modules/access/qtsound.m

   1 /*****************************************************************************
   2 * qtsound.m: qtkit (Mac OS X) based audio capture module
   3 *****************************************************************************
   4 * Copyright © 2011 VLC authors and VideoLAN
   5 *
   6 * Authors: Pierre d'Herbemont <pdherbemont@videolan.org>
   7 *          Gustaf Neumann <neumann@wu.ac.at>
   8 *          Michael S. Feurstein <michael.feurstein@wu.ac.at>
   9 *
  10 *****************************************************************************
  11 * This library is free software; you can redistribute it and/or
  12 * modify it under the terms of the GNU Lesser General Public License
  13 * as published by the Free Software Foundation; either version 2.1
  14 * of the License, or (at your option) any later version.
  15 *
  16 * This library is distributed in the hope that it will be useful,
  17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  19 * Lesser General Public License for more details.
  20 *
  21 * You should have received a copy of the GNU Lesser General Public
  22 * License along with this library; if not, write to the Free Software
  23 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110, USA
  24 *
  25 *****************************************************************************/
  26
  27 /*****************************************************************************
  28  * Preamble
  29  *****************************************************************************/
  30
  31 #ifdef HAVE_CONFIG_H
  32 # include "config.h"
  33 #endif
  34
  35 #include <vlc_common.h>
  36 #include <vlc_plugin.h>
  37 #include <vlc_aout.h>
  38
  39 #include <vlc_demux.h>
  40 #include <vlc_dialog.h>
  41
  42 #define QTKIT_VERSION_MIN_REQUIRED 70603
  43
  44 #import <QTKit/QTKit.h>
  45
  46 /*****************************************************************************
  47  * Local prototypes.
  48  *****************************************************************************/
  49 static int Open(vlc_object_t *p_this);
  50 static void Close(vlc_object_t *p_this);
  51 static int Demux(demux_t *p_demux);
  52 static int Control(demux_t *, int, va_list);
  53
  54 /*****************************************************************************
  55  * Module descriptor
  56  *****************************************************************************/
  57
  58 vlc_module_begin()
  59 set_shortname(N_("QTSound"))
  60 set_description(N_("QuickTime Sound Capture"))
  61 set_category(CAT_INPUT)
  62 set_subcategory(SUBCAT_INPUT_ACCESS)
  63 add_shortcut("qtsound")
  64 set_capability("access_demux", 0)
  65 set_callbacks(Open, Close)
  66 vlc_module_end ()
  67
  68
  69 /*****************************************************************************
  70  * QTKit Bridge
  71  *****************************************************************************/
  72 @interface VLCDecompressedAudioOutput : QTCaptureDecompressedAudioOutput
  73 {
  74     demux_t *p_qtsound;
  75     AudioBuffer *currentAudioBuffer;
  76     void *rawAudioData;
  77     UInt32 numberOfSamples;
  78     date_t date;
  79     mtime_t currentPts;
  80     mtime_t previousPts;
  81 }
  82 - (id)initWithDemux:(demux_t *)p_demux;
  83 - (void)outputAudioSampleBuffer:(QTSampleBuffer *)sampleBuffer fromConnection:(QTCaptureConnection *)connection;
  84 - (BOOL)checkCurrentAudioBuffer;
  85 - (void)freeAudioMem;
  86 - (mtime_t)getCurrentPts;
  87 - (void *)getCurrentAudioBufferData;
  88 - (UInt32)getCurrentTotalDataSize;
  89 - (UInt32)getNumberOfSamples;
  90
  91 @end
  92
  93 @implementation VLCDecompressedAudioOutput : QTCaptureDecompressedAudioOutput
  94 - (id)initWithDemux:(demux_t *)p_demux
  95 {
  96     if (self = [super init]) {
  97         p_qtsound = p_demux;
