@lvce-editor/renderer-process 30.42.2 → 30.43.1

package/dist/rendererProcessMain.js

@@ -8608,8 +8608,14 @@ const handleFocusOut = event => {
   }
   const uid = fromEvent(event);
   if (!relatedTarget && target && isInsideTitleBarMenu(target)) {
+    if (!target.isConnected) {
+      return;
+    }
     queueMicrotask(() => {
-      if (target.isConnected) {
+      const {
+        activeElement
+      } = document;
+      if (!activeElement || !isInsideTitleBarMenu(activeElement)) {
         closeMenu$1(uid);
       }
     });
@@ -10069,8 +10075,33 @@ const setupLevelMeter = (audioCtx, stream) => {
   source.connect(analyser);
   return analyser;
 };
+const isSpeechStoppedEvent = data => {
+  if (typeof data !== 'string') {
+    return false;
+  }
+  try {
+    const event = JSON.parse(data);
+    return event?.type === 'input_audio_buffer.speech_stopped';
+  } catch {
+    return false;
+  }
+};
+const startInputAudioRecorder = (micStream, audioDebugPort) => {
+  if (!audioDebugPort) {
+    return undefined;
+  }
+  const recorder = new MediaRecorder(micStream);
+  recorder.ondataavailable = event => {
+    if (event.data.size > 0) {
+      audioDebugPort.postMessage(event.data);
+    }
+  };
+  recorder.start();
+  return recorder;
+};
 const startWebRtcAudioStream = async options => {
   const {
+    audioDebugPort,
     elementLocator,
     port,
     trackAudioData,
@@ -10103,6 +10134,7 @@ const startWebRtcAudioStream = async options => {
   const micStream = await navigator.mediaDevices.getUserMedia({
     audio: true
   });
+  const inputAudioRecorder = startInputAudioRecorder(micStream, audioDebugPort);
   if (trackAudioData && audioCtx) {
     micAnalyzer = setupLevelMeter(audioCtx, micStream);
   }
@@ -10110,6 +10142,9 @@ const startWebRtcAudioStream = async options => {
   const dc = pc.createDataChannel('oai-events');
   dc.addEventListener('message', e => {
     port.postMessage(e.data);
+    if (inputAudioRecorder?.state === 'recording' && isSpeechStoppedEvent(e.data)) {
+      inputAudioRecorder.requestData();
+    }
   });
   port.onmessage = event => {
     const {
@@ -10126,8 +10161,10 @@ const startWebRtcAudioStream = async options => {
   await pc.setLocalDescription(offer);
   pcs[uid] = {
     audioCtx,
+    audioDebugPort,
     connection: pc,
     dataChannel: dc,
+    inputAudioRecorder,
     micAnalyzer,
     micStream,
     port,
@@ -10164,8 +10201,10 @@ const stopWebRtcAudioStream = async uid => {
   // TODO use disposableMap maybe?
   const {
     audioCtx,
+    audioDebugPort,
     connection,
     dataChannel,
+    inputAudioRecorder,
     micStream,
     port,
     remoteAudio
@@ -10174,10 +10213,15 @@ const stopWebRtcAudioStream = async uid => {
   connection.ontrack = null;
   connection.close();
   dataChannel.close();
+  if (inputAudioRecorder && inputAudioRecorder.state !== 'inactive') {
+    inputAudioRecorder.ondataavailable = null;
+    inputAudioRecorder.stop();
+  }
   for (const t of micStream.getTracks()) {
     t.stop();
   }
   port.close();
+  audioDebugPort?.close();
   remoteAudio.srcObject = null;
   if (audioCtx) {
     await audioCtx.close();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lvce-editor/renderer-process",
-  "version": "30.42.2",
+  "version": "30.43.1",
   "keywords": [
     "lvce-editor",
     "renderer-process"