fix vc android delay

android/app/src/main/assets/public/voice.js

@@ -24,10 +24,6 @@ let voiceLoopRestarts = 0;
 let voiceDeviceChangeHandler = null;
 let voiceDeviceChangeTimer = null;
 let isVoiceSpeakerMode = false;
-let voiceTapNode = null;
-let pcmBatch = null;
-let pcmBatchLen = 0;
-let voicePcmFlushTimer = null;
 
 const SPEAKING_HANGOVER_MS = 400;
 
@@ -123,37 +119,15 @@ function int16ToBase64(view) {
     return btoa(bin);
 }
 
-function flushPcmBatch() {
-    if (!pcmBatchLen || !pcmBatch) return;
-    const native = getNativeVoiceRouting();
-    if (native && native.nativeAudioFeed && native.nativeAudioStart) {
-        const slice = new Int16Array(pcmBatch.subarray(0, pcmBatchLen));
-        try {
-            native.nativeAudioFeed({ data: int16ToBase64(slice) }).catch(() => {});
-        } catch (e) {}
-    }
-    pcmBatchLen = 0;
-}
-
 function handlePcmChunk(ev) {
+    const native = getNativeVoiceRouting();
+    if (!native || !native.nativeAudioFeed || !ev.data) return;
     try {
-        if (!getNativeVoiceRouting() || !pcmBatch) return;
+        native.nativeAudioFeed({ data: int16ToBase64(new Int16Array(ev.data)) }).catch(() => {});
-        const chunk = new Int16Array(ev.data);
-        for (let i = 0; i < chunk.length; i++) {
-            if (pcmBatchLen >= pcmBatch.length) flushPcmBatch();
-            pcmBatch[pcmBatchLen++] = chunk[i];
-        }
     } catch (e) {}
 }
 
 function stopNativePlayback() {
-    if (voicePcmFlushTimer) { clearInterval(voicePcmFlushTimer); voicePcmFlushTimer = null; }
-    flushPcmBatch();
-    if (voiceTapNode) {
-        try { voiceTapNode.disconnect(); } catch (e) {}
-        try { voiceTapNode.port.onmessage = null; } catch (e) {}
-        voiceTapNode = null;
-    }
     const native = getNativeVoiceRouting();
     if (native && native.nativeAudioStop) native.nativeAudioStop().catch(() => {});
 }
@@ -226,15 +200,10 @@ function setupVoiceOutput() {
     // App + native plugin: the mixed call audio is played by a native AudioTrack
     // (USAGE_VOICE_COMMUNICATION), so it follows the call volume slider and the
     // earpiece/speaker switch. Nothing is output through the WebAudio graph.
+    // The receiver worklet forwards mixed PCM straight here (~20 ms packets).
     if (native && isMobileVoiceDevice()) {
         stopNativePlayback();
-        pcmBatch = new Int16Array(1536); // 32 ms @ 48 kHz
+        receiverNode.port.onmessage = handlePcmChunk;
-        pcmBatchLen = 0;
-        voiceTapNode = new AudioWorkletNode(audioContext, 'audio-tap');
-        voiceTapNode.port.onmessage = handlePcmChunk;
-        receiverNode.connect(voiceTapNode);
-        if (voicePcmFlushTimer) clearInterval(voicePcmFlushTimer);
-        voicePcmFlushTimer = setInterval(flushPcmBatch, 25);
         if (native.nativeAudioStart) native.nativeAudioStart({ sampleRate: audioContext.sampleRate }).catch(() => {});
         return;
     }
@@ -435,10 +404,20 @@ class AudioProcessor extends AudioWorkletProcessor {
     }
 }
 class AudioReceiver extends AudioWorkletProcessor {
-    constructor() {
+    // Jitter buffer targets in samples (@48 kHz):
+    //   PREBUF_START  - initial prebuffer before first playback (~43 ms)
+    //   PREBUF_RESUME - re-prebuffer after a starvation event (~32 ms)
+    //   HARD_LIMIT    - burst ceiling (~128 ms): snap back to target immediately
+    // Backlog slightly above target is drained by a gentle trim (1 sample per
+    // quantum) instead of letting latency creep up to the hard limit.
+    constructor(options) {
         super();
         this.users = {};
         this.volumes = {};
+        const opts = (options && options.processorOptions) || {};
+        this.forward = !!opts.forward;
+        this.acc = new Int16Array(960); // ~20 ms @ 48 kHz
+        this.accLen = 0;
 
