SetNet.Voice 1.1.0

<p align="center"> <img src="https://raw.githubusercontent.com/Povstalez/SetNet/master/assets/icon.png" alt="SetNet" width="96"> </p>

SetNet.Voice

Codec-agnostic voice-chat relay for SetNet.

A server-side relay for real-time voice. Clients join numeric channels and push opaque audio frames — the server never encodes or decodes anything; a frame is just bytes (an encoded Opus packet, raw PCM, Speex, whatever your capture produces). Each frame is fanned out to the channel's other members over the unreliable channel (loss-tolerant, low-latency, the right trade-off for voice). Every relayed frame carries a stable speaker id so the receiver can mix per-speaker. Added by composition — no base class.

Install

dotnet add package SetNet
dotnet add package SetNet.Voice

Usage

Call VoiceRuntime.Enable() once at startup on both ends (before handler discovery).

Server — just turn the relay on:

VoiceRuntime.Enable();
server.UseVoice();

Client — join a channel, push captured audio, play back what you receive:

VoiceRuntime.Enable();

var voice = client.UseVoice();
await voice.JoinChannel(1);

voice.FrameReceived += (speakerId, channel, audio) =>
{
    // decode `audio` (e.g. Opus → PCM) and play it, keyed by speakerId
    playback.Feed(speakerId, decoder.Decode(audio));
};

// from your microphone capture loop (~20 ms frames):
await voice.SendFrame(1, encoder.Encode(pcmFrame));

This package handles transport only. Capture, encode/decode (Opus etc.), jitter buffering, and mixing are yours — SetNet just moves the bytes.

API

Server: server.UseVoice() — enable the relay hub.

Client: var voice = client.UseVoice() → VoiceChannel

VoiceRuntime.Enable() — one-time bootstrap so the voice handlers are discovered.

Notes

• Reserved wire types 65503 / 65504 / 65505. Don't reuse them.
• Unreliable by design. Voice frames use DeliveryMethod.Unreliable; a UDP or Both transport gives the best latency. Over a reliable-only transport (plain TCP/WebSocket) they still deliver, but head-of-line blocking can add jitter.
• Bring your own MTU discipline. Keep encoded frames comfortably under the transport's datagram limit (SetNet's UDP default rejects payloads over ~1200 bytes); 20 ms Opus frames are well within this.
• No server-side mixing. The server forwards per-speaker frames; clients mix. This keeps the server cheap and codec-agnostic. Pair with SetNet.Rooms or SetNet.Party and map their ids to voice channels for lobby/party voice.
• No built-in auth. Anyone who can reach the server can join a channel — gate it with SetNet.Auth if channels are private.

Documentation & source

• 🐙 https://github.com/Povstalez/SetNet — full module catalog in docs/MODULES.md

License

MIT