The voice

Confluence-Voice

Spatial WebRTC voice for any OpenSimulator grid. One Windows program, one config file, and residents hear each other where they stand. No Vivox, no Linux VM, no Docker.

View on GitHub Quick start

Why it exists

Vivox has stopped taking new sign-ups for its free OpenSim voice service and has announced its end. Grids need a replacement they control before it stops. The existing WebRTC backend is built for Linux, with Docker or WSL for everyone else. Confluence-Voice gives Windows grid operators a native option.

Free and self-hosted

It costs only your own hardware and bandwidth: no account, no per-user pricing, and no outside service in the voice path.

Windows-native

A single confluence-voice.exe configured by one .toml file beside it. Download, edit one file, run.

Any OpenSim grid

Works with any OpenSimulator that has the os-webrtc-janus add-on, which upstream OpenSim includes. Not tied to Confluence-OpenSim.

Truly spatial

A separate mix for each listener, so voices pan and fade with distance the way residents expect.

Viewer-ready

Firestorm 7.1.10 and later work with no setup. Browser viewers can use the included reference client.

Secure by default

TLS is required, the region port takes an IP allow-list, and a TURN relay can be offered as a last resort for viewers behind strict networks.

How a call flows

The os-webrtc-janus add-on handles the region side. Confluence-Voice is what you point it at instead of a Janus gateway.

ViewerFirestorm 7.1.10+ or a browser viewer

→voice caps
(LLSD)

OpenSim regionwith os-webrtc-janus (required)

→JSON-RPC
over TLS

Confluence-Voiceanswers the offer, mixes audio, sends UDP media to viewers

Quick start

Run the exe once. It writes confluence-voice.toml beside itself and exits. Set at least your public IP, then run it again.

public_ip = "203.0.113.10"
rpc_bind = "0.0.0.0:9443"
media_port_lo = 40000
media_port_hi = 40999
max_sessions = 900
tls_cert = "tls\\fullchain.pem"
tls_key = "tls\\privkey.pem"

What you need

Windows, with a public IP and a DNS name pointing at it
A TLS certificate (required, not optional)
UDP 40000–40999 open, which also caps concurrent sessions
TCP 9443 reachable from your region hosts only
The os-webrtc-janus add-on in your regions

Full steps, including starting it automatically, are in the README.

Status

Released as 0.1.2. A release claims only what has been tested and ticked off in the project's testing checklist.

Area	Status	Detail
Spatial voice	Confirmed	With real viewers and real people in different countries (New Zealand and Indiana, USA)
Live grid	In use	Casperia moved all 15 regions from Vivox to WebRTC-only in a single pass
TURN relay	Done	Offered to viewers as a relay of last resort
Group and person-to-person calls	Testing	Next on the checklist
Shared secret on the region port	Planned	Needs a matching change on the region side first

Best on a WebRTC-only grid. A grid that mixes WebRTC with Vivox regions works, but Firestorm can stop its Vivox client after visiting a WebRTC region. Confluence-OpenSim carries a region-side fix for that; running WebRTC everywhere avoids it entirely.

Confluence-Voice is derived from wolfvoice by Wolf Software Systems Ltd, which donated its protocol handling, spatial mixer and session logic. Licensed under Apache 2.0.