         this.port.onmessage = (e) => {
             if (e.data.action === 'clear') { this.users = {}; this.volumes = {}; return; }
@@ -446,7 +425,7 @@ class AudioReceiver extends AudioWorkletProcessor {
             if (e.data.action === 'remove') { delete this.users[e.data.clientId]; return; }
 
             const { clientId, audioData } = e.data;
-            if (!this.users[clientId]) this.users[clientId] = { buffer: new Int16Array(65536), readPtr: 0, writePtr: 0, buffering: true };
+            if (!this.users[clientId]) this.users[clientId] = { buffer: new Int16Array(65536), readPtr: 0, writePtr: 0, buffering: true, started: false };
             const user = this.users[clientId];
             const intData = new Int16Array(audioData);
             for (let i = 0; i < intData.length; i++) {
@@ -462,12 +441,13 @@ class AudioReceiver extends AudioWorkletProcessor {
         for (const id in this.users) {
             const user = this.users[id];
             let available = (user.writePtr - user.readPtr + 65536) % 65536;
-            if (user.buffering && available >= 4096) user.buffering = false;
+            if (user.buffering && available >= (user.started ? 1536 : 2048)) { user.buffering = false; user.started = true; }
 
             if (!user.buffering) {
                 if (available < output.length) { user.buffering = true; }
                 else {
-                    if (available > 24576) user.readPtr = (user.writePtr - 8192 + 65536) % 65536;
+                    if (available > 6144) user.readPtr = (user.writePtr - 2048 + 65536) % 65536;
+                    else if (available > 2049) user.readPtr = (user.readPtr + 1) % 65536;
                     let vol = this.volumes[id] !== undefined ? this.volumes[id] : 1.0;
                     for (let i = 0; i < output.length; i++) {
                         output[i] += (user.buffer[user.readPtr] / 32768) * vol;
@@ -475,28 +455,21 @@ class AudioReceiver extends AudioWorkletProcessor {
                     }
                 }
             }
-        } return true;
         }
-}
+        if (this.forward && this.port) {
-class AudioTap extends AudioWorkletProcessor {
+            for (let i = 0; i < output.length; i++) {
-    constructor() { super(); this.acc = new Int16Array(512); this.accLen = 0; }
+                const s = Math.max(-1, Math.min(1, output[i]));
-    process(inputs) {
-        const ch = inputs[0] && inputs[0][0];
-        if (ch && this.port) {
-            for (let i = 0; i < ch.length; i++) {
-                const s = Math.max(-1, Math.min(1, ch[i]));
                 this.acc[this.accLen++] = s < 0 ? s * 0x8000 : s * 0x7FFF;
-            }
                 if (this.accLen >= this.acc.length) {
                     this.port.postMessage(this.acc.buffer, [this.acc.buffer]);
-                this.acc = new Int16Array(512);
+                    this.acc = new Int16Array(960);
                     this.accLen = 0;
                 }
             }
+        }
         return true;
     }
 }
-registerProcessor('audio-tap', AudioTap);
 registerProcessor('audio-sender', AudioProcessor);
 registerProcessor('audio-receiver', AudioReceiver);`;
 