  98         currentAudioBuffer = nil;
  99         date_Init(&date, 44100, 1);
 100         date_Set(&date,0);
 101         currentPts = 0;
 102         previousPts = 0;
 103     }
 104     return self;
 105 }
 106 - (void)dealloc
 107 {
 108     [super dealloc];
 109 }
 110
 111 - (void)outputAudioSampleBuffer:(QTSampleBuffer *)sampleBuffer fromConnection:(QTCaptureConnection *)connection
 112 {
 113     AudioBufferList *tempAudioBufferList;
 114     UInt32 totalDataSize = 0;
 115     UInt32 count = 0;
 116
 117     @synchronized (self) {
 118         numberOfSamples = [sampleBuffer numberOfSamples];
 119         date_Increment(&date,numberOfSamples);
 120         currentPts = date_Get(&date);
 121
 122         tempAudioBufferList = [sampleBuffer audioBufferListWithOptions:0];
 123         if (tempAudioBufferList->mNumberBuffers == 2) {
 124             /*
 125              * Compute totalDataSize as sum of all data blocks in the
 126              * audio buffer list:
 127              */
 128             for (count = 0; count < tempAudioBufferList->mNumberBuffers; count++)
 129                 totalDataSize += tempAudioBufferList->mBuffers[count].mDataByteSize;
 130
 131             /*
 132              * Allocate storage for the interleaved audio data
 133              */
 134             rawAudioData = malloc(totalDataSize);
 135             if (NULL == rawAudioData) {
 136                 msg_Err(p_qtsound, "Raw audiodata could not be allocated");
 137                 return;
 138             }
 139         } else {
 140             msg_Err(p_qtsound, "Too many or only one channel found: %i.",
 141                                tempAudioBufferList->mNumberBuffers);
 142             return;
 143         }
 144
 145         /*
 146          * Interleave raw data (provided in two separate channels as
 147          * F32L) with 2 samples per frame
 148          */
 149         if (totalDataSize) {
 150             unsigned short i;
 151             const float *b1Ptr, *b2Ptr;
 152             float *uPtr;
 153
 154             for (i = 0,
 155                  uPtr = (float *)rawAudioData,
 156                  b1Ptr = (const float *) tempAudioBufferList->mBuffers[0].mData,
 157                  b2Ptr = (const float *) tempAudioBufferList->mBuffers[1].mData;
 158                  i < numberOfSamples; i++) {
 159                 *uPtr++ = *b1Ptr++;
 160                 *uPtr++ = *b2Ptr++;
 161             }
 162
 163             if (currentAudioBuffer == nil) {
 164                 currentAudioBuffer = (AudioBuffer *)malloc(sizeof(AudioBuffer));
 165                 if (NULL == currentAudioBuffer) {
 166                     msg_Err(p_qtsound, "AudioBuffer could not be allocated.");
 167                     return;
 168                 }
 169             }
 170             currentAudioBuffer->mNumberChannels = 2;
 171             currentAudioBuffer->mDataByteSize = totalDataSize;
 172             currentAudioBuffer->mData = rawAudioData;
 173         }
 174     }
 175 }
 176
 177 - (BOOL)checkCurrentAudioBuffer
 178 {
 179     return (currentAudioBuffer) ? 1 : 0;
 180 }
 181
 182 - (void)freeAudioMem
 183 {
 184     FREENULL(rawAudioData);
 185 }
 186
 187 - (mtime_t)getCurrentPts
 188 {
 189     /* FIXME: can this getter be minimized? */
 190     mtime_t pts;
 191
 192     if(!currentAudioBuffer || currentPts == previousPts)
 193         return 0;
 194
 195     @synchronized (self) {
 196         pts = previousPts = currentPts;
 197     }
 198
 199     return (currentAudioBuffer->mData) ? currentPts : 0;
 200 }
 201
 202 - (void *)getCurrentAudioBufferData
 203 {
 204     return currentAudioBuffer->mData;
 205 }
 206
 207 - (UInt32)getCurrentTotalDataSize
 208 {
 209     return currentAudioBuffer->mDataByteSize;
 210 }
 211
 212 - (UInt32)getNumberOfSamples
 213 {
 214     return numberOfSamples;
 215 }
 216
