Next-gen Kaldi: Generate Subtitles'
ClientHeight = 623
ClientWidth = 852
OnClose = FormClose
OnCreate = FormCreate
LCLVersion = '3.4.0.0'
object FileNameEdt: TEdit
Left = 200
Height = 22
Top = 40
Width = 440
TabOrder = 0
OnChange = FileNameEdtChange
end
object SelectFileBtn: TButton
Left = 96
Height = 25
Top = 40
Width = 97
Caption = 'Select a file...'
TabOrder = 1
OnClick = SelectFileBtnClick
end
object StartBtn: TButton
Left = 656
Height = 25
Top = 37
Width = 75
Caption = 'Start'
TabOrder = 2
OnClick = StartBtnClick
end
object InitBtn: TButton
Left = 280
Height = 25
Top = 8
Width = 280
Caption = 'Click me to intialize models before you start'
TabOrder = 3
OnClick = InitBtnClick
end
object ResultMemo: TMemo
Left = 24
Height = 488
Top = 72
Width = 800
ScrollBars = ssAutoBoth
TabOrder = 4
end
object ProgressBar: TProgressBar
Left = 32
Height = 16
Top = 592
Width = 792
TabOrder = 5
end
object ProgressLabel: TLabel
Left = 770
Height = 16
Top = 568
Width = 8
Caption = '0'
end
object SelectFileDlg: TOpenDialog
Title = 'Open a wave file'
Left = 600
Top = 488
end
end

