OpenSimulator · EN

Install your own WebRTC voice server

Fresh independent install with TLS, ports, service and two keys. Do not apply blindly to an existing production host.

This documents one specific WebRTC stack. Respect compatible versions and consult linked sources before modifying a production deployment.

Official repositories and downloads

Three distinct components: voice server, simulator modules, and optional ad hoc conferences. Use versions compatible with your installation.

BloodMoon Voice — voice serverRelease v1.0.0 and installer
OpenSimulator WebRTC modulesRelease v0.9.3.1-webrtc.3, DLL/PDB ZIP
OpenSimulator AdHoc ConferencesPHP/MySQL source and installation guide

Prepare host and DNS

The repository guide targets Ubuntu 24.04 x86-64 with systemd. Prepare a publicly reachable IP for viewers and a DNS hostname resolving to the server. The TLS certificate must match that hostname. Configure NAT forwarding if necessary.

Understand ports before opening them

TCP 9443 handles simulator HTTPS signaling; UDP 40000–40999 transports viewer media. TCP 80 may be required for Let's Encrypt issuance or renewal depending on your setup. Do not publicly expose TCP 9443 until RPC-key enforcement is confirmed.

Example for your own installation: adapt all domains, paths and names.

sudo ufw status
sudo ufw allow 40000:40999/udp

Download the correct release installer

For a fresh installation, use the v1.0.0 installer archive, inspect setup.sh before execution, then follow prompts for IP and DNS. The installer handles the binary, service and TLS preparation, not keys, region modules or the conference API.

Example for your own installation: adapt all domains, paths and names.

mkdir -p "$HOME/bloodmoon-voice-install" && cd "$HOME/bloodmoon-voice-install"
curl -fL https://github.com/Vlad-BloodMoon/BloodMoon-Voice/releases/download/v1.0.0/BloodMoon-Voice-installer-v1.0.0.tar.gz -o installer.tar.gz
mkdir source && tar -xzf installer.tar.gz -C source
cd source
less setup.sh
sudo bash setup.sh

Create two keys without overwriting existing keys

A new independent deployment uses cred_key for voice credentials and rpc_key for simulator request authentication. The example preserves existing key files; do not rotate keys on an operational service.

Example for your own installation: adapt all domains, paths and names.

sudo bash <<'KEYS'
set -euo pipefail
install -d -o root -g wolfvoice -m 750 /etc/wolfvoice
for key in cred_key rpc_key; do
  path="/etc/wolfvoice/$key"
  if [ -e "$path" ]; then
    test -s "$path" || exit 1
  else
    umask 077
    openssl rand -hex 32 > "$path"
  fi
  chown root:wolfvoice "$path"
  chmod 640 "$path"
done
KEYS

Verify service health

Restart the service and inspect its status and logs. Before exposing 9443, verify using the repository's documented JSON-RPC/TLS test that unauthenticated requests are rejected.

Example for your own installation: adapt all domains, paths and names.

sudo systemctl restart wolfvoice.service
sudo systemctl status wolfvoice.service --no-pager -l
sudo journalctl -u wolfvoice.service -n 40 --no-pager

Open signaling after verification

Once the unauthenticated rejection test succeeds, allow TCP 9443 for intended simulator hosts while keeping key enforcement. UDP media ports must be reachable by viewers; adjust provider firewalls and NAT. Never enable UFW remotely without preserving SSH access.

Example for your own installation: adapt all domains, paths and names.

sudo ufw allow 9443/tcp
sudo ufw status

Do not reinstall a working production server

The guide explicitly warns against rerunning setup.sh just to update a live service. Back up binary, systemd unit, configuration, keys and certificates; use controlled replacement and a planned restart if necessary.

Further reading and sources