Readme

The WhisperPS.csproj project in this folder builds PowerShell wrapper for Whisper.

I wouldn’t call the wrapper particularly great, but it works on my computer.

This should handle use cases like “transcribe all files in a directory” or “export multiple formats”.

The supported PowerShell version is 5.1, the one I have preinstalled on my Windows 10 computer.

I wouldn’t expect it to work with the newer PowerShell Core, the runtime is different.

Installation

To install from PowerShell Gallery for the current user, open Windows PowerShell, and run the following command:

Install-Module -Name WhisperPS -Scope CurrentUser

To install from PowerShell Gallery for all users, open Windows PowerShell as an administrator, and run the following command:

Install-Module -Name WhisperPS

To install manually from github.com, download WhisperPS.zip from Releases page of this repository, and extract into the following folder:

%USERPROFILE%\Documents\WindowsPowerShell\Modules

Create that folder if you don’t yet have it.

Models

The binary files with the models are available for free download on Hugging Face.

I recommend ggml-medium.bin (1.42GB in size, but that web page says 1.53 GB), because I’ve mostly tested the software with that model.

Compressed models in ZIP format with mlmodelc in the file name are not supported.

Usage Example

Import-Module WhisperPS -DisableNameChecking
$Model = Import-WhisperModel D:\Data\Whisper\ggml-medium.bin
cd C:\Temp\2remove\Whisper
$Results = dir .\* -include *.wma, *.wav | Transcribe-File $Model
foreach ( $i in $Results ) { $txt = $i.SourceName + ".txt"; $i | Export-Text $txt; }
foreach ( $i in $Results ) { $txt = $i.SourceName + ".ts.txt"; $i | Export-Text $txt -timestamps; }

Commands

Here’s the list of commands implemented by this module.

• Get-Adapters prints names of the graphics adapters visible to DirectCompute.

You can use these strings for the -adapter argument of the Import-WhisperModel command.

• Import-WhisperModel loads the model from disk, returns the object which keeps the model
• Transcribe-File loads audio file from disk, and transcribes the audio into text. It returns the object which keeps both source file name, and transcribed text.
• Export-Text saves transcribed text into a text file, with or without timestamps.
• Export-SubRip saves transcribed text in *.srt format.
• Export-WebVTT saves transcribed text in *.vtt format.

You can use man (an alias for get-help) for detailed documentation on specific commands, example:

man Import-WhisperModel -full

Miscellaneous

By default, PowerShell doesn’t print any informational or debug messages.

If you want these messages, run these commands:

$InformationPreference="Continue"
$DebugPreference="Continue"

Apparently, the default value for these preference variables is SilentlyContinue so by default the messages go nowhere.

Unfortunately, PowerShell 5.1 uses legacy .NET framework 4.8.

That’s why I couldn’t simply consume WhisperNet library in this project.

Instead I implemented slightly different C# wrappers for the same C++ DLL.

Luckily, my ComLightInterop library is old, and it still supports legacy .NET framework, along with other old versions of the runtime like .NET Core 2.1.