@@ -599,7 +572,8 @@ async function startVoiceSession(serverUrl) {
             if (!isAudioWorkletLoaded) {
                 await audioContext.audioWorklet.addModule(URL.createObjectURL(new Blob([voiceWorkletCode], {type: 'application/javascript'})));
                 senderNode = new AudioWorkletNode(audioContext, 'audio-sender');
-                receiverNode = new AudioWorkletNode(audioContext, 'audio-receiver');
+                const forwardPcm = !!(getNativeVoiceRouting() && isMobileVoiceDevice());
+                receiverNode = new AudioWorkletNode(audioContext, 'audio-receiver', {processorOptions: {forward: forwardPcm}});
                 screenReceiverNode = new AudioWorkletNode(audioContext, 'audio-receiver');
                 isAudioWorkletLoaded = true;
             }
@@ -1499,8 +1473,6 @@ function endVoiceSession() {
     isVoiceSpeakerMode = false;
     voiceLoopRestarts = 0;
     stopNativePlayback();
-    pcmBatch = null;
-    pcmBatchLen = 0;
 
     if (currentVoiceCallId) {
         let payload = JSON.stringify({ string1: currentVoiceCallType, string2: currentVoiceCallId });

android/app/src/main/java/olcxja/miarven/VoiceRoutingPlugin.java

@@ -47,8 +47,16 @@ public class VoiceRoutingPlugin extends Plugin {
 
     private AudioTrack nativeTrack = null;
     private Thread nativeFeedThread = null;
-    private final LinkedBlockingQueue<byte[]> pcmQueue = new LinkedBlockingQueue<>();
+    private final LinkedBlockingQueue<PcmChunk> pcmQueue = new LinkedBlockingQueue<>();
     private volatile boolean nativePlaying = false;
+    private double pipeDelayEmaMs = -1;
+    private long lastPipeLogNanos = 0;
+
+    private static final class PcmChunk {
+        final byte[] data;
+        final long arrivalNanos;
+        PcmChunk(byte[] data) { this.data = data; this.arrivalNanos = System.nanoTime(); }
+    }
 
     private final AudioManager.OnAudioFocusChangeListener focusChangeListener = fi -> {};
 
@@ -266,6 +274,8 @@ public class VoiceRoutingPlugin extends Plugin {
 
         nativeTrack = builder.build();
         pcmQueue.clear();
+        pipeDelayEmaMs = -1;
+        lastPipeLogNanos = 0;
         nativePlaying = true;
         try {
             nativeTrack.play();
@@ -278,7 +288,7 @@ public class VoiceRoutingPlugin extends Plugin {
         nativeFeedThread = new Thread(() -> {
             android.os.Process.setThreadPriority(android.os.Process.THREAD_PRIORITY_URGENT_AUDIO);
             while (nativePlaying) {
-                byte[] chunk;
+                PcmChunk chunk;
                 try {
                     chunk = pcmQueue.poll(100, TimeUnit.MILLISECONDS);
                 } catch (InterruptedException ie) {
@@ -287,7 +297,22 @@ public class VoiceRoutingPlugin extends Plugin {
                 if (chunk == null) continue;
                 AudioTrack track = nativeTrack;
                 if (track == null || !nativePlaying) break;
-                int written = track.write(chunk, 0, chunk.length, AudioTrack.WRITE_BLOCKING);
+                double delayMs = (System.nanoTime() - chunk.arrivalNanos) / 1e6;
+                pipeDelayEmaMs = pipeDelayEmaMs < 0 ? delayMs : (pipeDelayEmaMs * 0.9 + delayMs * 0.1);
+                long now = System.nanoTime();
+                if (lastPipeLogNanos == 0) lastPipeLogNanos = now;
+                if (now - lastPipeLogNanos >= 1_000_000_000L) {
+                    lastPipeLogNanos = now;
+                    int underruns = 0;
+                    try {
+                        underruns = track.getUnderrunCount();
+                    } catch (Exception ignored) {}
+                    log(String.format(java.util.Locale.US,
+                            "pipe: queue=%d emaDelay=%.1f ms buffer=%.0f ms underruns=%d",
+                            pcmQueue.size(), pipeDelayEmaMs,
+                            track.getBufferSizeInFrames() * 1000.0 / track.getSampleRate(), underruns));
+                }
+                int written = track.write(chunk.data, 0, chunk.data.length, AudioTrack.WRITE_BLOCKING);
                 if (written < 0) {
                     Log.w(TAG, "AudioTrack.write error: " + written);
                     break;
@@ -320,7 +345,7 @@ public class VoiceRoutingPlugin extends Plugin {
         while (pcmQueue.size() > 10) {
             pcmQueue.poll(); // drop oldest to avoid unbounded growth
         }
-        pcmQueue.offer(pcm);
+        pcmQueue.offer(new PcmChunk(pcm));
         call.resolve();
     }
 

