node-mac-recorder 2.21.52 → 2.21.54

package/.claude/settings.local.json

@@ -42,7 +42,13 @@
       "Bash(echo:*)",
       "Bash(timeout 3 node:*)",
       "Bash(timeout 25 node:*)",
-      "Bash(ls:*)"
+      "Bash(ls:*)",
+      "Bash(for f in test-output/temp_camera_1764838803672.mov test-output/temp_audio_1764838803672.mov test-output/duration-test-1764838803672.mov)",
+      "Bash(do)",
+      "Bash(grep:*)",
+      "Bash(timeout 10 ffprobe:*)",
+      "Bash(ffmpeg:*)",
+      "Bash(timeout 30 node:*)"
     ],
     "deny": [],
     "ask": []

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "node-mac-recorder",
-	"version": "2.21.52",
+	"version": "2.21.54",
 	"description": "Native macOS screen recording package for Node.js applications",
 	"main": "index.js",
 	"keywords": [

package/src/camera_recorder.mm

@@ -525,8 +525,10 @@ didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer
 
     CMTime timestamp = CMSampleBufferGetPresentationTimeStamp(sampleBuffer);
 
-    // SYNC INTEGRATION: Register camera as active stream
-    MRSyncMarkAudioSample(timestamp);
+    // Hold camera frames until we see audio so timelines stay aligned
+    if (MRSyncShouldHoldVideoFrame(timestamp)) {
+        return;
+    }
 
     // Start writer session on first frame
     if (!self.writerStarted) {
@@ -536,11 +538,31 @@ didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer
         }
         [self.writer startSessionAtSourceTime:kCMTimeZero];  // CRITICAL: t=0 timeline
         self.writerStarted = YES;
-        self.startTime = timestamp;
+        
+        // LIP SYNC FIX: Align camera startTime with audio's first timestamp for perfect lip sync
+        // This ensures camera and audio start from the same reference point
+        CMTime audioFirstTimestamp = MRSyncAudioFirstTimestamp();
+        CMTime alignmentOffset = MRSyncVideoAlignmentOffset();
+        
+        if (CMTIME_IS_VALID(audioFirstTimestamp)) {
+            // Use audio's first timestamp as reference - this is the key to lip sync
+            self.startTime = audioFirstTimestamp;
+            CMTime offset = CMTimeSubtract(timestamp, audioFirstTimestamp);
+            double offsetMs = CMTimeGetSeconds(offset) * 1000.0;
+            MRLog(@"🎥 Camera writer started @ t=0 (aligned with audio first timestamp, offset: %.1fms)", offsetMs);
+        } else if (CMTIME_IS_VALID(alignmentOffset)) {
+            // If audio came first, use the alignment offset to sync
+            self.startTime = CMTimeSubtract(timestamp, alignmentOffset);
+            double offsetMs = CMTimeGetSeconds(alignmentOffset) * 1000.0;
+            MRLog(@"🎥 Camera writer started @ t=0 (using alignment offset: %.1fms)", offsetMs);
+        } else {
+            // Fallback: use camera's own timestamp (should not happen if sync is configured)
+            self.startTime = timestamp;
+            MRLog(@"🎥 Camera writer started @ t=0 (source PTS: %.3fs, no audio sync available)", CMTimeGetSeconds(timestamp));
+        }
+        
         g_cameraStartTimestamp = CFAbsoluteTimeGetCurrent();
 
-        MRLog(@"🎥 Camera writer started @ t=0 (source PTS: %.3fs)", CMTimeGetSeconds(timestamp));
-
         // Signal start completion
         [self completeStart:YES token:self.activeToken];
     }
@@ -551,12 +573,17 @@ didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer
     }
 
     // TIMESTAMP NORMALIZATION (audio_recorder.mm pattern)
+    // LIP SYNC FIX: Use audio-aligned startTime for perfect synchronization
     CMTime adjustedTimestamp = kCMTimeZero;
     if (CMTIME_IS_VALID(self.startTime)) {
         adjustedTimestamp = CMTimeSubtract(timestamp, self.startTime);
         if (CMTIME_COMPARE_INLINE(adjustedTimestamp, <, kCMTimeZero)) {
             adjustedTimestamp = kCMTimeZero;
         }
+    } else {
+        // Fallback: if startTime not set, use current timestamp as base
+        // This should not happen if sync is working correctly
+        adjustedTimestamp = kCMTimeZero;
     }
 
     // Get pixel buffer from sample

package/src/screen_capture_kit.mm

@@ -475,8 +475,7 @@ extern "C" NSString *ScreenCaptureKitCurrentAudioPath(void) {
     }
     
     CMTime presentationTime = CMSampleBufferGetPresentationTimeStamp(sampleBuffer);
-    MRSyncMarkAudioSample(presentationTime);
-    
+
     // Wait for audio to arrive before starting screen video to prevent leading frames.
     if (MRSyncShouldHoldVideoFrame(presentationTime)) {
         return;
@@ -632,6 +631,7 @@ extern "C" NSString *ScreenCaptureKitCurrentAudioPath(void) {
     }
     
     CMTime presentationTime = CMSampleBufferGetPresentationTimeStamp(sampleBuffer);
+    MRSyncMarkAudioSample(presentationTime);
     
     if (!g_audioWriterStarted) {
         if (![g_audioWriter startWriting]) {

package/output/temp_cursor_1764800554355.json

@@ -1 +0,0 @@
-[{"x":549,"y":1055,"timestamp":23,"unixTimeMs":1764800556470,"cursorType":"text","type":"move","coordinateSystem":"video-relative","recordingType":"display","videoInfo":{"width":2048,"height":1330,"offsetX":0,"offsetY":0},"displayInfo":{"displayId":1,"width":2048,"height":1330},"_syncMetadata":{"videoStartTime":1764800554355,"cursorStartTime":1764800556447,"offsetMs":2092}}]

package/output/temp_cursor_1764800784424.json

@@ -1 +0,0 @@
-[{"x":412,"y":989,"timestamp":25,"unixTimeMs":1764800788789,"cursorType":"text","type":"move","coordinateSystem":"video-relative","recordingType":"display","videoInfo":{"width":2048,"height":1330,"offsetX":0,"offsetY":0},"displayInfo":{"displayId":1,"width":2048,"height":1330},"_syncMetadata":{"videoStartTime":1764800784424,"cursorStartTime":1764800788764,"offsetMs":4340}}]