Merge pull request #645 from webadderallorg/fix/native-mac-duplicate-mic-audio

fix: avoid duplicate mic audio in native mac recordings

electron/ipc/recording/diagnostics.test.ts

@@ -134,6 +134,32 @@ describe("getCompanionAudioFallbackPaths", () => {
 		]);
 	});
 
+	it("prefers the mac mic companion alone when embedded audio already exists and no system sidecar is present", async () => {
+		const videoPath = path.join(tempRoot, "recording.mp4");
+		const micPath = path.join(tempRoot, "recording.mic.m4a");
+
+		await Promise.all([fs.writeFile(videoPath, "video"), fs.writeFile(micPath, "mic")]);
+
+		execFileMock.mockImplementation(
+			(
+				_file: string,
+				_args: string[],
+				_options: Record<string, unknown>,
+				callback: ExecFileCallback,
+			) => {
+				const error = new Error("ffmpeg probe found embedded audio") as Error & {
+					stderr?: string;
+				};
+				error.stderr = "Stream #0:1: Audio: aac";
+				callback(error, "", error.stderr);
+			},
+		);
+
+		const { getCompanionAudioFallbackPaths } = await import("./diagnostics");
+
+		await expect(getCompanionAudioFallbackPaths(videoPath)).resolves.toEqual([micPath]);
+	});
+
 	it("loads saved sidecar timing metadata alongside companion audio paths", async () => {
 		const videoPath = path.join(tempRoot, "recording.mp4");
 		const micPath = path.join(tempRoot, "recording.mic.webm");

electron/ipc/recording/diagnostics.ts

@@ -508,20 +508,41 @@ export async function getCompanionAudioFallbackInfo(videoPath: string) {
 
 	let paths: string[];
 	if (await hasEmbeddedAudioStream(videoPath)) {
-		const companionPaths = Array.from(
+		const hasUsableMacSystemCompanion = companionCandidates.some(
+			(candidate) =>
+				candidate.platform === "mac" &&
+				candidate.usablePaths.includes(candidate.systemPath),
+		);
+		const usableMacMicOnlyCompanions = Array.from(
 			new Set(
 				companionCandidates.flatMap((candidate) =>
-					candidate.usablePaths.filter(
-						(companionPath) => companionPath === candidate.micPath,
-					),
+					candidate.platform === "mac" &&
+					!candidate.usablePaths.includes(candidate.systemPath) &&
+					candidate.usablePaths.includes(candidate.micPath)
+						? [candidate.micPath]
+						: [],
 				),
 			),
 		);
-		if (companionPaths.length === 0) {
-			return { paths: [], startDelayMsByPath: {} };
-		}
 
-		paths = [videoPath, ...companionPaths];
+		if (!hasUsableMacSystemCompanion && usableMacMicOnlyCompanions.length > 0) {
+			paths = usableMacMicOnlyCompanions;
+		} else {
+			const companionPaths = Array.from(
+				new Set(
+					companionCandidates.flatMap((candidate) =>
+						candidate.usablePaths.filter(
+							(companionPath) => companionPath === candidate.micPath,
+						),
+					),
+				),
+			);
+			if (companionPaths.length === 0) {
+				return { paths: [], startDelayMsByPath: {} };
+			}
+
+			paths = [videoPath, ...companionPaths];
+		}
 	} else {
 		paths = Array.from(
 			new Set(companionCandidates.flatMap((candidate) => candidate.usablePaths)),

electron/native/ScreenCaptureKitRecorder.swift

@@ -77,9 +77,6 @@ final class ScreenCaptureRecorder: NSObject, SCStreamOutput, SCStreamDelegate {
 		}
 		writesSystemAudioToSeparateTrack = capturesSystemAudio
 		writesMicrophoneToSeparateTrack = capturesSystemAudio && capturesMicrophone
-		if capturesMicrophone && !capturesSystemAudio {
-			writesMicrophoneToSeparateTrack = true
-		}
 		let requestedFPS = max(targetCaptureFPS, config.fps ?? targetCaptureFPS)
 		streamConfig.minimumFrameInterval = CMTime(value: 1, timescale: CMTimeScale(requestedFPS))
 		streamConfig.queueDepth = 6

src/lib/exporter/audioEncoder.test.ts

@@ -161,6 +161,28 @@ describe("AudioProcessor offline render preparation", () => {
 		expect(renderAndMuxOfflineAudio).toHaveBeenCalled();
 	});
 
+	it("avoids the single-sidecar fast path for legacy mac mic sidecars that still need embedded audio", async () => {
+		const processor = new AudioProcessor() as unknown as OfflineRenderTestHarness;
+		const loadAudioFileDemuxer = vi.spyOn(processor, "loadAudioFileDemuxer");
+		const renderAndMuxOfflineAudio = vi
+			.spyOn(processor, "renderAndMuxOfflineAudio")
+			.mockResolvedValue();
+
+		await processor.process(
+			{} as never,
+			{} as never,
+			"file:///tmp/recording.mp4",
+			[],
+			[],
+			undefined,
+			[],
+			["/tmp/recording.mic.m4a"],
+		);
+
+		expect(loadAudioFileDemuxer).not.toHaveBeenCalled();
+		expect(renderAndMuxOfflineAudio).toHaveBeenCalled();
+	});
+
 	it("soft-limits mixed peaks before encoding or WAV conversion", () => {
 		const samples = new Float32Array([
 			-1.6,

src/lib/exporter/audioEncoder.ts

@@ -263,12 +263,18 @@ export class AudioProcessor {
 			videoUrl,
 			sortedSourceAudioFallbackPaths,
 		);
+		const requiresLegacyMacMicSidecarMix =
+			routingPolicy.includeEmbeddedInExport &&
+			!routingPolicy.hasEmbeddedSourceAudio &&
+			routingPolicy.playbackPaths.length === 1 &&
+			routingPolicy.playbackPaths[0]?.toLowerCase().endsWith(".mic.m4a") === true;
 		const hasTimedCompanionAudio = routingPolicy.playbackPaths.some(
 			(audioPath) => (sourceAudioFallbackStartDelayMsByPath?.[audioPath] ?? 0) > 0,
 		);
 		const needsSourceAudioMixing =
 			routingPolicy.playbackPaths.length > 1 ||
 			(routingPolicy.hasEmbeddedSourceAudio && routingPolicy.playbackPaths.length > 0) ||
+			requiresLegacyMacMicSidecarMix ||
 			hasTimedCompanionAudio;
 
 		// When speed edits, audio regions, or multiple audio sources need mixing, use offline AudioContext pipeline.