ElBruno.Realtime 0.6.0

ElBruno.Realtime

A pluggable real-time audio conversation framework for .NET, following Microsoft.Extensions.AI patterns. Build voice-powered apps with local STT, TTS, VAD, and any LLM — all running on your machine, no cloud required.

Powered By

This project is built on two core Microsoft frameworks for AI and conversation management:

• Microsoft.Extensions.AI (MEAI) — Provides unified abstractions for chat clients (IChatClient) and speech-to-text (ISpeechToTextClient), enabling pluggable LLM and STT providers throughout the pipeline.
• Microsoft Agent Framework — Manages conversation sessions, per-user state, and dialogue continuity, ensuring each user gets a consistent, stateful conversation experience.

Together with industry-standard models (Whisper STT, Silero VAD, ONNX Runtime), these frameworks provide a production-ready foundation for real-time voice applications.

Architecture

    Microphone (Audio Input)
        │ raw PCM audio
        ▼
    🔇 Silero VAD ─── Voice Activity Detection (~2 MB ONNX)
        │ speech segments
        ▼
    🎙️ Whisper STT ─── Speech-to-Text (~75 MB GGML)
        │ transcribed text
        ▼
    🤖 Any IChatClient ─── LLM Chat (Ollama / OpenAI / Azure)
        │ response text
        ▼
    🗣️ Any TTS ─── Text-to-Speech (pluggable)
        │ WAV audio
        ▼
    Speaker (Audio Output)

All models download automatically on first use. The LLM is pluggable via IChatClient — use Ollama, OpenAI, Azure, or any provider.

Features

• Local-First — All audio processing runs locally. No data leaves your machine.
• Microsoft.Extensions.AI — Implements ISpeechToTextClient and follows M.E.AI patterns throughout
• Pluggable Providers — Swap STT, TTS, VAD, or LLM independently
• Auto Model Download — Models download from HuggingFace/Whisper.net on first use
• DI-Ready — One-line setup with AddPersonaPlexRealtime() + fluent builder
• Streaming — Full async streaming via IAsyncEnumerable for real-time processing
• Multi-Target — Supports .NET 8.0 and .NET 10.0

Quick Start

Install

dotnet add package ElBruno.Realtime          # Core abstractions + pipeline
dotnet add package ElBruno.Realtime.Whisper   # Local STT (Whisper.net)
dotnet add package ElBruno.Realtime.SileroVad # Local VAD (Silero)

Basic Usage

using ElBruno.Realtime;
using ElBruno.Realtime.Whisper;
using ElBruno.Realtime.SileroVad;
using Microsoft.Extensions.AI;

// 1. Configure the pipeline (models auto-download on first use)
builder.Services.AddPersonaPlexRealtime(opts =>
{
    opts.DefaultSystemPrompt = "You are a helpful voice assistant.";
})
.UseWhisperStt("whisper-tiny.en")   // 75MB, or "whisper-base.en" for accuracy
.UseSileroVad();                    // Voice activity detection
// .UseYourTts()  — plug in any ITextToSpeechClient

// 2. Register your LLM (any IChatClient provider)
builder.Services.AddChatClient(new OllamaChatClient(
    new Uri("http://localhost:11434"), "phi4-mini"));

// 3. Use the pipeline
var conversation = app.Services.GetRequiredService<IRealtimeConversationClient>();

// One-shot turn
using var audio = File.OpenRead("question.wav");
var turn = await conversation.ProcessTurnAsync(audio);
Console.WriteLine($"User: {turn.UserText}");
Console.WriteLine($"AI: {turn.ResponseText}");

// Streaming full-duplex
await foreach (var evt in conversation.ConverseAsync(microphoneStream))
{
    if (evt.Kind == ConversationEventKind.ResponseTextChunk)
        Console.Write(evt.ResponseText);
}

Packages

Samples

Run a Sample

# Prerequisites: Ollama running with phi4-mini
ollama pull phi4-mini
ollama serve

# Run the console sample
cd src/samples/scenario-01-console
dotnet run

Auto-Downloaded Models

All models are cached in %LOCALAPPDATA%/ElBruno/Realtime/ and shared across apps:

Documentation

Building from Source

git clone https://github.com/elbruno/ElBruno.Realtime.git
cd ElBruno.Realtime
dotnet build
dotnet test

Requirements

• .NET 8.0 or .NET 10.0 SDK
• ONNX Runtime compatible platform (Windows, Linux, macOS)
• Ollama (or any IChatClient provider) for the LLM
• Sufficient disk space for model files

Contributing

Contributions are welcome! Here's how to get started:

1. Fork the repository
2. Create a branch for your feature or fix: git checkout -b feature/my-feature
3. Make your changes and ensure the solution builds: dotnet build
4. Run tests: dotnet test
5. Submit a pull request with a clear description of the changes

Please open an issue first for major changes or new features to discuss the approach.

👋 About the Author

Hi! I'm ElBruno 🧡, a passionate developer and content creator exploring AI, .NET, and modern development practices.

Made with ❤️ by ElBruno

If you like this project, consider following my work across platforms:

• 📻 Podcast: No Tienen Nombre — Spanish-language episodes on AI, development, and tech culture
• 💻 Blog: ElBruno.com — Deep dives on embeddings, RAG, .NET, and local AI
• 📺 YouTube: youtube.com/elbruno — Demos, tutorials, and live coding
• 🔗 LinkedIn: @elbruno — Professional updates and insights
• 𝕏 Twitter: @elbruno — Quick tips, releases, and tech news

License

This project is licensed under the MIT License — see the LICENSE file for details.

Related Projects

• ElBruno.PersonaPlex — NVIDIA PersonaPlex-7B-v1 ONNX inference
• ElBruno.QwenTTS — QwenTTS text-to-speech
• ElBruno.VibeVoiceTTS — VibeVoiceTTS
• ElBruno.Text2Image — Text-to-image generation
• ElBruno.HuggingFace.Downloader — HuggingFace model downloader