feat: add dual-track recording to capture each side of a call separately

Adds a "Record each side separately" toggle that captures the uplink
(your mic) and downlink (the other party) as two independent, time-synced
files instead of one mixed recording, so each speaker can be clearly
attributed for transcription.

Author: R2bEEaton

README.md

@@ -26,6 +26,7 @@ It can also be seen as an on-device wrapper for [scrcpy-server](https://github.c
     - Ignore specific contacts
     - Ignore all contacts
 - Saves recordings with **Opus** or **AAC** codec.
+- Optionally records each side of the call (uplink/downlink) as two separate, time-synced files, useful for transcription with clear speaker separation.
 - The app runs only on phone event changes, no persistent background process and notifications
 
 ## Requirements

app/src/main/aidl/com/kitsumed/shizucallrecorder/IShellService.aidl

@@ -40,6 +40,31 @@ interface IShellService {
      */
     void stopRecording() = 3;
 
+    /**
+     * Starts a second, independent audio-capture pipeline in parallel with [startRecording].
+     * Used for dual-track recording, where each side of a call (uplink/downlink) is captured
+     * as its own scrcpy-server process and its own pipe, so the two streams never mix.
+     *
+     * @param audioSource        scrcpy audio_source parameter (e.g. "voice-call-downlink").
+     * @param audioCodec         scrcpy audio_codec parameter (e.g. "opus", "aac").
+     * @param audioBitRate       scrcpy audio_bit_rate in bps (e.g. 16000 for 16 kbps Opus).
+     * @param serverPath      Absolute path to scrcpy-server.jar in shared storage.
+     * @param isDebuggingModeEnabled  When true, logs relay throughput every second.
+     * @return The read-end [ParcelFileDescriptor] of the audio pipe, or null on failure.
+     */
+    ParcelFileDescriptor startSecondaryRecording(
+        String audioSource,
+        String audioCodec,
+        int audioBitRate,
+        String serverPath,
+        boolean isDebuggingModeEnabled
+    ) = 7;
+
+    /**
+     * Stops the secondary audio capture pipeline started by [startSecondaryRecording].
+     */
+    void stopSecondaryRecording() = 8;
+
     /**
      * Grants an AppOp permission at the package level for a specific user profile.
      * @param packageName The package name of the app to grant the permission to.

app/src/main/java/com/kitsumed/shizucallrecorder/data/AppPreferences.kt

@@ -45,6 +45,7 @@ class AppPreferences(context: Context) {
         const val VIBRATION_ENABLED = true
         val CALL_DETECTION_MODE = CallDetectionMode.getDefaultModeForDevice().key
         const val RECORD_THIRD_PARTY_CALLS = false
+        const val DUAL_TRACK_RECORDING_ENABLED = false
 
         const val POST_RECORDING_FILE_ACTIONS_NOTIFICATION_ENABLED = false
         const val AUTO_RECORD_INCOMING = false
@@ -123,7 +124,8 @@ class AppPreferences(context: Context) {
         SHIZUKU_KEEP_ALIVE("shizuku_keep_alive"),
         SHIZUKU_AUTH_KEY("shizuku_auth_key"),
         CALL_DETECTION_MODE("call_detection_mode"),
-        RECORD_THIRD_PARTY_CALLS("record_third_party_calls");
+        RECORD_THIRD_PARTY_CALLS("record_third_party_calls"),
+        DUAL_TRACK_RECORDING_ENABLED("dual_track_recording_enabled");
     }
 
     // -------- Nested enums
@@ -377,6 +379,15 @@ class AppPreferences(context: Context) {
     /** Sets the configured audio bitrate. */
     fun setAudioBitRate(bitRate: Int) = setInt(Key.AUDIO_BITRATE, bitRate)
 
+    /**
+     * Checks if dual-track recording is enabled. When enabled, each side of the call is captured
+     * separately (uplink/downlink) into two files instead of one, ignoring [getAudioSource].
+     */
+    fun isDualTrackRecordingEnabled() = getBoolean(Key.DUAL_TRACK_RECORDING_ENABLED, DefaultsValue.DUAL_TRACK_RECORDING_ENABLED)
+
+    /** Sets whether dual-track recording is enabled. */
+    fun setDualTrackRecordingEnabled(enabled: Boolean) = setBoolean(Key.DUAL_TRACK_RECORDING_ENABLED, enabled)
+
     // -------- File Naming --------
 
     /** Gets the user configured file name template. */

app/src/main/java/com/kitsumed/shizucallrecorder/integrations/scrcpy/ScrcpyAudioMuxer.kt

@@ -37,10 +37,15 @@ import java.nio.ByteBuffer
  *
  * @param outputFileDescriptor Writable [java.io.FileDescriptor] for the output file (from SAF).
  * @param outputDisplayPath    Human-readable path for logs (e.g. "Recordings/call_….ogg").
+ * @param sharedOriginNanos    Optional [System.nanoTime] instant to use as PTS=0 instead of this
+ *                             muxer's own first packet. Used for dual-track recording so both
+ *                             tracks' timelines share the exact same wall-clock origin and stay
+ *                             in sync when merged externally (e.g. for transcription).
  */
 class ScrcpyAudioMuxer(
     private val outputFileDescriptor: java.io.FileDescriptor,
-    private val outputDisplayPath: String
+    private val outputDisplayPath: String,
+    private val sharedOriginNanos: Long? = null
 ) : Closeable {
 
     // Muxer state
@@ -144,7 +149,7 @@ class ScrcpyAudioMuxer(
         // dead time from our internal timeline so the final output file doesn't have a huge silence.
         // TODO: This code was added to fix most phone audio player behaving badly with the large silence, ideally we would like to keep the silence and not corrupt the file by doing so.
         if (firstPacketTimeNanos == -1L) {
-            firstPacketTimeNanos = nowNanos
+            firstPacketTimeNanos = sharedOriginNanos ?: nowNanos
             lastPacketWallClockNanos = nowNanos
             AppLogger.d( "First audio frame: wall-clock origin set, pts=0")
         } else {