 217 @end
 218
 219 /*****************************************************************************
 220  * Struct
 221  *****************************************************************************/
 222
 223 struct demux_sys_t {
 224     QTCaptureSession * session;
 225     QTCaptureDevice * audiodevice;
 226     VLCDecompressedAudioOutput * audiooutput;
 227     es_out_id_t *p_es_audio;
 228 };
 229
 230 /*****************************************************************************
 231  * Open: initialize interface
 232  *****************************************************************************/
 233 static int Open(vlc_object_t *p_this)
 234 {
 235     demux_t *p_demux = (demux_t*)p_this;
 236     demux_sys_t *p_sys;
 237     es_format_t audiofmt;
 238     char *psz_uid = NULL;
 239     int audiocodec;
 240     bool success;
 241     NSString *qtk_curraudiodevice_uid;
 242     NSArray *myAudioDevices, *audioformat_array;
 243     QTFormatDescription *audio_format;
 244     QTCaptureDeviceInput *audioInput;
 245     NSError *o_returnedAudioError;
 246
 247     @autoreleasepool {
 248         if(p_demux->psz_location && *p_demux->psz_location)
 249             psz_uid = p_demux->psz_location;
 250
 251         msg_Dbg(p_demux, "qtsound uid = %s", psz_uid);
 252         qtk_curraudiodevice_uid = [[NSString alloc] initWithFormat:@"%s", psz_uid];
 253
 254         p_demux->p_sys = p_sys = calloc(1, sizeof(demux_sys_t));
 255         if(!p_sys)
 256             return VLC_ENOMEM;
 257
 258         msg_Dbg(p_demux, "qtsound : uid = %s", [qtk_curraudiodevice_uid UTF8String]);
 259         myAudioDevices = [[[QTCaptureDevice inputDevicesWithMediaType:QTMediaTypeSound]
 260                            arrayByAddingObjectsFromArray:[QTCaptureDevice inputDevicesWithMediaType:QTMediaTypeMuxed]] retain];
 261         if([myAudioDevices count] == 0) {
 262             vlc_dialog_display_error(p_demux, _("No Audio Input device found"),
 263                 _("Your Mac does not seem to be equipped with a suitable audio input device."
 264                 "Please check your connectors and drivers."));
 265             msg_Err(p_demux, "Can't find any Audio device");
 266
 267             goto error;
 268         }
 269         unsigned iaudio;
 270         for (iaudio = 0; iaudio < [myAudioDevices count]; iaudio++) {
 271             QTCaptureDevice *qtk_audioDevice;
 272             qtk_audioDevice = [myAudioDevices objectAtIndex:iaudio];
 273             msg_Dbg(p_demux, "qtsound audio %u/%lu localizedDisplayName: %s uniqueID: %s",
 274                     iaudio, [myAudioDevices count],
 275                     [[qtk_audioDevice localizedDisplayName] UTF8String],
 276                     [[qtk_audioDevice uniqueID] UTF8String]);
 277             if ([[[qtk_audioDevice uniqueID] stringByTrimmingCharactersInSet:
 278                   [NSCharacterSet whitespaceCharacterSet]] isEqualToString:qtk_curraudiodevice_uid]) {
 279                 msg_Dbg(p_demux, "Device found");
 280                 break;
 281             }
 282         }
 283
 284         audioInput = nil;
 285         if(iaudio < [myAudioDevices count])
 286             p_sys->audiodevice = [myAudioDevices objectAtIndex:iaudio];
 287         else {
 288             /* cannot find designated audio device, fall back to open default audio device */
 289             msg_Dbg(p_demux, "Cannot find designated uid audio device as %s. Fall back to open default audio device.", [qtk_curraudiodevice_uid UTF8String]);
 290             p_sys->audiodevice = [QTCaptureDevice defaultInputDeviceWithMediaType: QTMediaTypeSound];
 291         }
 292         if(!p_sys->audiodevice) {
 293             vlc_dialog_display_error(p_demux, _("No audio input device found"),
 294                 _("Your Mac does not seem to be equipped with a suitable audio input device."