502
lazarus-examples/generate_subtitles/unit1.pas Normal file
View File

@@ -0,0 +1,502 @@
unit Unit1;
{$mode objfpc}{$H+}
{$IFDEF DARWIN}
{$modeswitch objectivec1} {For getting resource directory}
{$ENDIF}
interface
uses
Classes, SysUtils, StrUtils, Forms, Controls,
Graphics, Dialogs, StdCtrls,
sherpa_onnx, ComCtrls;
type
{ TForm1 }
TForm1 = class(TForm)
InitBtn: TButton;
ProgressBar: TProgressBar;
ResultMemo: TMemo;
StartBtn: TButton;
SelectFileDlg: TOpenDialog;
SelectFileBtn: TButton;
FileNameEdt: TEdit;
ProgressLabel: TLabel;
procedure FileNameEdtChange(Sender: TObject);
procedure FormClose(Sender: TObject; var CloseAction: TCloseAction);
procedure InitBtnClick(Sender: TObject);
procedure SelectFileBtnClick(Sender: TObject);
procedure FormCreate(Sender: TObject);
procedure StartBtnClick(Sender: TObject);
private
public
procedure UpdateResult(
Msg: AnsiString;
StartTime: Single;
StopTime: Single;
TotalDuration: Single);
procedure UpdateProgress(StopTime: Single; TotalDuration: Single);
public
Vad: TSherpaOnnxVoiceActivityDetector;
OfflineRecognizer: TSherpaOnnxOfflineRecognizer;
end;
var
Form1: TForm1;
implementation
uses
my_worker
{$IFDEF DARWIN}
,MacOSAll
,CocoaAll
{$ENDIF}
;
{See https://wiki.lazarus.freepascal.org/Locating_the_macOS_application_resources_directory}
{$IFDEF DARWIN}
{Note: The returned path contains a trailing /}
function GetResourcesPath(): AnsiString;
var
pathStr: shortstring;
status: Boolean = false;
begin
status := CFStringGetPascalString(CFStringRef(NSBundle.mainBundle.resourcePath), @pathStr, 255, CFStringGetSystemEncoding());
if status = true then
Result := pathStr + PathDelim
else
raise Exception.Create('Error in GetResourcesPath()');
end;
{$ENDIF}
function CreateVad(VadFilename: AnsiString): TSherpaOnnxVoiceActivityDetector;
var
Config: TSherpaOnnxVadModelConfig;
SampleRate: Integer;
WindowSize: Integer;
begin
Initialize(Config);
SampleRate := 16000; {Please don't change it unless you know the details}
WindowSize := 512; {Please don't change it unless you know the details}
Config.SileroVad.Model := VadFilename;
Config.SileroVad.MinSpeechDuration := 0.5;
Config.SileroVad.MinSilenceDuration := 0.5;
Config.SileroVad.Threshold := 0.5;
Config.SileroVad.WindowSize := WindowSize;
Config.NumThreads:= 2;
Config.Debug:= True;
Config.Provider:= 'cpu';
Config.SampleRate := SampleRate;
Result := TSherpaOnnxVoiceActivityDetector.Create(Config, 30);
end;
function CreateOfflineRecognizerTransducer(
Tokens: AnsiString;
Encoder: AnsiString;
Decoder: AnsiString;
Joiner: AnsiString;
ModelType: AnsiString): TSherpaOnnxOfflineRecognizer;
var
Config: TSherpaOnnxOfflineRecognizerConfig;
begin
Initialize(Config);
Config.ModelConfig.Transducer.Encoder := Encoder;
Config.ModelConfig.Transducer.Decoder := Decoder;
Config.ModelConfig.Transducer.Joiner := Joiner;
Config.ModelConfig.ModelType := ModelType;
Config.ModelConfig.Tokens := Tokens;
Config.ModelConfig.Provider := 'cpu';
Config.ModelConfig.NumThreads := 2;
Config.ModelConfig.Debug := False;
Result := TSherpaOnnxOfflineRecognizer.Create(Config);
end;
function CreateOfflineRecognizerTeleSpeech(
Tokens: AnsiString;
TeleSpeech: AnsiString): TSherpaOnnxOfflineRecognizer;
var
Config: TSherpaOnnxOfflineRecognizerConfig;
begin
Initialize(Config);
Config.ModelConfig.TeleSpeechCtc := TeleSpeech;
Config.ModelConfig.Tokens := Tokens;
Config.ModelConfig.Provider := 'cpu';
Config.ModelConfig.NumThreads := 2;
Config.ModelConfig.Debug := False;
Result := TSherpaOnnxOfflineRecognizer.Create(Config);
end;
function CreateOfflineRecognizerParaformer(
Tokens: AnsiString;
Paraformer: AnsiString): TSherpaOnnxOfflineRecognizer;
var
Config: TSherpaOnnxOfflineRecognizerConfig;
begin
Initialize(Config);
Config.ModelConfig.Paraformer.Model := Paraformer;
Config.ModelConfig.Tokens := Tokens;
Config.ModelConfig.Provider := 'cpu';
Config.ModelConfig.NumThreads := 2;
Config.ModelConfig.Debug := False;
Result := TSherpaOnnxOfflineRecognizer.Create(Config);
end;
function CreateOfflineRecognizerSenseVoice(
Tokens: AnsiString;
SenseVoice: AnsiString): TSherpaOnnxOfflineRecognizer;
var
Config: TSherpaOnnxOfflineRecognizerConfig;
begin
Initialize(Config);
Config.ModelConfig.SenseVoice.Model := SenseVoice;
