MythEclipse/dc-recorder
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

refactor: implement robust Opus decoding with error-resilient streams and transition to pull-based audio transmission for Discord output

Browse Source
This commit is contained in:
baharsah
2026-05-13 02:58:11 +07:00
parent ad7dcde47c
commit aa85dd9beb
3 changed files with 114 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/webserver.ts
+81 -26
View File
@@ -8,20 +8,31 @@ import { discordPlayer } from "./player";
const activeUsers = new Map<string, { username: string, avatar: string, speaking: boolean }>();
let wsClients = new Set<any>();
// --- Upsampling: 24kHz mono s16le → 48kHz stereo s16le (pure JS, no FFmpeg) ---
// Each input sample is duplicated into 2 stereo pairs to double the sample rate.
function upsample24kMonoTo48kStereo(mono24k: Buffer): Buffer {
const out = Buffer.alloc(mono24k.length * 4); // 2x rate * 2ch = 4x bytes
// Upsample 24kHz mono s16le → 48kHz stereo s16le (pure JS)
function upsample(mono24k: Buffer): Buffer {
const out = Buffer.alloc(mono24k.length * 4);
for (let i = 0; i < mono24k.length / 2; i++) {
const s = mono24k.readInt16LE(i * 2);
out.writeInt16LE(s, i * 8); // t=0 L
out.writeInt16LE(s, i * 8 + 2); // t=0 R
out.writeInt16LE(s, i * 8 + 4); // t=1 L (duplicate for 2x rate)
out.writeInt16LE(s, i * 8 + 6); // t=1 R
out.writeInt16LE(s, i * 8);
out.writeInt16LE(s, i * 8 + 2);
out.writeInt16LE(s, i * 8 + 4);
out.writeInt16LE(s, i * 8 + 6);
}
return out;
}
// Calculate RMS dB level of a PCM s16le buffer
function rmsDb(pcm: Buffer): number {
let sum = 0;
const samples = pcm.length / 2;
for (let i = 0; i < samples; i++) {
const s = pcm.readInt16LE(i * 2) / 32768;
sum += s * s;
}
const rms = Math.sqrt(sum / samples);
return 20 * Math.log10(Math.max(rms, 1e-10));
}
export function startWebserver(port: number = 3000) {
const app = express();
const server = http.createServer(app);
@@ -32,7 +43,7 @@ export function startWebserver(port: number = 3000) {
app.use(express.static(path.join(__dirname, "../public")));
// --- Inbound: Discord PCM → tagged chunks → browser (set in recorder.ts) ---
// Inbound: Discord PCM → tagged chunks → browser
(global as any).broadcastPcmToWeb = (chunk: Buffer, userId: string) => {
let hash = 0;
for (let i = 0; i < userId.length; i++) {
@@ -62,33 +73,80 @@ export function startWebserver(port: number = 3000) {
});
}
// --- Outbound: browser PCM (24kHz mono) → Opus → Discord, NO FFmpeg ---
// --- Outbound: browser PCM (24kHz mono) → Opus → Discord ---
const RATE = 48000;
const CHANNELS = 2;
const FRAME_SIZE = 960; // 20ms @ 48kHz
const BYTES_PER_FRAME = FRAME_SIZE * CHANNELS * 2; // 3840 bytes
const FRAME_SIZE = 960;
const BYTES_PER_FRAME = FRAME_SIZE * CHANNELS * 2; // 3840 bytes = 20ms
const SILENCE_TAIL_MS = 300; // continue sending silence for 300ms after browser stops
const MAX_BUF_BYTES = BYTES_PER_FRAME * 50; // cap at 1 second to avoid runaway buffer
const opusEncoder = new prism.opus.Encoder({ rate: RATE, channels: CHANNELS, frameSize: FRAME_SIZE });
const oggBitstream = new prism.opus.OggLogicalBitstream({
opusHead: new prism.opus.OpusHead({ channelCount: CHANNELS, sampleRate: RATE }),
pageSizeControl: { maxPackets: 10 },
pageSizeControl: { maxPackets: 1 }, // 1 packet per page = 20ms latency
crc: true,
});
opusEncoder.on('error', () => {});
opusEncoder.pipe(oggBitstream);
// Prime the encoder immediately so OGG headers are emitted before player reads
// Prime OGG headers before player starts reading
opusEncoder.write(Buffer.alloc(BYTES_PER_FRAME, 0));
discordPlayer.playStream(oggBitstream);
discordPlayer.pause();
let pcmBuffer = Buffer.alloc(0);
let lastBrowserAudioTime = 0;
let playerPaused = true;
const SILENCE_FRAME = Buffer.alloc(BYTES_PER_FRAME, 0);
// Keep encoder alive with silence when browser isn't sending
// Log level every 2 seconds
let dbAccum = 0, dbCount = 0;
setInterval(() => {
if (Date.now() - lastBrowserAudioTime > 40) {
opusEncoder.write(SILENCE_FRAME);
if (dbCount > 0) {
const avg = dbAccum / dbCount;
console.log(`[transmit] Audio level: ${avg.toFixed(1)} dBFS (${dbCount} frames/2s)`);
dbAccum = 0; dbCount = 0;
}
}, 2000);
// PULL-BASED encode loop: fires every 20ms, pulls exactly one frame from buffer.
// This avoids the timing conflict where browser bursts and silence timer collide.
setInterval(() => {
const msSinceAudio = Date.now() - lastBrowserAudioTime;
let frame: Buffer | null = null;
if (pcmBuffer.length >= BYTES_PER_FRAME) {
// Real audio available
frame = pcmBuffer.slice(0, BYTES_PER_FRAME);
pcmBuffer = pcmBuffer.slice(BYTES_PER_FRAME);
// Track level for logging
dbAccum += rmsDb(frame);
dbCount++;
if (playerPaused) {
discordPlayer.unpause();
playerPaused = false;
console.log("[transmit] Transmitting — Discord indicator ON");
}
} else if (msSinceAudio < SILENCE_TAIL_MS && msSinceAudio > 0) {
// Buffer drained but audio was recent — pad silence to avoid OGG gap
frame = SILENCE_FRAME;
} else if (!playerPaused && msSinceAudio >= SILENCE_TAIL_MS) {
// No audio for a while — pause Discord indicator
discordPlayer.pause();
playerPaused = true;
console.log("[transmit] Stopped — Discord indicator OFF");
return;
} else {
return; // already paused, nothing to do
}
// Write one frame. If encoder is backpressured, skip this tick to avoid stalling.
const ok = opusEncoder.write(frame);
if (!ok) {
opusEncoder.once('drain', () => {}); // re-arm drain without blocking
}
}, 20);
@@ -105,15 +163,12 @@ export function startWebserver(port: number = 3000) {
if (!Buffer.isBuffer(data)) return;
lastBrowserAudioTime = Date.now();
// Upsample browser 24kHz mono → 48kHz stereo
const upsampled = upsample24kMonoTo48kStereo(data);
pcmBuffer = Buffer.concat([pcmBuffer, upsampled]);
// Upsample 24kHz mono → 48kHz stereo and add to buffer
const upsampled = upsample(data);
// Encode complete Opus frames
while (pcmBuffer.length >= BYTES_PER_FRAME) {
const frame = pcmBuffer.slice(0, BYTES_PER_FRAME);
pcmBuffer = pcmBuffer.slice(BYTES_PER_FRAME);
opusEncoder.write(frame);
// Cap buffer to avoid runaway growth during stall
if (pcmBuffer.length < MAX_BUF_BYTES) {
pcmBuffer = Buffer.concat([pcmBuffer, upsampled]);
}
});