node_modules/@capacitor/app/android/build/intermediates/incremental/lintVitalAnalyzeRelease/release-artifact-libraries.xml

@@ -1,9 +1,9 @@
 <libraries>
   <library
       name=":@@:capacitor-android::release"
-      jars="/home/olcxja/.gradle/caches/8.14.3/transforms/a60990329a178b65acb9987d031670ff/transformed/out/jars/classes.jar:/home/olcxja/.gradle/caches/8.14.3/transforms/a60990329a178b65acb9987d031670ff/transformed/out/jars/libs/R.jar"
+      jars="/home/olcxja/.gradle/caches/8.14.3/transforms/7687406e09940ba1a8559cc8a3590167/transformed/out/jars/classes.jar:/home/olcxja/.gradle/caches/8.14.3/transforms/7687406e09940ba1a8559cc8a3590167/transformed/out/jars/libs/R.jar"
       resolved="android:capacitor-android:unspecified"
-      folder="/home/olcxja/.gradle/caches/8.14.3/transforms/a60990329a178b65acb9987d031670ff/transformed/out"
+      folder="/home/olcxja/.gradle/caches/8.14.3/transforms/7687406e09940ba1a8559cc8a3590167/transformed/out"
       manifest="AndroidManifest.xml"
       resFolder="res"
       assetsFolder="assets"

node_modules/@capacitor/status-bar/android/build/intermediates/incremental/lintVitalAnalyzeRelease/release-artifact-libraries.xml

@@ -1,9 +1,9 @@
 <libraries>
   <library
       name=":@@:capacitor-android::release"
-      jars="/home/olcxja/.gradle/caches/8.14.3/transforms/a60990329a178b65acb9987d031670ff/transformed/out/jars/classes.jar:/home/olcxja/.gradle/caches/8.14.3/transforms/a60990329a178b65acb9987d031670ff/transformed/out/jars/libs/R.jar"
+      jars="/home/olcxja/.gradle/caches/8.14.3/transforms/7687406e09940ba1a8559cc8a3590167/transformed/out/jars/classes.jar:/home/olcxja/.gradle/caches/8.14.3/transforms/7687406e09940ba1a8559cc8a3590167/transformed/out/jars/libs/R.jar"
       resolved="android:capacitor-android:unspecified"
-      folder="/home/olcxja/.gradle/caches/8.14.3/transforms/a60990329a178b65acb9987d031670ff/transformed/out"
+      folder="/home/olcxja/.gradle/caches/8.14.3/transforms/7687406e09940ba1a8559cc8a3590167/transformed/out"
       manifest="AndroidManifest.xml"
       resFolder="res"
       assetsFolder="assets"

webroot/voice.js

@@ -24,10 +24,6 @@ let voiceLoopRestarts = 0;
 let voiceDeviceChangeHandler = null;
 let voiceDeviceChangeTimer = null;
 let isVoiceSpeakerMode = false;
-let voiceTapNode = null;
-let pcmBatch = null;
-let pcmBatchLen = 0;
-let voicePcmFlushTimer = null;
 
 const SPEAKING_HANGOVER_MS = 400;
 
