Add Custom Voice Commands to Your .NET C# App

Looking to integrate custom voice commands into an enterprise .NET C# application? Imagine users interacting naturally with commands like "Make me a large espresso" or "Set the temperature to 72 degrees". Adding voice command functionality to your C# application is simpler than ever with Rhino Speech-to-Intent.

With Rhino Speech-to-Intent, developers can train AI models with custom voice commands, allowing your app to understand and act on spoken language. These commands run entirely on-device, eliminating cloud dependencies, latency issues, and privacy concerns—all while delivering exceptional accuracy.

In this tutorial, we'll walk you through the process of integrating voice command functionality into your .NET application using Rhino Speech-to-Intent. You'll learn everything from creating a custom context to implementing the code.

Implement Voice Commands in a .NET App

Ensure your environment meets Rhino Speech-to-Intent's .NET requirements:

• Windows (x86_64): .NET Framework 4.6.1+, .NET Standard 2.0+, or .NET Core 3.0+
• macOS (x86_64): .NET Standard 2.0+ or .NET Core 3.0+
• macOS (arm64), Windows (arm64), Linux (x86_64), Raspberry Pi (3, 4, 5): .NET 6.0+

1. Get Your AccessKey

Sign up for a Picovoice Console account for free and obtain your AccessKey.

2. Train Custom Voice Command Models

Rhino Speech-to-Intent uses a context file to define what voice commands your app should recognize. Here's how to create one:

1. Go to the Rhino page on Picovoice Console.
2. Create a context (e.g. SmartHomeCommands)
3. Add intents (like TurnOnLight or SetThermostat)
4. Define phrases for each intent (e.g. "Set the thermostat to 72 degrees")

Once complete, download the context model file (.rhn). If your context is in a non-English language, download the corresponding language model.

3. Install the NuGet Package

Install the Rhino NuGet package:

dotnet add package Rhino

4. Initialize Rhino Speech-to-Intent

Initialize Rhino with your Picovoice AccessKey, your context file, and model file (if your context file is not in English):

using Pv;

Rhino rhino = Rhino.Create(
        "${ACCESS_KEY}", // Your Picovoice AccessKey
        "${CONTEXT_FILE_PATH}", // Your custom context file (.rhn)
        modelPath: "${MODEL_FILE_PATH}"); // Your model file (.pv) (for non-English contexts)

5. Start Listening For Voice Commands

Send audio frames to Rhino and handle the resulting inferences:

short[] GetNextAudioFrame()
{
    // ...
}

while (true)
{
    bool isFinalized = rhino.Process(GetNextAudioFrame());
    if (isFinalized)
    {
        Inference inference = rhino.GetInference();
        if (inference.IsUnderstood)
        {
            string intent = inference.Intent;
            Dictionary<string, string> slots = inference.Slots;
            // take action based on inferred intent and slot values
        }
    }
}

Whenever speech is detected, Rhino returns an Inference object containing the detected intent and slots values.

6. Clean Up

Once you no longer need Rhino, call Dispose() to free acquired memory:

rhino.Dispose();

Complete Demo: Voice Commands in .NET

Putting it all together, here's a fully functional console demo in C#. PvRecorder is used for audio capture:

using System;
using System.Threading;
using Pv;

class RhinoDemo
{
    static void Main()
    {
        Console.WriteLine("Initializing Rhino & PvRecorder...");
        Rhino rhino = Rhino.Create(
                "${ACCESS_KEY}", // Your Picovoice AccessKey
                "${CONTEXT_FILE_PATH}"); // Your custom context file (.rhn)
        PvRecorder recorder = PvRecorder.Create(frameLength: rhino.FrameLength);

        var processingThread = new Thread(() =>
        {
            while (recorder.IsRecording)
            {
                short[] audioFrame = recorder.Read();
                bool isFinalized = rhino.Process(audioFrame);
                if (isFinalized)
                {
                    Inference inference = rhino.GetInference();
                    if (inference.IsUnderstood)
                    {
                        string intent = inference.Intent;
                        Dictionary<string, string> slots = inference.Slots;
                        Console.WriteLine($"Intent: {intent}");
                        foreach (var kvp in slots)
                        {
                            Console.WriteLine($"{kvp.Key} > {kvp.Value}");
                        }
                    }
                    else
                    {
                        Console.WriteLine("Intent not understood.");
                    }
                }
            }
        });

        recorder.Start();
        processingThread.Start();

        Console.WriteLine("Listening for commands... Press Enter to stop.");
        Console.ReadLine();
        recorder.Stop();
        processingThread.Join();

        rhino.Dispose();
        recorder.Dispose();

        Console.WriteLine("Stopped and resources released.");
    }
}

For a complete .NET application, see the Rhino .NET demo on GitHub.

This tutorial uses the following package:

• Rhino Speech-to-Intent NuGet Package
• PvRecorder NuGet Package

Explore our documentation for more details:

• Rhino Speech-to-Intent .NET Quick Start
• Rhino Speech-to-Intent .NET API

Best Practices for Voice Command Systems

• Audio format: Record audio in the format expected by your voice command engine. Rhino Speech-to-Intent requires mono, 16-bit PCM audio, with the sample rate set to rhino.SampleRate and frame length set to rhino.FrameLength.
• Intent clarity: Ensure the phrases in your custom context closely match what users are likely to say, so voice commands are accurately understood.
• Threading: Run audio capture and processing on a separate thread to keep your UI or main application logic responsive.

Next Steps

Enhance your voice-controlled app by combining Rhino Speech-to-Intent with:

• Porcupine Wake Word: start listening for voice commands only after a custom wake word is spoken.
• Cheetah Streaming Speech-to-Text: transcribe spoken content after a command, e.g., "Transcribe this meeting for me".