rustrtc

A high-performance, full-stack real-time communication library — WebRTC, RTP/SRTP, T.38 Fax, and UPnP NAT traversal — all through a unified PeerConnection API.

Features

• High performance — ~2.7× faster than webrtc-rs and ~2.4× faster than pion (Go) in throughput, with ~30% less memory than webrtc-rs (see the benchmark below).
• Full protocol stack — WebRTC, RTP, SRTP, and T.38 fax in a single library, plus UPnP IGD NAT traversal. Few moving parts, no missing pieces.
• Unified PeerConnection API — one interface for every transport mode (WebRtc ICE/DTLS/SRTP, Srtp, Rtp, and T.38). No fragmented APIs.
• WebRTC compliant — interoperable with Chrome/WebRTC and pion; offer/answer, renegotiation, and standard SDP attributes.
• Complete media pipeline — packetizer/depacketizer, jitter buffer, NACK/FIR/PLI, TWCC, and REMB for audio and video.
• Full ICE — STUN, TURN (UDP + TCP), ICE Lite, ICE TCP (RFC 6544), and single-port UDP mux for SFU/WHEP deployments.
• NAT traversal & deployment — RTP latching, UPnP IGD port mapping, and firewall-friendly port ranges (rtp_start_port/rtp_end_port).
• Production extras — RTP rewrite bridge (SSRC/PT/sequence remapping) and a WebRTC-compatible stats model.

Benchmark game (rustrtc vs webrtc-rs & pion) in 0.3.113

CPU: AMD Ryzen 7 5700X 8-Core Processor OS 5.15.0-118-generic #128-Ubuntu
Compiler rustc 1.97.1 (8bab26f4f 2026-07-14), go version go1.23.0 linux/amd64

nice@miuda.ai rustrtc % cargo run -r --example benchmark

Comparison (Baseline: webrtc)
Metric               | webrtc     | rustrtc    | pion      
--------------------------------------------------------------------------------
Duration (s)         | 10.05      | 10.05      | 10.08     
Setup Latency (ms)   | 0.92       | 0.20       | 0.90      
Throughput (MB/s)    | 285.64     | 758.67     | 310.14    
Msg Rate (msg/s)     | 292495.02  | 776880.40  | 317586.71 
CPU Usage (%)        | 1348.20    | 1341.00    | 1154.67   
Memory (MB)          | 33.00      | 23.00      | 44.00     
--------------------------------------------------------------------------------

Performance Charts
==================

Throughput (MB/s) (Higher is better)
webrtc     | ███████████████                          285.64
rustrtc    | ████████████████████████████████████████ 758.67
pion       | ████████████████                         310.14

Message Rate (msg/s) (Higher is better)
webrtc     | ███████████████                          292495.02
rustrtc    | ████████████████████████████████████████ 776880.40
pion       | ████████████████                         317586.71

Setup Latency (ms) (Lower is better)
webrtc     | ████████████████████████████████████████ 0.92
rustrtc    | █████████                                  0.20
pion       | ███████████████████████████████████████  0.90

CPU Usage (%) (Lower is better)
webrtc     | ████████████████████████████████████████ 1348.20
rustrtc    | ███████████████████████████████████████  1341.00
pion       | ██████████████████████████████████       1154.67

Memory (MB) (Lower is better)
webrtc     | ██████████████████████████████           33.00
rustrtc    | █████████████████████                    23.00
pion       | ████████████████████████████████████████ 44.00

Key Findings:

• Throughput: rustrtc is ~2.7x faster than webrtc-rs and ~2.4x faster than pion.
• Memory: rustrtc uses ~30% less memory than webrtc-rs and ~48% less than pion.
• Setup Latency: Significantly faster connection setup (0.20ms vs 0.92ms/0.90ms).

Usage

Here is a simple example of how to create a PeerConnection and handle an offer:

use rustrtc::{PeerConnection, RtcConfiguration, SessionDescription, SdpType};

#[tokio::main]
async fn main() {
    let config = RtcConfiguration::default();
    let pc = PeerConnection::new(config);

    // Create a Data Channel
    let dc = pc.create_data_channel("data", None).unwrap();

    // Handle received messages
    let dc_clone = dc.clone();
    tokio::spawn(async move {
        while let Some(event) = dc_clone.recv().await {
            if let rustrtc::DataChannelEvent::Message(data) = event {
                println!("Received: {:?}", String::from_utf8_lossy(&data));
            }
        }
    });

    // Create an offer
    let offer = pc.create_offer().unwrap();
    pc.set_local_description(offer).unwrap();

    // Wait for ICE gathering to complete
    pc.wait_for_gathering_complete().await;

