Ios – How to record a conversation / phone call on iOS

audioaudio-recordingiosiphone

Is it theoretically possible to record a phone call on iPhone?

I'm accepting answers which:

  • may or may not require the phone to be jailbroken
  • may or may not pass apple's guidelines due to use of private API's (I don't care; it is not for the App Store)
  • may or may not use private SDKs

I don't want answers just bluntly saying "Apple does not allow that".
I know there would be no official way of doing it, and certainly not for an App Store application, and I know there are call recording apps which place outgoing calls through their own servers.

Best Answer

Here you go. Complete working example. Tweak should be loaded in mediaserverd daemon. It will record every phone call in /var/mobile/Media/DCIM/result.m4a. Audio file has two channels. Left is microphone, right is speaker. On iPhone 4S call is recorded only when the speaker is turned on. On iPhone 5, 5C and 5S call is recorded either way. There might be small hiccups when switching to/from speaker but recording will continue.

#import <AudioToolbox/AudioToolbox.h>
#import <libkern/OSAtomic.h>

//CoreTelephony.framework
extern "C" CFStringRef const kCTCallStatusChangeNotification;
extern "C" CFStringRef const kCTCallStatus;
extern "C" id CTTelephonyCenterGetDefault();
extern "C" void CTTelephonyCenterAddObserver(id ct, void* observer, CFNotificationCallback callBack, CFStringRef name, void *object, CFNotificationSuspensionBehavior sb);
extern "C" int CTGetCurrentCallCount();
enum
{
    kCTCallStatusActive = 1,
    kCTCallStatusHeld = 2,
    kCTCallStatusOutgoing = 3,
    kCTCallStatusIncoming = 4,
    kCTCallStatusHanged = 5
};

NSString* kMicFilePath = @"/var/mobile/Media/DCIM/mic.caf";
NSString* kSpeakerFilePath = @"/var/mobile/Media/DCIM/speaker.caf";
NSString* kResultFilePath = @"/var/mobile/Media/DCIM/result.m4a";

OSSpinLock phoneCallIsActiveLock = 0;
OSSpinLock speakerLock = 0;
OSSpinLock micLock = 0;

ExtAudioFileRef micFile = NULL;
ExtAudioFileRef speakerFile = NULL;

BOOL phoneCallIsActive = NO;

void Convert()
{
    //File URLs
    CFURLRef micUrl = CFURLCreateWithFileSystemPath(NULL, (CFStringRef)kMicFilePath, kCFURLPOSIXPathStyle, false);
    CFURLRef speakerUrl = CFURLCreateWithFileSystemPath(NULL, (CFStringRef)kSpeakerFilePath, kCFURLPOSIXPathStyle, false);
    CFURLRef mixUrl = CFURLCreateWithFileSystemPath(NULL, (CFStringRef)kResultFilePath, kCFURLPOSIXPathStyle, false);

    ExtAudioFileRef micFile = NULL;
    ExtAudioFileRef speakerFile = NULL;
    ExtAudioFileRef mixFile = NULL;

    //Opening input files (speaker and mic)
    ExtAudioFileOpenURL(micUrl, &micFile);
    ExtAudioFileOpenURL(speakerUrl, &speakerFile);

    //Reading input file audio format (mono LPCM)
    AudioStreamBasicDescription inputFormat, outputFormat;
    UInt32 descSize = sizeof(inputFormat);
    ExtAudioFileGetProperty(micFile, kExtAudioFileProperty_FileDataFormat, &descSize, &inputFormat);
    int sampleSize = inputFormat.mBytesPerFrame;

    //Filling input stream format for output file (stereo LPCM)
    FillOutASBDForLPCM(inputFormat, inputFormat.mSampleRate, 2, inputFormat.mBitsPerChannel, inputFormat.mBitsPerChannel, true, false, false);

    //Filling output file audio format (AAC)
    memset(&outputFormat, 0, sizeof(outputFormat));
    outputFormat.mFormatID = kAudioFormatMPEG4AAC;
    outputFormat.mSampleRate = 8000;
    outputFormat.mFormatFlags = kMPEG4Object_AAC_Main;
    outputFormat.mChannelsPerFrame = 2;

    //Opening output file
    ExtAudioFileCreateWithURL(mixUrl, kAudioFileM4AType, &outputFormat, NULL, kAudioFileFlags_EraseFile, &mixFile);
    ExtAudioFileSetProperty(mixFile, kExtAudioFileProperty_ClientDataFormat, sizeof(inputFormat), &inputFormat);

    //Freeing URLs
    CFRelease(micUrl);
    CFRelease(speakerUrl);
    CFRelease(mixUrl);

