Three live demos on the marketing site let visitors hear positional mix features in under a minute — a looping sine tone orbiting with TTS, panned clip playback, and proximity-based room voice. All three share one Voice+Data showcase project and the spatial-showcase agent template. Visitors pick a mode with a DataChannel join message; the browser never sends clip URLs.

Use stereo headphones for every spatial demo — panning is equal-power stereo, not HRTF binaural.

Try the live demos

Orbiting voice — continuous per-client sine tone plus TTS circle your head (setPlayPose, setTtsPose). Drag the marker to place the speaker; orbit cycling stops until you resume.
Spatial soundboard — drop clips in space with play({ volume, pose }).
Proximity room — mix groups, listener poses, and per-peer mute (setClientPose, setListenerMute).

What each demo exercises

Demo	Agent APIs	Visitor experience
Orbit	`setPlayPose`, `setTtsPose`, `speak`	Continuous sine (per-session frequency and start/stop), TTS with Say-it lock until `orbit_say_status` ended, drag-to-place marker; stage dot tracks speaker position.
Soundboard	`play` with `volume` and world `pose`; `setPlayPose` via `pad_move`	Tap a pad to preview near center; click the field to place the selected clip; drag stage markers to reposition; Left / Center / Right place in front and to the sides of the listener.
Proximity	`createMixGroup`, `setClientPose`, `setListenerMute`	Avatars move on a ~50 m stage (zoom to inspect nearby); voices get louder when closer.

First-frame spatial pose (agent 0.7.11+)

In @voicethere/agent 0.7.11 and later, the spatial-showcase template emits the first orbit_pose and proximity room_state before the greeting speak(). Browsers can apply spatial pose on the first TTS chunk instead of waiting for the second pose update.

Run it yourself

Dashboard: Create project → Template → Spatial audio showcase (spatial-showcase).
Set runner mode to Voice+Data and enable shared_child_per_session (required so mix-group members share one runner process). See Runner settings.
Deploy to cloud and connect a browser client with a client API key.
Send the first DataChannel message: { "type": "join", "demo": "orbit" | "soundboard" | "proximity" }. Optional assetOrigin tells the agent where to resolve showcase sound files (allowlisted origins only).

Template source: templates/spatial-showcase/

DataChannel protocol

The browser sends commands only — never HTTPS clip URLs. The agent resolves clipId to ${assetOrigin}/showcase/sounds/<file> after validating assetOrigin.

Direction	type	Purpose
Browser → agent	`join`	`demo` + optional `assetOrigin`
Browser → agent	`orbit`	`action: set \| say \| place` — radius, period, `frequencyHz` (110–1760), `sinePlaying`, text, or place `x`/`z`
Browser → agent	`pad`	`clipId`, `x`, `z`, optional `volume` (0–1) — world pose only
Browser → agent	`pad_move`	Reposition an existing play: `playId`, `x`, `z`
Browser → agent	`move`	Listener pose `x`, `z`, `yawDeg`
Browser → agent	`mute_peer`	`peerId`, `muted`
Agent → browser	`orbit_pose`	Speaker position for the stage canvas
Agent → browser	`orbit_say_status`	TTS playback `playing` or `ended` — Say it stays disabled until ended
Agent → browser	`room_state`	Peer list with poses and mute flags
Agent → browser	`showcase_ack`	Command result; pad acks include `playId`

Two volume layers

Per-clip — soundboard pads send volume (0–1) on pad; the agent passes it to play({ volume }). This scales that clip in the server mix.
Browser master — each showcase page has an agent volume slider backed by a Web Audio GainNode on the inbound track (persisted in localStorage["vt-showcase-volume"]). It affects TTS, clips, and remote mics equally — it does not change per-pad mix levels server-side.

Sourcing showcase sounds

The live soundboard catalog uses short WAV clips served from your site under /showcase/sounds/. When building your own project, you can source effects from:

Kenney.nl asset packs (UI Audio, Interface, Impact, Sci-Fi) — check each pack license.
Freesound — filter for CC0 and record attribution in your project docs.
Procedural or in-house WAV files you generate and host on HTTPS (same pattern as the showcase procedural chimes).

Always host clips on HTTPS URLs your runner can fetch; keep files small, document license and author in your repo, and never pass raw URLs from an untrusted browser — send ids or paths your agent validates.