Config.ModelConfig.SenseVoice.Language := 'auto';
Config.ModelConfig.SenseVoice.UseItn := True;
Config.ModelConfig.Tokens := Tokens;
Config.ModelConfig.Provider := 'cpu';
Config.ModelConfig.NumThreads := 2;
Config.ModelConfig.Debug := False;
Result := TSherpaOnnxOfflineRecognizer.Create(Config);
end;
function CreateOfflineRecognizerWhisper(
Tokens: AnsiString;
WhisperEncoder: AnsiString;
WhisperDecoder: AnsiString): TSherpaOnnxOfflineRecognizer;
var
Config: TSherpaOnnxOfflineRecognizerConfig;
begin
Initialize(Config);
Config.ModelConfig.Whisper.Encoder := WhisperEncoder;
Config.ModelConfig.Whisper.Decoder := WhisperDecoder;
Config.ModelConfig.Tokens := Tokens;
Config.ModelConfig.Provider := 'cpu';
Config.ModelConfig.NumThreads := 2;
Config.ModelConfig.Debug := False;
Result := TSherpaOnnxOfflineRecognizer.Create(Config);
end;
{$R *.lfm}
{ TForm1 }
procedure TForm1.FormCreate(Sender: TObject);
begin
StartBtn.Enabled := False;
SelectFileDlg.Filter := 'All Files|*.wav';
FileNameEdt.Enabled := False;
SelectFileBtn.Enabled := False;
ResultMemo.Lines.Add('1. It supports only 1 channel, 16-bit, 16000Hz wav files');
ResultMemo.Lines.Add('2. There should be no Chinese characters in the file path.');
ProgressBar.Position := 0;
ProgressLabel.Caption := '';
end;
procedure TForm1.StartBtnClick(Sender: TObject);
begin
if StartBtn.Caption = 'Stop' then
begin
if (MyWorkerThread <> nil) and not MyWorkerThread.Finished then
MyWorkerThread.Terminate;
StartBtn.Caption := 'Start';
Exit;
end;
ResultMemo.Lines.Clear();
ResultMemo.Lines.Add('Start processing');
ProgressBar.Position := 0;
ProgressLabel.Caption := Format('%d%%', [ProgressBar.Position]);
MyWorkerThread := TMyWorkerThread.Create(False, FileNameEdt.Text);
StartBtn.Caption := 'Stop';
end;
procedure TForm1.SelectFileBtnClick(Sender: TObject);
begin
if SelectFileDlg.Execute then
begin
FileNameEdt.Text := SelectFileDlg.FileName;
end;
end;
procedure TForm1.FileNameEdtChange(Sender: TObject);
begin
if FileExists(FileNameEdt.Text) then
StartBtn.Enabled := True
else
StartBtn.Enabled := False;
end;
procedure TForm1.FormClose(Sender: TObject; var CloseAction: TCloseAction);
begin
if (MyWorkerThread <> nil) and not MyWorkerThread.Finished then
begin
MyWorkerThread.Terminate;
MyWorkerThread.WaitFor;
end;
FreeAndNil(Vad);
FreeAndNil(OfflineRecognizer);
end;
procedure TForm1.UpdateProgress(StopTime: Single; TotalDuration: Single);
var
Percent: Single;
begin
if (StopTime <> 0) and (TotalDuration <> 0) then
begin
Percent := StopTime / TotalDuration * 100;
ProgressBar.Position := Round(Percent);
ProgressLabel.Caption := Format('%d%%', [ProgressBar.Position]);
end;
end;
procedure TForm1.UpdateResult(
Msg: AnsiString;
StartTime: Single;
StopTime: Single;
TotalDuration: Single);
var
NewResult: AnsiString;
begin
UpdateProgress(StopTime, TotalDuration);
if (Msg = 'DONE!') or
(Msg = 'Cancelled!') or
EndsStr('16-bit encoded wave files', Msg) or
EndsStr('. Please select a new file', Msg) then
begin
Form1.StartBtn.Caption := 'Start';
NewResult := Msg;
end
else
begin
NewResult := Format('%.3f -- %.3f %s', [StartTime, StopTime, Msg]);
end;
if Msg = 'DONE!' then
begin
ProgressBar.Position := 100;
ProgressLabel.Caption := '100%';
end;
Form1.ResultMemo.Lines.Add(NewResult);
end;
procedure TForm1.InitBtnClick(Sender: TObject);
var
Msg: AnsiString;
ModelDir: AnsiString;
VadFilename: AnsiString;
Tokens: AnsiString;
WhisperEncoder: AnsiString;
WhisperDecoder: AnsiString;
SenseVoice: AnsiString;
Paraformer: AnsiString;
TeleSpeech: AnsiString;
TransducerEncoder: AnsiString; // from icefall
TransducerDecoder: AnsiString;
TransducerJoiner: AnsiString;
NeMoTransducerEncoder: AnsiString;
NeMoTransducerDecoder: AnsiString;
NeMoTransducerJoiner: AnsiString;
begin
{$IFDEF DARWIN}
ModelDir := GetResourcesPath;
{$ELSE}
ModelDir := './';
{$ENDIF}
VadFilename := ModelDir + 'silero_vad.onnx';
Tokens := ModelDir + 'tokens.txt';
{
Please refer to
https://k2-fsa.github.io/sherpa/onnx/pretrained_models/whisper/export-onnx.html#available-models
for a list of whisper models.
In the code, we use the normalized filename whisper-encoder.onnx, whisper-decoder.onnx, and tokens.txt
You need to rename the existing model files.