 295                 "Please check your connectors and drivers."));
 296             msg_Err(p_demux, "Can't find any Audio device");
 297
 298             goto error;
 299         }
 300
 301         if(![p_sys->audiodevice open: &o_returnedAudioError]) {
 302             msg_Err(p_demux, "Unable to open the audio capture device (%ld)", [o_returnedAudioError code]);
 303             goto error;
 304         }
 305
 306         if([p_sys->audiodevice isInUseByAnotherApplication] == YES) {
 307             msg_Err(p_demux, "default audio capture device is exclusively in use by another application");
 308             goto error;
 309         }
 310         audioInput = [[QTCaptureDeviceInput alloc] initWithDevice: p_sys->audiodevice];
 311         if(!audioInput) {
 312             msg_Err(p_demux, "can't create a valid audio capture input facility");
 313             goto error;
 314         } else
 315             msg_Dbg(p_demux, "created valid audio capture input facility");
 316
 317         p_sys->audiooutput = [[VLCDecompressedAudioOutput alloc] initWithDemux:p_demux];
 318         msg_Dbg (p_demux, "initialized audio output");
 319
 320         /* Get the formats */
 321         /*
 322          FIXME: the format description gathered here does not seem to be the same
 323          in comparison to the format description collected from the actual sampleBuffer.
 324          This information needs to be updated some other place. For the time being this shall suffice.
 325
 326          The following verbose output is an example of what is read from the input device during the below block
 327          [0x3042138] qtsound demux debug: Audio localized format summary: Linear PCM, 24 bit little-endian signed integer, 2 channels, 44100 Hz
 328          [0x3042138] qtsound demux debug: Sample Rate: 44100; Format ID: lpcm; Format Flags: 00000004; Bytes per Packet: 8; Frames per Packet: 1; Bytes per Frame: 8; Channels per Frame: 2; Bits per Channel: 24
 329          [0x3042138] qtsound demux debug: Flag float 0 bigEndian 0 signedInt 1 packed 0 alignedHigh 0 non interleaved 0 non mixable 0
 330          canonical 0 nativeFloatPacked 0 nativeEndian 0
 331
 332          However when reading this information from the sampleBuffer during the delegate call from
 333          - (void)outputAudioSampleBuffer:(QTSampleBuffer *)sampleBuffer fromConnection:(QTCaptureConnection *)connection;
 334          the following data shows up
 335          2011-09-23 22:06:03.077 VLC[23070:f103] Audio localized format summary: Linear PCM, 32 bit little-endian floating point, 2 channels, 44100 Hz
 336          2011-09-23 22:06:03.078 VLC[23070:f103] Sample Rate: 44100; Format ID: lpcm; Format Flags: 00000029; Bytes per Packet: 4; Frames per Packet: 1; Bytes per Frame: 4; Channels per Frame: 2; Bits per Channel: 32
 337          2011-09-23 22:06:03.078 VLC[23070:f103] Flag float 1 bigEndian 0 signedInt 0 packed 1 alignedHigh 0 non interleaved 1 non mixable 0
 338          canonical 1 nativeFloatPacked 1 nativeEndian 0
 339
 340          Note the differences
 341          24bit vs. 32bit
 342          little-endian signed integer vs. little-endian floating point
 343          format flag 00000004 vs. 00000029
 344          bytes per packet 8 vs. 4
 345          packed 0 vs. 1
 346          non interleaved 0 vs. 1 -> this makes a major difference when filling our own buffer