@@ -123,37 +119,15 @@ function int16ToBase64(view) {
     return btoa(bin);
 }
 
-function flushPcmBatch() {
-    if (!pcmBatchLen || !pcmBatch) return;
-    const native = getNativeVoiceRouting();
-    if (native && native.nativeAudioFeed && native.nativeAudioStart) {
-        const slice = new Int16Array(pcmBatch.subarray(0, pcmBatchLen));
-        try {
-            native.nativeAudioFeed({ data: int16ToBase64(slice) }).catch(() => {});
-        } catch (e) {}
-    }
-    pcmBatchLen = 0;
-}
-
 function handlePcmChunk(ev) {
+    const native = getNativeVoiceRouting();
+    if (!native || !native.nativeAudioFeed || !ev.data) return;
     try {
-        if (!getNativeVoiceRouting() || !pcmBatch) return;
+        native.nativeAudioFeed({ data: int16ToBase64(new Int16Array(ev.data)) }).catch(() => {});
-        const chunk = new Int16Array(ev.data);
-        for (let i = 0; i < chunk.length; i++) {
-            if (pcmBatchLen >= pcmBatch.length) flushPcmBatch();
-            pcmBatch[pcmBatchLen++] = chunk[i];
-        }
     } catch (e) {}
 }
 
 function stopNativePlayback() {
-    if (voicePcmFlushTimer) { clearInterval(voicePcmFlushTimer); voicePcmFlushTimer = null; }
-    flushPcmBatch();
-    if (voiceTapNode) {
-        try { voiceTapNode.disconnect(); } catch (e) {}
-        try { voiceTapNode.port.onmessage = null; } catch (e) {}
-        voiceTapNode = null;
-    }
     const native = getNativeVoiceRouting();
     if (native && native.nativeAudioStop) native.nativeAudioStop().catch(() => {});
 }
@@ -226,15 +200,10 @@ function setupVoiceOutput() {
     // App + native plugin: the mixed call audio is played by a native AudioTrack
     // (USAGE_VOICE_COMMUNICATION), so it follows the call volume slider and the
     // earpiece/speaker switch. Nothing is output through the WebAudio graph.
+    // The receiver worklet forwards mixed PCM straight here (~20 ms packets).
     if (native && isMobileVoiceDevice()) {
         stopNativePlayback();
-        pcmBatch = new Int16Array(1536); // 32 ms @ 48 kHz
+        receiverNode.port.onmessage = handlePcmChunk;
-        pcmBatchLen = 0;
-        voiceTapNode = new AudioWorkletNode(audioContext, 'audio-tap');
-        voiceTapNode.port.onmessage = handlePcmChunk;
-        receiverNode.connect(voiceTapNode);
-        if (voicePcmFlushTimer) clearInterval(voicePcmFlushTimer);
-        voicePcmFlushTimer = setInterval(flushPcmBatch, 25);
         if (native.nativeAudioStart) native.nativeAudioStart({ sampleRate: audioContext.sampleRate }).catch(() => {});
         return;
     }
@@ -435,10 +404,20 @@ class AudioProcessor extends AudioWorkletProcessor {
     }
 }
 class AudioReceiver extends AudioWorkletProcessor {
-    constructor() {
+    // Jitter buffer targets in samples (@48 kHz):
+    //   PREBUF_START  - initial prebuffer before first playback (~43 ms)
+    //   PREBUF_RESUME - re-prebuffer after a starvation event (~32 ms)
+    //   HARD_LIMIT    - burst ceiling (~128 ms): snap back to target immediately
+    // Backlog slightly above target is drained by a gentle trim (1 sample per
+    // quantum) instead of letting latency creep up to the hard limit.
+    constructor(options) {
         super();
         this.users = {};
         this.volumes = {};
+        const opts = (options && options.processorOptions) || {};
+        this.forward = !!opts.forward;
+        this.acc = new Int16Array(960); // ~20 ms @ 48 kHz
+        this.accLen = 0;
 
         this.port.onmessage = (e) => {
             if (e.data.action === 'clear') { this.users = {}; this.volumes = {}; return; }
@@ -446,7 +425,7 @@ class AudioReceiver extends AudioWorkletProcessor {
             if (e.data.action === 'remove') { delete this.users[e.data.clientId]; return; }
 