For instance, if you use sherpa-onnx-whisper-tiny.en, you have to do
mv tiny.en-tokens.txt tokens.txt
mv tiny.en-encoder.onnx whisper-encoder.onnx
mv tiny.en-decoder.onnx whisper-decoder.onnx
// or use the int8.onnx
mv tiny.en-encoder.int8.onnx whisper-encoder.onnx
mv tiny.en-decoder.int8.onnx whisper-decoder.onnx
}
WhisperEncoder := ModelDir + 'whisper-encoder.onnx';
WhisperDecoder := ModelDir + 'whisper-decoder.onnx';
{
Please refer to
https://k2-fsa.github.io/sherpa/onnx/sense-voice/pretrained.html#pre-trained-models
to download models for SenseVoice.
In the code, we use the normalized model name sense-voice.onnx. You have
to rename the downloaded model files.
For example, you need to use
mv model.onnx sense-voice.onnx
// or use the int8.onnx
mv model.int8.onnx sense-voice.onnx
}
SenseVoice := ModelDir + 'sense-voice.onnx';
{
Please refer to
https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-paraformer/index.html
to download paraformer models.
Note that you have to rename model.onnx or model.int8.onnx to paraformer.onnx.
An example is given below for the rename:
cp model.onnx paraformer.onnx
// or use int8.onnx
cp model.int8.onnx paraformer.onnx
}
Paraformer := ModelDir + 'paraformer.onnx';
{
please refer to
https://k2-fsa.github.io/sherpa/onnx/pretrained_models/telespeech/models.html
to download TeleSpeech models.
Note that you have to rename model files after downloading. The following
is an example
mv model.onnx telespeech.onnx
// or to use int8.onnx
mv model.int8.onnx telespeech.onnx
}
TeleSpeech := ModelDir + 'telespeech.onnx';
{
Please refer to
https://k2-fsa.github.io/sherpa/onnx/pretrained_models/index.html
to download an icefall offline transducer model. Note that you need to rename the
model files to transducer-encoder.onnx, transducer-decoder.onnx, and
transducer-joiner.onnx
}
TransducerEncoder := ModelDir + 'transducer-encoder.onnx';
TransducerDecoder := ModelDir + 'transducer-decoder.onnx';
TransducerJoiner := ModelDir + 'transducer-joiner.onnx';
{
Please visit
https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
to donwload a NeMo transducer model.
}
NeMoTransducerEncoder := ModelDir + 'nemo-transducer-encoder.onnx';
NeMoTransducerDecoder := ModelDir + 'nemo-transducer-decoder.onnx';
NeMoTransducerJoiner := ModelDir + 'nemo-transducer-joiner.onnx';
if not FileExists(VadFilename) then
begin
ShowMessage(VadFilename + ' does not exist! Please download it from' +
sLineBreak + 'https://github.com/k2-fsa/sherpa-onnx/tree/asr-models'
);
Exit;
end;
Self.Vad := CreateVad(VadFilename);
if not FileExists(Tokens) then
begin
ShowMessage(Tokens + ' not found. Please download a non-streaming ASR model first!');
Exit;
end;
if FileExists(WhisperEncoder) and FileExists(WhisperDecoder) then
begin
OfflineRecognizer := CreateOfflineRecognizerWhisper(Tokens, WhisperEncoder, WhisperDecoder);
Msg := 'Whisper';
end
else if FileExists(SenseVoice) then
begin
OfflineRecognizer := CreateOfflineRecognizerSenseVoice(Tokens, SenseVoice);
Msg := 'SenseVoice';
end
else if FileExists(Paraformer) then
begin
OfflineRecognizer := CreateOfflineRecognizerParaformer(Tokens, Paraformer);
Msg := 'Paraformer';
end
else if FileExists(TeleSpeech) then
begin
OfflineRecognizer := CreateOfflineRecognizerTeleSpeech(Tokens, TeleSpeech);
Msg := 'TeleSpeech';
end
else if FileExists(TransducerEncoder) and FileExists(TransducerDecoder) and FileExists(TransducerJoiner) then
begin
OfflineRecognizer := CreateOfflineRecognizerTransducer(Tokens,
TransducerEncoder, TransducerDecoder, TransducerJoiner, 'transducer');
Msg := 'Zipformer transducer';
end
else if FileExists(NeMoTransducerEncoder) and FileExists(NeMoTransducerDecoder) and FileExists(NeMoTransducerJoiner) then
begin
OfflineRecognizer := CreateOfflineRecognizerTransducer(Tokens,
NeMoTransducerEncoder, NeMoTransducerDecoder, NeMoTransducerJoiner, 'nemo_transducer');
Msg := 'NeMo transducer';
end
else
begin
ShowMessage('Please download at least one non-streaming speech recognition model first.');
Exit;
end;
MessageDlg('Congrat! The ' + Msg + ' model is initialized succesfully!', mtInformation, [mbOk], 0);
FileNameEdt.Enabled := True;
SelectFileBtn.Enabled := True;
InitBtn.Enabled := False;
end;
end.