 347          canonical 0 vs. 1
 348          nativeFloatPacked 0 vs. 1
 349
 350          One would assume we'd need to feed the (es_format_t)audiofmt with the data collected here.
 351          This is not the case. Audio will be transmitted in artefacts, due to wrong information.
 352
 353          At the moment this data is set manually, however one should consider trying to set this data dynamically
 354          */
 355         audioformat_array = [p_sys->audiodevice formatDescriptions];
 356         audio_format = NULL;
 357         for(int k = 0; k < [audioformat_array count]; k++) {
 358             audio_format = (QTFormatDescription *)[audioformat_array objectAtIndex:k];
 359
 360             msg_Dbg(p_demux, "Audio localized format summary: %s", [[audio_format localizedFormatSummary] UTF8String]);
 361             msg_Dbg(p_demux, "Audio format description attributes: %s",[[[audio_format formatDescriptionAttributes] description] UTF8String]);
 362
 363             AudioStreamBasicDescription asbd = {0};
 364             NSValue *asbdValue =  [audio_format attributeForKey:QTFormatDescriptionAudioStreamBasicDescriptionAttribute];
 365             [asbdValue getValue:&asbd];
 366
 367             char formatIDString[5];
 368             UInt32 formatID = CFSwapInt32HostToBig (asbd.mFormatID);
 369             bcopy (&formatID, formatIDString, 4);
 370             formatIDString[4] = '\0';
 371
 372             /* kept for development purposes */
 373 #if 0
 374             msg_Dbg(p_demux, "Sample Rate: %.0lf; Format ID: %s; Format Flags: %.8x; Bytes per Packet: %d; Frames per Packet: %d; Bytes per Frame: %d; Channels per Frame: %d; Bits per Channel: %d",
 375                     asbd.mSampleRate,
 376                     formatIDString,
 377                     asbd.mFormatFlags,
 378                     asbd.mBytesPerPacket,
 379                     asbd.mFramesPerPacket,
 380                     asbd.mBytesPerFrame,
 381                     asbd.mChannelsPerFrame,
 382                     asbd.mBitsPerChannel);
 383
 384             msg_Dbg(p_demux, "Flag float %d bigEndian %d signedInt %d packed %d alignedHigh %d non interleaved %d non mixable %d\ncanonical %d nativeFloatPacked %d nativeEndian %d",
 385                     (asbd.mFormatFlags & kAudioFormatFlagIsFloat) != 0,
 386                     (asbd.mFormatFlags & kAudioFormatFlagIsBigEndian) != 0,
 387                     (asbd.mFormatFlags & kAudioFormatFlagIsSignedInteger) != 0,
 388                     (asbd.mFormatFlags & kAudioFormatFlagIsPacked) != 0,
 389                     (asbd.mFormatFlags & kAudioFormatFlagIsAlignedHigh) != 0,
 390                     (asbd.mFormatFlags & kAudioFormatFlagIsNonInterleaved) != 0,
 391                     (asbd.mFormatFlags & kAudioFormatFlagIsNonMixable) != 0,
 392
 393                     (asbd.mFormatFlags & kAudioFormatFlagsCanonical) != 0,
 394                     (asbd.mFormatFlags & kAudioFormatFlagsNativeFloatPacked) != 0,
 395                     (asbd.mFormatFlags & kAudioFormatFlagsNativeEndian) != 0
 396                     );
 397 #endif
 398         }
 399
 400         if([audioformat_array count])
 401             audio_format = [audioformat_array objectAtIndex:0];
 402         else
 403             goto error;
 404
 405         /* Now we can init */
 406         audiocodec = VLC_CODEC_FL32;
 407         es_format_Init(&audiofmt, AUDIO_ES, audiocodec);
 408
 409         audiofmt.audio.i_format = audiocodec;
 410         audiofmt.audio.i_rate = 44100;
 411         /*
 412          * i_physical_channels Describes the channels configuration of the
 413          * samples (ie. number of channels which are available in the
 414          * buffer, and positions).
 415          */
 416         audiofmt.audio.i_physical_channels = AOUT_CHAN_RIGHT | AOUT_CHAN_LEFT;
 417         /*
 418          * Please note that it may be completely arbitrary - buffers are not
 419          * obliged to contain a integral number of so-called "frames". It's
 420          * just here for the division:
 421          * buffer_size = i_nb_samples * i_bytes_per_frame / i_frame_length
 422          */
 423         audiofmt.audio.i_bitspersample = 32;
 424         audiofmt.audio.i_channels = 2;
 425         audiofmt.audio.i_blockalign = audiofmt.audio.i_channels * (audiofmt.audio.i_bitspersample / 8);
 426         audiofmt.i_bitrate = audiofmt.audio.i_channels * audiofmt.audio.i_rate * audiofmt.audio.i_bitspersample;
 427
 428         p_sys->session = [[QTCaptureSession alloc] init];
 429
 430         success = [p_sys->session addInput:audioInput error: &o_returnedAudioError];
 431         if(!success) {
