Britney

Voice assistant project for our Software Engineering course in the Introduction to AI subject taught by Sir Bennett Tanyag.

Group members:

• Jacob Tarun @MarshmallowAndroid
• Ian Alvarez @proj-jra
• Marc Briones @Pluggedaxe
• Kyle Fabila @elykfabl

Building

Open the solution, then build GooeyWpf.

Running

This project makes use of the NVIDIA CUDA Toolkit for acceleration. Download it here for your version of Windows. Otherwise, CPU fallback will be used which is not ideal.

Whisper dependencies

Download a model of your choice here. Models with en will support only English. Otherwise, the model has translation support.

This project was tested on ggml-medium-q8_0.

Piper dependencies

Download the latest release of piper for Windows, amd64.

Additionally, this project requires the semaine voice from piper. You need to download two files:

• en_GB-semaine-medium.onnx
• en_GB-semaine-medium.onnx.json

To run the application, find the executable and ensure your directory structure matches the following:

<GooeyWpf.exe directory>
|   GooeyWpf.dll
|   GooeyWpf.exe
|   ...
|
+---Piper
|       piper.exe
|       onnxruntime.dll
|       ...
|
+---PiperVoice
|       en_GB-semaine-medium.onnx
|       en_GB-semaine-medium.onnx.json
|
\---WhisperCppModel
        ggml-xxxxxxxx-xxxx.bin

Credits

• whisper.net, C# bindings for whisper.cpp for voice transcription.
• piper for voice synthesis.