    // Get the complete SDP with candidates
    let complete_offer = pc.local_description().unwrap();
    println!("Offer SDP: {}", complete_offer.to_sdp_string());
}

Configuration

All configuration goes through RtcConfiguration (or its builder RtcConfigurationBuilder):

Transport & Network

• transport_mode — TransportMode::WebRtc (default), TransportMode::Srtp, or TransportMode::Rtp.
• ice_servers — STUN/TURN server list.
• ice_transport_policy — All or Relay.
• rtp_start_port / rtp_end_port — Restrict RTP/ICE to a port range.
• external_ip — Override the external IP for ICE candidates (NAT scenarios).
• bind_ip — Bind to a specific local IP.
• disable_ipv6 — Disable IPv6 candidate gathering.
• enable_ice_lite — Enable ICE Lite mode.
• ice_tcp_policy — IceTcpPolicy::Disabled (default), IceTcpPolicy::Enabled, or IceTcpPolicy::PassiveOnly. Controls ICE TCP candidate support per RFC 6544.
• ice_udp_mux / ice_udp_mux_port — Share a single UDP socket across many PeerConnections (single-port multiplexing for SFU/WHEP). Set ice_udp_mux = true and ice_udp_mux_port = <port>; incoming packets are demuxed by the server ufrag in the STUN Binding Request, then by remote source address.

UPnP

• enable_upnp — Auto-map ports via UPnP IGD.
• upnp_lease_duration — UPnP port mapping lease duration in seconds (default: 3600).

RTP Latching

• enable_latching — Enable dynamic remote address detection for RTP-only mode.
• probation_max_packets — Number of packets to observe before committing a latched address.

Media Capabilities

• media_capabilities — Configure audio/video/image (T.38) codecs and SCTP port via MediaCapabilities.
• ssrc_start — Starting SSRC value for local tracks.

SCTP (Data Channels)

• sctp_rto_initial, sctp_rto_min, sctp_rto_max, sctp_max_association_retransmits, sctp_receive_window, sctp_heartbeat_interval, sctp_max_heartbeat_failures, sctp_max_burst, sctp_max_cwnd

RTP Buffer

• rtp_buffer_capacity — Per-SSRC receive buffer capacity.
• buffer_drop_strategy — DropNew or DropOldest when buffer is full.

use rustrtc::{
    PeerConnection, RtcConfiguration, RtcConfigurationBuilder,
    IceServer, TransportMode, config::T38Capability,
};

// Using builder
let config = RtcConfigurationBuilder::new()
    .transport_mode(TransportMode::Rtp)
    .enable_latching(true)
    .probation_max_packets(Some(5))
    .rtp_port_range(50000, 50100)
    .enable_upnp(true)
    .ice_tcp_policy(config::IceTcpPolicy::Enabled)
    .ice_server(IceServer::new(vec!["stun:stun.l.google.com:19302"]))
    .build();

let pc = PeerConnection::new(config);

// Direct field access
let mut config = RtcConfiguration::default();
config.transport_mode = TransportMode::WebRtc;
config.enable_latching = true;
config.rtp_start_port = Some(50000);
config.rtp_end_port = Some(50100);
config.enable_upnp = true;

Examples

You can run the examples provided in the repository.

SFU (Selective Forwarding Unit)

A multi-user video conferencing server. It receives media from each participant and forwards it to others.

1. Run the server:

   cargo run --example rustrtc_sfu

2. Open your browser and navigate to http://127.0.0.1:8081. Open multiple tabs/windows to simulate multiple users.

Echo Server

The echo server example demonstrates how to accept a WebRTC connection, receive data on a data channel, and echo it back. It also supports video playback if an IVF file is provided.

1. Run the server:

   cargo run --example echo_server

2. Open your browser and navigate to http://127.0.0.1:3000.

DataChannel Chat

A multi-user chat room using WebRTC DataChannels.

1. Run the server:

   cargo run --example datachannel_chat

2. Open your browser and navigate to http://127.0.0.1:3000. Open multiple tabs to chat between them.

Audio Saver

Records audio from the browser's microphone and saves it to a file (output.ulaw) on the server.

1. Run the server:

   cargo run --example audio_saver

2. Open your browser and navigate to http://127.0.0.1:3000. Click "Start" to begin recording.

RTP Play (FFmpeg)

Streams a video file (examples/static/output.ivf) via RTP to a UDP port, which can be played back using ffplay.

1. Run the server:

   cargo run --example rtp_play

2. In a separate terminal, run ffplay (requires ffmpeg installed):

   ffplay -protocol_whitelist file,udp,rtp -i examples/rtp_play.sdp

License

This project is licensed under the MIT License.