diff --git a/Voice-and-Video.md b/Voice-and-Video.md
new file mode 100644
index 0000000..0e154a9
--- /dev/null
+++ b/Voice-and-Video.md
@@ -0,0 +1,125 @@
+# Voice & Video
+
+Dumpster uses LiveKit for real-time voice and video communication. LiveKit is an open-source WebRTC SFU (Selective Forwarding Unit) that handles media routing between participants.
+
+## Architecture
+
+```
+Client ──(WS)──> Dumpster API ──(token)──> Client
+Client ──(WebRTC)──> LiveKit SFU ──(WebRTC)──> Other Clients
+```
+
+- **Dumpster backend** generates room tokens and enforces permissions
+- **LiveKit** handles all media (audio, video, screen share)
+- **coturn** provides TURN relay for clients behind NAT
+
+## Setup
+
+### 1. Generate LiveKit API Keys
+
+```bash
+# Generate a random API key and secret
+echo "API_KEY: $(openssl rand -hex 12)"
+echo "API_SECRET: $(openssl rand -hex 24)"
+```
+
+Add to `.env`:
+```
+LIVEKIT_API_KEY=<generated_key>
+LIVEKIT_API_SECRET=<generated_secret>
+LIVEKIT_URL=wss://livekit.your.domain
+```
+
+### 2. Start Services
+
+```bash
+docker compose -f docker/compose.yml up -d livekit coturn
+```
+
+### 3. Verify
+
+```bash
+# Check LiveKit is running
+curl http://localhost:7880
+
+# Check coturn is running
+docker logs dumpster-coturn
+```
+
+## API Endpoints
+
+### Get Voice Token
+
+```
+POST /api/v1/voice/token
+Authorization: session cookie
+
+{
+  "room_name": "voice-channel-id"
+}
+
+Response:
+{
+  "token": "<jwt_token>",
+  "livekit_url": "wss://livekit.your.domain"
+}
+```
+
+### List Participants
+
+```
+GET /api/v1/voice/rooms/{roomID}/participants
+Authorization: session cookie
+
+Response:
+[
+  { "identity": "username", "state": "connected" },
+  { "identity": "username2", "state": "disconnected" }
+]
+```
+
+## WebSocket Events
+
+| Event | Data | Description |
+|-------|------|-------------|
+| `VOICE_JOIN` | `{ room_name, user_id, username }` | User joined voice channel |
+| `VOICE_LEAVE` | `{ room_name, user_id, username }` | User left voice channel |
+| `VOICE_MUTE` | `{ room_name, user_id, muted }` | User muted/unmuted |
+| `VOICE_DEAFEN` | `{ room_name, user_id, deafened }` | User deafened/undeafened |
+
+## Frontend Components
+
+| Component | Location | Description |
+|-----------|----------|-------------|
+| `VoiceChannel.tsx` | Sidebar | `🔊 channel-name` with participant count |
+| `VoicePanel.tsx` | Chat area | Box-drawn frame with participant list |
+| `VoiceControls.tsx` | VoicePanel | `[m]` mute, `[d]` deafen, `[cam]` video, `[share]` screen, `[x]` disconnect |
+| `stores/voice.ts` | Stores | Zustand store for voice state management |
+
+## NAT Traversal
+
+coturn runs as a TURN relay for clients behind restrictive NAT (carrier NAT, corporate firewalls). LiveKit is configured to use coturn automatically.
+
+**Ports required:**
+- 3478/tcp+udp (TURN)
+- 5349/tcp+udp (TURNS)
+- 10000-20000/udp (TURN relay range)
+- 7880-7882 (LiveKit)
+- 50000-50100/udp (LiveKit relay)
+
+## Troubleshooting
+
+### Can't connect to voice channel
+1. Check LiveKit is running: `docker logs dumpster-livekit`
+2. Check API keys match in `.env` and `docker/livekit.yaml`
+3. Check browser console for WebRTC errors
+
+### No audio
+1. Check browser permissions for microphone
+2. Check device selection in settings
+3. Try deafen/undeafen toggle
+
+### NAT issues
+1. Check coturn is running: `docker logs dumpster-coturn`
+2. Verify TURN ports are open (3478, 5349, 10000-20000)
+3. Check `DUMPSTER_HOST` is set to your public domain
diff --git a/_Sidebar.md b/_Sidebar.md
index 4d2cba3..f5ede5b 100644
--- a/_Sidebar.md
+++ b/_Sidebar.md
@@ -5,6 +5,7 @@
 - [Architecture](Architecture)
 - [API Reference](API-Reference)
 - [Profiles](Profiles)
+- [Voice & Video](Voice-and-Video)
 - [Frontend Design](Frontend-Design)
 - [Self-Hosting](Self-Hosting)
 - [Roadmap](Roadmap)