 432             msg_Err(p_demux, "the audio capture device could not be added to capture session (%ld)", [o_returnedAudioError code]);
 433             goto error;
 434         }
 435
 436         success = [p_sys->session addOutput:p_sys->audiooutput error: &o_returnedAudioError];
 437         if(!success) {
 438             msg_Err(p_demux, "audio output could not be added to capture session (%ld)", [o_returnedAudioError code]);
 439             goto error;
 440         }
 441
 442         [p_sys->session startRunning];
 443
 444         /* Set up p_demux */
 445         p_demux->pf_demux = Demux;
 446         p_demux->pf_control = Control;
 447         p_demux->info.i_update = 0;
 448         p_demux->info.i_title = 0;
 449         p_demux->info.i_seekpoint = 0;
 450
 451         msg_Dbg(p_demux, "New audio es %d channels %dHz",
 452                 audiofmt.audio.i_channels, audiofmt.audio.i_rate);
 453
 454         p_sys->p_es_audio = es_out_Add(p_demux->out, &audiofmt);
 455
 456         [audioInput release];
 457
 458         msg_Dbg(p_demux, "QTSound: We have an audio device ready!");
 459
 460         return VLC_SUCCESS;
 461     error:
 462         [audioInput release];
 463
 464         free(p_sys);
 465
 466         return VLC_EGENERIC;
 467     }
 468 }
 469
 470 /*****************************************************************************
 471  * Close: destroy interface
 472  *****************************************************************************/
 473 static void Close(vlc_object_t *p_this)
 474 {
 475     @autoreleasepool {
 476         demux_t *p_demux = (demux_t*)p_this;
 477         demux_sys_t *p_sys = p_demux->p_sys;
 478
 479         [p_sys->session performSelectorOnMainThread:@selector(stopRunning) withObject:nil waitUntilDone:NO];
 480         [p_sys->audiooutput performSelectorOnMainThread:@selector(release) withObject:nil waitUntilDone:NO];
 481         [p_sys->session performSelectorOnMainThread:@selector(release) withObject:nil waitUntilDone:NO];
 482
 483         free(p_sys);
 484     }
 485 }
 486
 487 /*****************************************************************************
 488  * Demux:
 489  *****************************************************************************/
 490 static int Demux(demux_t *p_demux)
 491 {
 492     demux_sys_t *p_sys = p_demux->p_sys;
 493     block_t *p_blocka = nil;
 494
 495     @autoreleasepool {
 496         @synchronized (p_sys->audiooutput) {
 497             if ([p_sys->audiooutput checkCurrentAudioBuffer]) {
 498                 unsigned i_buffer_size = [p_sys->audiooutput getCurrentTotalDataSize];
 499                 p_blocka = block_Alloc(i_buffer_size);
 500
 501                 if(!p_blocka) {
 502                     msg_Err(p_demux, "cannot get audio block");
 503                     return 0;
 504                 }
 505
 506                 memcpy(p_blocka->p_buffer, [p_sys->audiooutput getCurrentAudioBufferData], i_buffer_size);
 507                 p_blocka->i_nb_samples = [p_sys->audiooutput getNumberOfSamples];
 508                 p_blocka->i_pts = [p_sys->audiooutput getCurrentPts];
 509
 510                 [p_sys->audiooutput freeAudioMem];
 511             }
 512         }
 513     }
 514
 515     if (p_blocka) {
 516         if (!p_blocka->i_pts) {
 517             block_Release(p_blocka);
 518
 519             // Nothing to transfer yet, just forget
 520             msleep(10000);
 521             return 1;
 522         }
 523
 524         es_out_SetPCR(p_demux->out, p_blocka->i_pts);
 525         es_out_Send(p_demux->out, p_sys->p_es_audio, p_blocka);
 526     }
 527
 528     return 1;
 529 }
 530
 531 /*****************************************************************************
 532  * Control:
 533  *****************************************************************************/
 534 static int Control(demux_t *p_demux, int i_query, va_list args)
 535 {
 536     bool *pb;
 537     int64_t *pi64;
 538
 539     switch(i_query) {
 540             /* Special for access_demux */
 541         case DEMUX_CAN_PAUSE:
 542         case DEMUX_CAN_SEEK:
 543         case DEMUX_SET_PAUSE_STATE:
 544         case DEMUX_CAN_CONTROL_PACE:
 545             pb = (bool*)va_arg(args, bool *);
 546             *pb = false;
 547             return VLC_SUCCESS;
 548
 549         case DEMUX_GET_PTS_DELAY:
 550             pi64 = (int64_t*)va_arg(args, int64_t *);
 551             *pi64 = INT64_C(1000) * var_InheritInteger(p_demux, "live-caching");
 552             return VLC_SUCCESS;
 553
 554         default:
 555             return VLC_EGENERIC;
 556     }
 557     return VLC_EGENERIC;
 558 }