Neural Speech Engine with NVIDIA NeMo and ONNX Runtime screenshot 1
K

Publisher

kaiida

Neural Speech Engine with NVIDIA NeMo and ONNX Runtime

Game Kits
Speech Recognition Synthesis VAD C# NVIDIA NeMo ONNX

A sample Godot project of using NeMoOnnxSharp. NeMoOnnxSharp is a neural speech engine which is based on NVIDIA NeMo and ONNX Runtime written purely in C#. This project supports VAD (voice activation detection), speech recognition and speech synthesis.

Download Latest (0.1)
Source Code

NeMoOnnxGodot

A sample Godot project of using NeMoOnnxSharp. This project supports VAD (voice activation detection), speech recognition and speech synthesis. NeMoOnnxSharp is a inference library written in C# for NVIDIA NeMo with ONNX Runtime.

screen shot

Requirement

  • Visual Studio 2022
  • Godot 4.1.2
  • Windows 11 (or versions for which Godot supports)

Building

  • Open the directory with Godot Editor
  • Open the solution with Visual Studio 2022
  • From menu Build -> Build Solution
  • Then run with debugger (F5) from Godot Editor or export to Windows exe.

Running

English and German are supported.

The applicatin needs model files downloaded from GitHub. Click "Download models" button to download them.

Click Transcribe button to start VAD and speech recognition using a microphone. Click Speak to generate audio from the text.

Importing to your project

NeMoOnnxSharp needs ONNX Runtime NuGet package. Add these to your .csproj file.

Script.cs is the main script to use. Attach it to a Godot node in the inspector.

screen shot