prepend first section of audio while detecting speech (was cut before)

web/client.html

@@ -51,7 +51,7 @@
         let isSpeaking = false;
         let soundDetected = false;
         let speakingCount = 0;
-        let SILENCE_DELAY_MS = 100;
+        let SILENCE_DELAY_MS = 50; //was 100 with good results
         const volumeHistory = [];
 
         let canvas = document.getElementById("canvas");
@@ -100,9 +100,9 @@
             if (averageVolume > threshold) {
                 if (autosend.checked && speakingCount == 0 && audioRecorder) {
                     console.log("startint new recording");
+                    soundDetected = false;
                     audioRecorder.stop();
                     audioRecorder.start();
-                    soundDetected = true;
                 }
                 speakingCount++;
                 if (speakingCount > 7) {
@@ -111,13 +111,14 @@
                     isSpeaking = true;
                     console.log("Was silent and is now speaking. (" + averageVolume + " averageVolume).");
                 }
-            } else {
+            } else if (averageVolume - 5 < threshold) {
                 speakingCount = 0;
                 if (isSpeaking) {
                     silenceCount++;
-                    if (silenceCount > 5) {
+                    if (silenceCount > 3) {
                         if (autosend.checked) {
                             console.log("Was speakng and is now silence. (" + averageVolume + " averageVolume). Sending audio to server.");
+                            soundDetected = true;
                             audioRecorder.stop();
                             audioRecorder.start();
                         }
@@ -182,7 +183,6 @@
                 });
         };
 
-
         function onmessage(event) {
             let latency = Date.now() - serverTime;
             console.log("Received message from server: " + event.data + " (latency: " + latency + "ms)");
@@ -197,6 +197,10 @@
             }
         }
 
+
+        const preBufferDuration = 500 ; // duration of pre-buffer in ms
+
+
         function startListening() {
             // Initialize canvas
             canvasCtx.fillStyle = "green";
@@ -204,14 +208,39 @@
             navigator.mediaDevices.getUserMedia({ audio: { sampleRate: 16000, echoCancellation: true } }).then((stream) => {
                 audioRecorder = new MediaRecorder(stream);
                 audioRecorder.start();
+
+                console.log("Started listening to microphone (sample rate: " + stream.getAudioTracks()[0].getSettings().sampleRate + " Hz, echoCancellation: " + stream.getAudioTracks()[0].getSettings().echoCancellation + ", " +audioRecorder.mimeType.audioChannels + " channels)");
+                //const preBufferLength = Math.ceil(preBufferDuration * audioRecorder.mimeType.audioSampleRate * audioRecorder.mimeType.audioChannels);
+                const preBufferLength = Math.ceil(preBufferDuration  * 48000 * 2 );
+                var preBuffer = [];
+
                 audioRecorder.addEventListener("dataavailable", (event) => {
                     console.log("Audio data available: " + event.data.size + " bytes");
                     if (!soundDetected && autosend.checked) {
                         console.log("discarding audio data because not speaking");
+                        //store last 100ms of audio data
+                        preBuffer = [];
+                        const audioData = event.data;
+                        //const start = audioBlob.size - preBufferLength; // calculate offset to trim pre-buffer
+                        //const audioBlob = new Blob([...preBuffer, audioData], { type: "audio/wav" });
+                        //const end = audioBlob.size;
+                        //const slicedAudio = audioBlob.slice(start, end);
+                        const audioBlob = new Blob([...preBuffer, audioData], { type: "audio/ogg; codecs=opus" });
+                        const start = audioBlob.size - 500 * 48 * 2 / 8; // Assuming 48 kHz sampling rate and 16-bit PCM audio
+                        const end = audioBlob.size;
+                        const slicedAudio = audioBlob.slice(start, end);
+                        preBuffer = slicedAudio;
+                        //sendAudioToServer(event.data);
                         return;
                     }
                     if (event.data.size > 0) {
-                        sendAudioToServer(event.data);
+                        let data = event.data;
+                        if (preBuffer.length > 0) {
+                            const audioBlob = new Blob([...preBuffer, audioData], { type: "audio/ogg; codecs=opus" });
+                            const newEventChunk = new Blob([audioBlob, preBuffer], { type: "audio/ogg; codecs=opus" });
+                            data = newEventChunk;
+                        }
+                        sendAudioToServer(data);
                         soundDetected = false;
                     }
                 });
@@ -223,6 +252,21 @@
             recordButton.classList.add("recording");
         }
 
+        function getlast500ms(audioData, preBuffer) {
+            const audioBlob = new Blob([...preBuffer, audioData], { type: "audio/ogg; codecs=opus" });
+            const start = audioBlob.size - 500 * 48000  * 2 / 8; // Assuming 48 kHz sampling rate and 16-bit PCM audio
+            const end = audioBlob.size;
+            const slicedAudio = audioBlob.slice(start, end);
+          
+            // Create a new Blob object with the remaining audio data
+            const remainingAudio = audioBlob.slice(0, start);
+            const newEvent = new Blob([remainingAudio], { type: "audio/ogg; codecs=opus" });
+          
+            // Replace the original event data with the new event data
+            const newEventChunk = new Blob([newEvent, slicedAudio], { type: "audio/ogg; codecs=opus" });
+            return newEventChunk;
+          }
+
         function stopListening() {
             recording = false;
             audioRecorder.stop();
@@ -262,7 +306,7 @@
                         return device.kind === 'audioinput';
                     });
                     console.log(audioInputDevices.length + ' audio input devices found');
-                    
+
                     // If more than one audio input device is available, populate the select list
                     if (audioInputDevices.length > 1) {
                         audioInputDevices.forEach(function (device) {
@@ -312,7 +356,7 @@
             //transcription = document.getElementById("transcription");
             //autosend = document.getElementById("autosend");
             statusRecording = document.getElementById("status-recording");
-            
+
             enumerateDevices();
             connect(socket);
         };

web/server.js

@@ -35,7 +35,7 @@ wss.on('connection', (ws) => {
             }
         };
 
-
+        //"yyyymmdd-hhMMss"
         var timestampfilename = Date.now("yyyymmdd-hhMMss");
 
         //save the audio data to a file to /rec subfolder 
@@ -85,7 +85,6 @@ app.get('/wsurl', (req, res) => {
     res.send(process.env.WS_URL, 200, { 'Content-Type': 'text/plain' });
 });
 
-
 app.listen(8080, () => {
     console.log('Server listening on port 8080');
 });