1
pascal-api-examples/non-streaming-asr/run-nemo-ctc.sh
View File

@@ -31,6 +31,7 @@ if [ ! -f ./sherpa-onnx-nemo-fast-conformer-ctc-be-de-en-es-fr-hr-it-pl-ru-uk-20
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./nemo_ctc.pas

1
pascal-api-examples/non-streaming-asr/run-nemo-transducer.sh
View File

@@ -32,6 +32,7 @@ if [ ! -f ./sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-r
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./nemo_transducer.pas

1
pascal-api-examples/non-streaming-asr/run-paraformer-itn.sh
View File

@@ -40,6 +40,7 @@ if [ ! -f ./itn_zh_number.fst ]; then
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./paraformer_itn.pas

1
pascal-api-examples/non-streaming-asr/run-paraformer.sh
View File

@@ -32,6 +32,7 @@ if [ ! -f ./sherpa-onnx-paraformer-zh-2023-09-14/tokens.txt ]; then
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./paraformer.pas

1
pascal-api-examples/non-streaming-asr/run-sense-voice.sh
View File

@@ -31,6 +31,7 @@ if [ ! -f ./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/tokens.txt ]; the
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./sense_voice.pas

1
pascal-api-examples/non-streaming-asr/run-telespeech-ctc.sh
View File

@@ -32,6 +32,7 @@ if [ ! -f ./sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04/tokens.txt ]; then
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./telespeech_ctc.pas

1
pascal-api-examples/non-streaming-asr/run-whisper.sh
View File

@@ -32,6 +32,7 @@ if [ ! -f ./sherpa-onnx-whisper-tiny.en/tiny.en-tokens.txt ]; then
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./whisper.pas

1
pascal-api-examples/non-streaming-asr/run-zipformer-transducer.sh
View File

@@ -32,6 +32,7 @@ if [ ! -f ./sherpa-onnx-zipformer-gigaspeech-2023-12-12/tokens.txt ]; then
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./zipformer_transducer.pas

1
pascal-api-examples/read-wav/run.sh
View File

@@ -28,6 +28,7 @@ if [ ! -f ./lei-jun-test.wav ]; then
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./main.pas

1
pascal-api-examples/streaming-asr/run-nemo-transducer.sh
View File

@@ -31,6 +31,7 @@ if [ ! -f ./sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms/tokens.
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./nemo_transducer.pas

1
pascal-api-examples/streaming-asr/run-paraformer.sh
View File

@@ -32,6 +32,7 @@ if [ ! -f ./sherpa-onnx-streaming-paraformer-bilingual-zh-en/tokens.txt ]; then
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./paraformer.pas

1
pascal-api-examples/streaming-asr/run-zipformer-ctc-hlg.sh
View File

@@ -31,6 +31,7 @@ if [ ! -f ./sherpa-onnx-streaming-zipformer-ctc-small-2024-03-18/tokens.txt ]; t
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./zipformer_ctc_hlg.pas

1
pascal-api-examples/streaming-asr/run-zipformer-ctc.sh
View File

@@ -31,6 +31,7 @@ if [ ! -f ./sherpa-onnx-streaming-zipformer-ctc-small-2024-03-18/tokens.txt ]; t
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./zipformer_ctc.pas

1
pascal-api-examples/streaming-asr/run-zipformer-transducer.sh
View File

@@ -32,6 +32,7 @@ fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./zipformer_transducer.pas

1
pascal-api-examples/vad-with-non-streaming-asr/run-vad-with-sense-voice.sh
View File

@@ -38,6 +38,7 @@ if [ ! -f ./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/tokens.txt ]; the
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./vad_with_sense_voice.pas

1
pascal-api-examples/vad-with-non-streaming-asr/run-vad-with-whisper.sh
View File

@@ -39,6 +39,7 @@ if [ ! -f ./sherpa-onnx-whisper-tiny.en/tiny.en-tokens.txt ]; then
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./vad_with_whisper.pas

1
pascal-api-examples/vad/run-circular-buffer.sh
View File

@@ -24,6 +24,7 @@ if [[ ! -f ../../build/install/lib/libsherpa-onnx-c-api.dylib && ! -f ../../bui
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./circular_buffer.pas