    //Setting up audio buffers
    int bufferSizeInSamples = 64 * 1024;

    AudioBufferList micBuffer;
    micBuffer.mNumberBuffers = 1;
    micBuffer.mBuffers[0].mNumberChannels = 1;
    micBuffer.mBuffers[0].mDataByteSize = sampleSize * bufferSizeInSamples;
    micBuffer.mBuffers[0].mData = malloc(micBuffer.mBuffers[0].mDataByteSize);

    AudioBufferList speakerBuffer;
    speakerBuffer.mNumberBuffers = 1;
    speakerBuffer.mBuffers[0].mNumberChannels = 1;
    speakerBuffer.mBuffers[0].mDataByteSize = sampleSize * bufferSizeInSamples;
    speakerBuffer.mBuffers[0].mData = malloc(speakerBuffer.mBuffers[0].mDataByteSize);

    AudioBufferList mixBuffer;
    mixBuffer.mNumberBuffers = 1;
    mixBuffer.mBuffers[0].mNumberChannels = 2;
    mixBuffer.mBuffers[0].mDataByteSize = sampleSize * bufferSizeInSamples * 2;
    mixBuffer.mBuffers[0].mData = malloc(mixBuffer.mBuffers[0].mDataByteSize);

    //Converting
    while (true)
    {
        //Reading data from input files
        UInt32 framesToRead = bufferSizeInSamples;
        ExtAudioFileRead(micFile, &framesToRead, &micBuffer);
        ExtAudioFileRead(speakerFile, &framesToRead, &speakerBuffer);
        if (framesToRead == 0)
        {
            break;
        }

        //Building interleaved stereo buffer - left channel is mic, right - speaker
        for (int i = 0; i < framesToRead; i++)
        {
            memcpy((char*)mixBuffer.mBuffers[0].mData + i * sampleSize * 2, (char*)micBuffer.mBuffers[0].mData + i * sampleSize, sampleSize);
            memcpy((char*)mixBuffer.mBuffers[0].mData + i * sampleSize * 2 + sampleSize, (char*)speakerBuffer.mBuffers[0].mData + i * sampleSize, sampleSize);
        }

        //Writing to output file - LPCM will be converted to AAC
        ExtAudioFileWrite(mixFile, framesToRead, &mixBuffer);
    }

    //Closing files
    ExtAudioFileDispose(micFile);
    ExtAudioFileDispose(speakerFile);
    ExtAudioFileDispose(mixFile);

    //Freeing audio buffers
    free(micBuffer.mBuffers[0].mData);
    free(speakerBuffer.mBuffers[0].mData);
    free(mixBuffer.mBuffers[0].mData);
}

void Cleanup()
{
    [[NSFileManager defaultManager] removeItemAtPath:kMicFilePath error:NULL];
    [[NSFileManager defaultManager] removeItemAtPath:kSpeakerFilePath error:NULL];
}

void CoreTelephonyNotificationCallback(CFNotificationCenterRef center, void *observer, CFStringRef name, const void *object, CFDictionaryRef userInfo)
{
    NSDictionary* data = (NSDictionary*)userInfo;

    if ([(NSString*)name isEqualToString:(NSString*)kCTCallStatusChangeNotification])
    {
        int currentCallStatus = [data[(NSString*)kCTCallStatus] integerValue];

        if (currentCallStatus == kCTCallStatusActive)
        {
            OSSpinLockLock(&phoneCallIsActiveLock);
            phoneCallIsActive = YES;
            OSSpinLockUnlock(&phoneCallIsActiveLock);
        }
        else if (currentCallStatus == kCTCallStatusHanged)
        {
            if (CTGetCurrentCallCount() > 0)
            {
                return;
            }

            OSSpinLockLock(&phoneCallIsActiveLock);
            phoneCallIsActive = NO;
            OSSpinLockUnlock(&phoneCallIsActiveLock);

            //Closing mic file
            OSSpinLockLock(&micLock);
            if (micFile != NULL)
            {
                ExtAudioFileDispose(micFile);
            }
            micFile = NULL;
            OSSpinLockUnlock(&micLock);

            //Closing speaker file
            OSSpinLockLock(&speakerLock);
            if (speakerFile != NULL)
            {
                ExtAudioFileDispose(speakerFile);
            }
            speakerFile = NULL;
            OSSpinLockUnlock(&speakerLock);