             const { clientId, audioData } = e.data;
-            if (!this.users[clientId]) this.users[clientId] = { buffer: new Int16Array(65536), readPtr: 0, writePtr: 0, buffering: true };
+            if (!this.users[clientId]) this.users[clientId] = { buffer: new Int16Array(65536), readPtr: 0, writePtr: 0, buffering: true, started: false };
             const user = this.users[clientId];
             const intData = new Int16Array(audioData);
             for (let i = 0; i < intData.length; i++) {
@@ -462,12 +441,13 @@ class AudioReceiver extends AudioWorkletProcessor {
         for (const id in this.users) {
             const user = this.users[id];
             let available = (user.writePtr - user.readPtr + 65536) % 65536;
-            if (user.buffering && available >= 4096) user.buffering = false;
+            if (user.buffering && available >= (user.started ? 1536 : 2048)) { user.buffering = false; user.started = true; }
 
             if (!user.buffering) {
                 if (available < output.length) { user.buffering = true; }
                 else {
-                    if (available > 24576) user.readPtr = (user.writePtr - 8192 + 65536) % 65536;
+                    if (available > 6144) user.readPtr = (user.writePtr - 2048 + 65536) % 65536;
+                    else if (available > 2049) user.readPtr = (user.readPtr + 1) % 65536;
                     let vol = this.volumes[id] !== undefined ? this.volumes[id] : 1.0;
                     for (let i = 0; i < output.length; i++) {
                         output[i] += (user.buffer[user.readPtr] / 32768) * vol;
@@ -475,28 +455,21 @@ class AudioReceiver extends AudioWorkletProcessor {
                     }
                 }
             }
-        } return true;
         }
-}
+        if (this.forward && this.port) {
-class AudioTap extends AudioWorkletProcessor {
+            for (let i = 0; i < output.length; i++) {
-    constructor() { super(); this.acc = new Int16Array(512); this.accLen = 0; }
+                const s = Math.max(-1, Math.min(1, output[i]));
-    process(inputs) {
-        const ch = inputs[0] && inputs[0][0];
-        if (ch && this.port) {
-            for (let i = 0; i < ch.length; i++) {
-                const s = Math.max(-1, Math.min(1, ch[i]));
                 this.acc[this.accLen++] = s < 0 ? s * 0x8000 : s * 0x7FFF;
-            }
                 if (this.accLen >= this.acc.length) {
                     this.port.postMessage(this.acc.buffer, [this.acc.buffer]);
-                this.acc = new Int16Array(512);
+                    this.acc = new Int16Array(960);
                     this.accLen = 0;
                 }
             }
+        }
         return true;
     }
 }
-registerProcessor('audio-tap', AudioTap);
 registerProcessor('audio-sender', AudioProcessor);
 registerProcessor('audio-receiver', AudioReceiver);`;
 
@@ -599,7 +572,8 @@ async function startVoiceSession(serverUrl) {
             if (!isAudioWorkletLoaded) {
                 await audioContext.audioWorklet.addModule(URL.createObjectURL(new Blob([voiceWorkletCode], {type: 'application/javascript'})));
                 senderNode = new AudioWorkletNode(audioContext, 'audio-sender');
-                receiverNode = new AudioWorkletNode(audioContext, 'audio-receiver');
+                const forwardPcm = !!(getNativeVoiceRouting() && isMobileVoiceDevice());
+                receiverNode = new AudioWorkletNode(audioContext, 'audio-receiver', {processorOptions: {forward: forwardPcm}});
                 screenReceiverNode = new AudioWorkletNode(audioContext, 'audio-receiver');
                 isAudioWorkletLoaded = true;
             }
@@ -1499,8 +1473,6 @@ function endVoiceSession() {
     isVoiceSpeakerMode = false;
     voiceLoopRestarts = 0;
     stopNativePlayback();
-    pcmBatch = null;
-    pcmBatchLen = 0;
 
     if (currentVoiceCallId) {
         let payload = JSON.stringify({ string1: currentVoiceCallType, string2: currentVoiceCallId });