1
pascal-api-examples/vad/run-remove-silence.sh
View File

@@ -32,6 +32,7 @@ if [ ! -f ./lei-jun-test.wav ]; then
fi
fpc \
-dSHERPA_ONNX_USE_SHARED_LIBS \
-Fu$SHERPA_ONNX_DIR/sherpa-onnx/pascal-api \
-Fl$SHERPA_ONNX_DIR/build/install/lib \
./remove_silence.pas

4
scripts/apk/generate-vad-asr-apk-script.py
View File

@@ -71,7 +71,7 @@ def get_models():
Model(
model_name="sherpa-onnx-paraformer-zh-2023-09-14",
idx=0,
lang="zh",
lang="zh_en",
short_name="paraformer",
rule_fsts="itn_zh_number.fst",
cmd="""
@@ -109,7 +109,7 @@ def get_models():
Model(
model_name="sherpa-onnx-paraformer-zh-small-2024-03-09",
idx=14,
lang="zh",
lang="zh_en",
short_name="small_paraformer",
rule_fsts="itn_zh_number.fst",
cmd="""

81
scripts/lazarus/build-generate-subtitles.sh.in Normal file
View File

@@ -0,0 +1,81 @@
#!/usr/bin/env bash
#
# It expects that there are 4 directories inside /tmp
#
# macos-x64
# macos-arm64
# linux-x64
# windows-x64
#
# Generated files are saved in /tmp/out/*.tar.bz2
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)
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
os_array=(
linux-x64
macos-x64
macos-arm64
windows-x64
)
for os in ${os_array[@]}; do
if [[ $os == macos-x64 || $os == macos-arm64 ]]; then
cp -v silero_vad.onnx /tmp/$os/Contents/Resources/
else
cp -v silero_vad.onnx /tmp/$os/
fi
done
{% for model in model_list %}
model_name={{ model.model_name }}
lang={{ model.lang }}
short_name={{ model.short_name }}
curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/${model_name}.tar.bz2
tar xvf ${model_name}.tar.bz2
rm ${model_name}.tar.bz2
{{ model.cmd }}
ls -lh $model_name
for os in ${os_array[@]}; do
dst=sherpa-onnx-$SHERPA_ONNX_VERSION-generate-subtitles-$os-$short_name-$lang
src=/tmp/$os
cp -a $src $dst
d=$dst
if [[ $os == macos-x64 || $os == macos-arm64 ]]; then
mv $dst $dst.app
dst=$dst.app
d=$dst/Contents/Resources
fi
cp -v $model_name/*.onnx $d
cp -v $model_name/tokens.txt $d
d=$dst
tar cjvf $d.tar.bz2 $d
ls -lh
mkdir -p /tmp/out
mv $d.tar.bz2 /tmp/out
ls -lh /tmp/out
done
rm -rf $model_name
{% endfor %}

176
scripts/lazarus/generate-subtitles.py Executable file
View File

@@ -0,0 +1,176 @@
#!/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
lang: str
short_name: str = ""
cmd: str = ""
def get_models():
models = [
Model(
model_name="sherpa-onnx-whisper-tiny.en",
lang="en",
short_name="whisper_tiny.en",
cmd="""
pushd $model_name
rm -fv tiny.en-encoder.onnx
rm -fv tiny.en-decoder.onnx
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
""",
),
Model(
model_name="sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17",
lang="zh_en_ko_ja_yue",
short_name="sense_voice",
cmd="""
pushd $model_name
rm -fv model.onnx
mv -v model.int8.onnx sense-voice.onnx
popd
""",
),
Model(
model_name="sherpa-onnx-paraformer-zh-2023-09-14",
lang="zh_en",
short_name="paraformer_2023_09_14",
cmd="""
pushd $model_name
rm -fv model.onnx
mv -v model.int8.onnx paraformer.onnx
popd
""",
),
Model(
model_name="sherpa-onnx-paraformer-zh-small-2024-03-09",
lang="zh_en",
short_name="paraformer_small_2024_03_09",
cmd="""
pushd $model_name
rm -fv model.onnx
mv -v model.int8.onnx paraformer.onnx
popd
""",
),
Model(
model_name="sherpa-onnx-zipformer-gigaspeech-2023-12-12",
lang="en",
short_name="zipformer_gigaspeech_2023_12_12",
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
rm -fv encoder-epoch-30-avg-1.onnx
rm -fv decoder-epoch-30-avg-1.int8.onnx
rm -fv joiner-epoch-30-avg-1.onnx
popd
""",
),
Model(
model_name="icefall-asr-zipformer-wenetspeech-20230615",
lang="zh",
short_name="zipformer_wenetspeech",
cmd="""
pushd $model_name
rm -rfv test_wavs
rm -fv README.md
mv -v data/lang_char/tokens.txt ./
rm -rfv data/lang_char
mv -v exp/encoder-epoch-12-avg-4.int8.onnx ./
mv -v exp/decoder-epoch-12-avg-4.onnx ./
mv -v exp/joiner-epoch-12-avg-4.int8.onnx ./
rm -rfv exp
mv -v encoder-epoch-12-avg-4.int8.onnx transducer-encoder.onnx
mv -v decoder-epoch-12-avg-4.onnx transducer-decoder.onnx
mv -v joiner-epoch-12-avg-4.int8.onnx transducer-joiner.onnx
ls -lh
popd
""",
),
]
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 = [
"./build-generate-subtitles.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()

