Talknado

English | Русский | 中文

Description

A lightweight, low-latency real-time communication application. Provides voice chat, text messaging, and hardware-accelerated screen sharing in a single executable with an embedded server - no separate server installation required.

<p align="center"> <img src="/docs/MainWindow-en.png" alt="Main window screenshot"> </p>

Key Features

• Voice chat with noise suppression (RNNoise) and packet loss concealment (Opus PLC)
• Screen sharing at 30 fps with hardware H.264 encoding (NVENC → AMF → libx264 fallback)
• CoreAudio audio stack (WASAPI/MMDevice) for low-latency capture and playback
• Adaptive audio playback with jitter estimation and dynamic buffer sizing
• Adaptive screen-share playback with per-frame delta timing and target-buffer logic
• UPnP automatic port forwarding via Open.NAT (optional, one click)
• End-to-end encryption: ECDH key exchange → AES-256-CBC session key per participant
• TCP + UDP dual-channel transport (LiteNetLib): control over TCP, media over UDP
• Transparent reconnect: 5-second grace window on connection loss, no re-authentication
• Embedded server - starts only when the user chooses to host; zero startup cost otherwise
• MVVM architecture with dynamic DI containers (Microsoft.Extensions.DependencyInjection)

System Components

Client (WPF .NET 8)

• UI with dark theme, Russian and Chinese localization
• WASAPI audio capture (microphone + system loopback) with RNNoise noise suppression
• Opus codec (VOIP mode, 480-sample / 10 ms frames, PLC on packet loss)
• H.264 encode/decode via FFmpeg (avcodec, swscale)
• Desktop Duplication API capture (SharpDX) with cursor overlay
• TCP/UDP network client with transparent reconnect

Embedded Server (DLL .NET 8)

• Session creation, password verification, user management
• MSG/CMD packet routing - server never decrypts message content
• Media (audio/video) relay between participants
• Per-session AES session key, distributed via XOR with each client's ECDH shared secret
• Screen-share slot arbitration (one presenter at a time)

Dedicated Linux Server

• Separate repository: talknado-server-linux
• Same session handling and media relay as the embedded server, self-contained linux-x64 binary
• Installation:

curl -fsSL https://raw.githubusercontent.com/SodiumCXI/talknado-server-linux/main/install.sh | bash

Security

The connection handshake runs in seven steps:

1. Version check - incompatible clients are rejected before any key material is shared
2. ECDH key exchange - client and server exchange public keys; each independently derives the shared secret
3. Password verification - SHA-256(password) encrypted with AES-256-CBC using the shared secret
4. Session key delivery - server XORs the 32-byte session key with the client's unique shared secret; XOR is self-inverse, so no extra decryption step is needed
5. Identity exchange - username and assigned user ID, both encrypted with the session key
6. UDP registration - encrypted userId sent over LiteNetLib; server confirms with #UCC
7. State sync - server delivers participant list and current screen-share state

From step 4 onward all traffic - control and media - is encrypted with the session key (AES-256-CBC).

How to Use Over the Internet

Option A - Automatic (UPnP)

When starting a server, check "Enable UPnP port forwarding". Talknado will attempt to map the port automatically via Open.NAT. The mapping is removed when the server stops. Requires a static public IP.

Option B - ZeroTier (always works)

1. Install ZeroTier on each device.
2. All participants join the same ZeroTier network (Network ID).
3. Start Talknado and enter the connection key from the host's application.