CrowdStream

Real-time live streaming infrastructure built for scale. CrowdStream enables low-latency broadcaster-to-audience sessions with live chat, reactions, co-broadcasting, and viewer presence — self-hosted, no third-party streaming dependency.

---

Features

• Live video and audio broadcasting over WebRTC
• Multi-viewer delivery via MediaSoup SFU (no peer-to-peer mesh)
• Real-time chat during live sessions
• Real-time emoji reactions
• Co-broadcasting (multiple active broadcasters in a single room)
• Live viewer count
• NAT traversal via Coturn (TURN/STUN)

---

Architecture

      Client
        |
        | WebSocket (Signaling) + WebRTC (Media)
        |
Node.js Signaling Server  <------->  MongoDB
        |                            (Room state, viewer sessions)
        |
MediaSoup SFU (Workers per CPU core)
        |
   [Broadcaster Producer]
        |
   [Viewer Consumers] 
        |
     Coturn
  (TURN/STUN Relay)

---

Tech Stack

Backend

• Node.js — Signaling server, WebSocket handling, REST API
• MediaSoup — WebRTC SFU. Spawns one worker process per CPU core. Media plane runs in C++ off the Node.js event loop.
• MongoDB — Room documents, viewer session tracking, broadcaster metadata
• Coturn — Self-hosted TURN/STUN server for ICE relay and NAT traversal
• FFmpeg — Media processing pipeline (image/thumbnail handling, future recording)

Infrastructure

• AWS EC2 — Compute. Required for static public IP (MediaSoup ICE candidates) and consistent CPU core allocation for MediaSoup workers.

---

Current Status

Broadcaster and viewer are functional on a local network. Cross-network WebRTC transport configuration (public IP announcement + Coturn relay path) is in progress.

---

Roadmap

• Cross-network WebRTC (public transport configuration)
• Authentication and session tokens
• Co-broadcast (multi-producer rooms)
• Chat and reaction system
• Viewer count persistence
• Stream recording pipeline
• CDN delivery for large audiences

---