2
sherpa-onnx/csrc/wave-reader.cc
View File

@@ -224,8 +224,6 @@ std::vector<float> ReadWaveImpl(std::istream &is, int32_t *sampling_rate,
// header.subchunk2_size contains the number of bytes in the data.
// As we assume each sample contains two bytes, so it is divided by 2 here
std::vector<int16_t> samples(header.subchunk2_size / 2);
SHERPA_ONNX_LOGE("%d samples, bytes: %d", (int)samples.size(),
header.subchunk2_size);
is.read(reinterpret_cast<char *>(samples.data()), header.subchunk2_size);
if (!is) {

58
sherpa-onnx/pascal-api/sherpa_onnx.pas
View File

@@ -309,19 +309,50 @@ uses
SysUtils;
const
{See https://www.freepascal.org/docs-html/prog/progap7.html}
{
See
- https://www.freepascal.org/docs-html/prog/progap7.html
- https://downloads.freepascal.org/fpc/docs-pdf/
- https://downloads.freepascal.org/fpc/docs-pdf/CinFreePascal.pdf
}
{$IFDEF WINDOWS}
SherpaOnnxLibName = 'sherpa-onnx-c-api.dll';
{$ENDIF}
{ For windows, we always use dynamic link. See
https://forum.lazarus.freepascal.org/index.php/topic,15712.msg84781.html#msg84781
We need to rebuild the static lib for windows using Mingw or cygwin
}
SherpaOnnxLibName = 'sherpa-onnx-c-api.dll';
{$ELSE}
{$IFNDEF SHERPA_ONNX_USE_SHARED_LIBS}
{static link for linux and macos}
{$linklib sherpa-onnx-c-api}
{$linklib sherpa-onnx-core}
{$linklib kaldi-decoder-core}
{$linklib sherpa-onnx-kaldifst-core}
{$linklib sherpa-onnx-fstfar}
{$linklib sherpa-onnx-fst}
{$linklib kaldi-native-fbank-core}
{$linklib piper_phonemize}
{$linklib espeak-ng}
{$linklib ucd}
{$linklib onnxruntime}
{$linklib ssentencepiece_core}
{$IFDEF DARWIN}
SherpaOnnxLibName = 'sherpa-onnx-c-api';
{$linklib sherpa-onnx-c-api}
{$ENDIF}
{$IFDEF LINUX}
{$linklib m}
{$LINKLIB stdc++}
{$LINKLIB gcc_s}
{$ENDIF}
{$IFDEF LINUX}
SherpaOnnxLibName = 'libsherpa-onnx-c-api.so';
{$IFDEF DARWIN}
{$linklib c++}
{$ENDIF}
SherpaOnnxLibName = '';
{$ELSE}
{dynamic link for linux and macos}
SherpaOnnxLibName = 'sherpa-onnx-c-api';
{$linklib sherpa-onnx-c-api}
{$ENDIF}
{$ENDIF}
type
@@ -621,10 +652,17 @@ var
PWave: PSherpaOnnxWave;
I: Integer;
begin
Result.Samples := nil;
Result.SampleRate := 0;
PFilename := PAnsiChar(Filename);
PWave := SherpaOnnxReadWaveWrapper(PFilename);
Result.Samples := nil;
if PWave = nil then
Exit;
SetLength(Result.Samples, PWave^.NumSamples);
Result.SampleRate := PWave^.SampleRate;