            Convert();
            Cleanup();
        }
    }
}

OSStatus(*AudioUnitProcess_orig)(AudioUnit unit, AudioUnitRenderActionFlags *ioActionFlags, const AudioTimeStamp *inTimeStamp, UInt32 inNumberFrames, AudioBufferList *ioData);
OSStatus AudioUnitProcess_hook(AudioUnit unit, AudioUnitRenderActionFlags *ioActionFlags, const AudioTimeStamp *inTimeStamp, UInt32 inNumberFrames, AudioBufferList *ioData)
{
    OSSpinLockLock(&phoneCallIsActiveLock);
    if (phoneCallIsActive == NO)
    {
        OSSpinLockUnlock(&phoneCallIsActiveLock);
        return AudioUnitProcess_orig(unit, ioActionFlags, inTimeStamp, inNumberFrames, ioData);
    }
    OSSpinLockUnlock(&phoneCallIsActiveLock);

    ExtAudioFileRef* currentFile = NULL;
    OSSpinLock* currentLock = NULL;

    AudioComponentDescription unitDescription = {0};
    AudioComponentGetDescription(AudioComponentInstanceGetComponent(unit), &unitDescription);
    //'agcc', 'mbdp' - iPhone 4S, iPhone 5
    //'agc2', 'vrq2' - iPhone 5C, iPhone 5S
    if (unitDescription.componentSubType == 'agcc' || unitDescription.componentSubType == 'agc2')
    {
        currentFile = &micFile;
        currentLock = &micLock;
    }
    else if (unitDescription.componentSubType == 'mbdp' || unitDescription.componentSubType == 'vrq2')
    {
        currentFile = &speakerFile;
        currentLock = &speakerLock;
    }

    if (currentFile != NULL)
    {
        OSSpinLockLock(currentLock);

        //Opening file
        if (*currentFile == NULL)
        {
            //Obtaining input audio format
            AudioStreamBasicDescription desc;
            UInt32 descSize = sizeof(desc);
            AudioUnitGetProperty(unit, kAudioUnitProperty_StreamFormat, kAudioUnitScope_Input, 0, &desc, &descSize);

            //Opening audio file
            CFURLRef url = CFURLCreateWithFileSystemPath(NULL, (CFStringRef)((currentFile == &micFile) ? kMicFilePath : kSpeakerFilePath), kCFURLPOSIXPathStyle, false);
            ExtAudioFileRef audioFile = NULL;
            OSStatus result = ExtAudioFileCreateWithURL(url, kAudioFileCAFType, &desc, NULL, kAudioFileFlags_EraseFile, &audioFile);
            if (result != 0)
            {
                *currentFile = NULL;
            }
            else
            {
                *currentFile = audioFile;

                //Writing audio format
                ExtAudioFileSetProperty(*currentFile, kExtAudioFileProperty_ClientDataFormat, sizeof(desc), &desc);
            }
            CFRelease(url);
        }
        else
        {
            //Writing audio buffer
            ExtAudioFileWrite(*currentFile, inNumberFrames, ioData);
        }

        OSSpinLockUnlock(currentLock);
    }

    return AudioUnitProcess_orig(unit, ioActionFlags, inTimeStamp, inNumberFrames, ioData);
}

__attribute__((constructor))
static void initialize()
{
    CTTelephonyCenterAddObserver(CTTelephonyCenterGetDefault(), NULL, CoreTelephonyNotificationCallback, NULL, NULL, CFNotificationSuspensionBehaviorHold);

    MSHookFunction(AudioUnitProcess, AudioUnitProcess_hook, &AudioUnitProcess_orig);
}

A few words about what's going on. AudioUnitProcess function is used for processing audio streams in order to apply some effects, mix, convert etc. We are hooking AudioUnitProcess in order to access phone call's audio streams. While phone call is active these streams are being processed in various ways.

We are listening for CoreTelephony notifications in order to get phone call status changes. When we receive audio samples we need to determine where they come from - microphone or speaker. This is done using componentSubType field in AudioComponentDescription structure. Now, you might think, why don't we store AudioUnit objects so that we don't need to check componentSubType every time. I did that but it will break everything when you switch speaker on/off on iPhone 5 because AudioUnit objects will change, they are recreated. So, now we open audio files (one for microphone and one for speaker) and write samples in them, simple as that. When phone call ends we will receive appropriate CoreTelephony notification and close the files. We have two separate files with audio from microphone and speaker that we need to merge. This is what void Convert() is for. It's pretty simple if you know the API. I don't think I need to explain it, comments are enough.

About locks. There are many threads in mediaserverd. Audio processing and CoreTelephony notifications are on different threads so we need some kind synchronization. I chose spin locks because they are fast and because the chance of lock contention is small in our case. On iPhone 4S and even iPhone 5 all the work in AudioUnitProcess should be done as fast as possible otherwise you will hear hiccups from device speaker which obviously not good.