ezspeak

TeamSpeak-like voice chat in the browser. No downloads, no accounts. Just open a link and talk.

Quick Start

npm install
npm start
# → http://localhost:3000

Share the link: http://localhost:3000/server/my-crew

Docker Compose

docker compose up --build

Exposes port 3111.

Features

• WebRTC full mesh — every peer connects directly to every other peer in a channel. Browser natively mixes multiple audio streams.
• Channels — default Lobby, Channel 1, Channel 2; creator can add more
• Password-protected servers — optional on creation, required on join
• Text chat — per-channel messages with join/leave system notices
• Speaking indicators — avatars glow green when audio is detected (AnalyserNode threshold)
• Per-user volume & mute — volume slider (0–100%) and mute button per peer. Exponential curve for natural volume control.
• Mute/Deafen status broadcast — your mute 🤐 and deafen 🙉 state is visible to everyone in the channel. Persisted on the server (survives reloads).
• Sound effects — synthesized beeps for join, leave, and channel switch
• Persistent identity — username saved to localStorage (default: ezhuman-chrome-4821)
• Recent servers — modal with last 10 servers, accessible from all views
• In-server rename — change your display name live, everyone sees it
• Creator tools — add channels, set/remove server password
• Mobile-optimized — slide-out sidebar, 44px touch targets, WakeLock API
• Autoplay recovery — per-audio-element retry on user interaction

Architecture

Browser A ←──WebRTC mesh──→ Browser B
    │                           │
    └────── WebSocket ──────────┘
                 │
          signaling server
           (Node.js / Express / ws)

Server (server.js)

Directory	Concern
server.js	HTTP server, Express static files, WebSocket upgrade
server/state.js	In-memory state: servers, channels, users, broadcasting
server/handlers.js	Per-message-type handler functions

State is purely in-memory. No database. Servers persist as long as at least one user is connected.

Client (public/)

File	Concern
index.html	Page structure, modals, control bar
style.css	Dark theme, layout, animations
js/config.js	Constants (RTC config, localStorage keys)
js/storage.js	localStorage helpers (username, recent servers)
js/audio.js	Audio stream, mute/deafen, speaking detection, beeps
js/net.js	WebSocket signaling connection, keep-alive
js/webrtc.js	WebRTC full-mesh peer connections, per-user volume/mute
js/chat.js	Chat message rendering
js/ui.js	DOM rendering, channel list, user list with controls, modals
js/app.js	Entry point, state, routing, signaling dispatch, event wiring

WebSocket Protocol

All messages are JSON.

Client → Server

Type	Fields	Notes
join-server	serverName, username, password?	Creates or joins server
join-channel	channelName	Auto-leaves current
leave-channel	—	Closes all WebRTC
add-channel	channelName	Creator only
change-username	username	Broadcasts to all
set-password	password	Creator only, empty = remove
chat-message	message	Broadcasts to channel
mute-state-changed	value	Self mute toggle (boolean)
deafen-state-changed	value	Self deafen toggle (boolean)
webrtc-offer	targetId, offer	SDP offer
webrtc-answer	targetId, answer	SDP answer
webrtc-ice-candidate	targetId, candidate	ICE candidate

Server → Client

Type	Fields	Notes
server-state	serverName, creator, hasPassword, channels, users, yourUserId	Full state on join. Users include isMuted/isDeafened.
password-required	serverName	Server needs a password
user-joined-server	userId, username	Another user joined
user-left-server	userId	Another user left
user-renamed	userId, oldUsername, newUsername	Name change
password-updated	hasPassword	Password set/removed
user-channel-update	userId, username, channelName	Moved channels
channel-added	channelKey, channelName	New channel
joined-channel	channelName, existingPeers, peerDetails, totalUsers	You joined. peerDetails includes mute/deafen state.
left-channel	channelName	You left
peer-joined-channel	userId, username, channelName, totalUsers, isMuted, isDeafened	Another joined your channel
peer-left-channel	userId, username, channelName	Another left your channel
mute-state-changed	userId, username, value	Peer mute state changed
deafen-state-changed	userId, username, value	Peer deafen state changed
chat-message	userId, username, message, timestamp	Chat message
webrtc-offer	fromId, fromUsername, offer	Relayed offer
webrtc-answer	fromId, fromUsername, answer	Relayed answer
webrtc-ice-candidate	fromId, fromUsername, candidate	Relayed ICE
error	message	Error feedback

WebRTC Mesh Topology

When a user joins a channel:

1. Server notifies all existing peers via peer-joined-channel → each creates an RTCPeerConnection + SDP offer
2. New peer receives offers via webrtc-offer → creates answers
3. Bidirectional audio flows after ICE completes

With N users in a channel, each user has N−1 peer connections. All audio streams are played through <audio> elements and mixed natively by the browser.

Volume & Mute

Per-peer volume uses audio.volume with an exponential slider curve (Math.pow(x, 1.5)) for natural perceptual range. Per-peer mute sets audio.muted. Global deafen overrides both.

Per-user Controls

Each remote user in the channel list has:

• Volume slider (0–100%, exponential curve)
• Mute button (🔊/🔇) — stops audio from that specific user

Self mute (🤐) and deafen (🙉) status is broadcast to all channel members and persisted on the server so it survives page reloads.