libavcodec/libfdk-aacenc.c

   1 /*
   2  * AAC encoder wrapper
   3  * Copyright (c) 2012 Martin Storsjo
   4  *
   5  * This file is part of Libav.
   6  *
   7  * Permission to use, copy, modify, and/or distribute this software for any
   8  * purpose with or without fee is hereby granted, provided that the above
   9  * copyright notice and this permission notice appear in all copies.
  10  *
  11  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  12  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  13  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
  14  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  15  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  16  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  17  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  18  */
  19
  20 #include <fdk-aac/aacenc_lib.h>
  21
  22 #include "libavutil/channel_layout.h"
  23 #include "libavutil/common.h"
  24 #include "libavutil/opt.h"
  25 #include "avcodec.h"
  26 #include "audio_frame_queue.h"
  27 #include "internal.h"
  28
  29 #ifdef AACENCODER_LIB_VL0
  30 #define FDKENC_VER_AT_LEAST(vl0, vl1) \
  31     ((AACENCODER_LIB_VL0 > vl0) || \
  32      (AACENCODER_LIB_VL0 == vl0 && AACENCODER_LIB_VL1 >= vl1))
  33 #else
  34 #define FDKENC_VER_AT_LEAST(vl0, vl1) 0
  35 #endif
  36
  37 typedef struct AACContext {
  38     const AVClass *class;
  39     HANDLE_AACENCODER handle;
  40     int afterburner;
  41     int eld_sbr;
  42     int eld_v2;
  43     int signaling;
  44     int latm;
  45     int header_period;
  46     int vbr;
  47
  48     AudioFrameQueue afq;
  49 } AACContext;
  50
  51 static const AVOption aac_enc_options[] = {
  52     { "afterburner", "Afterburner (improved quality)", offsetof(AACContext, afterburner), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
  53     { "eld_sbr", "Enable SBR for ELD (for SBR in other configurations, use the -profile parameter)", offsetof(AACContext, eld_sbr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
  54 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
  55     { "eld_v2", "Enable ELDv2 (LD-MPS extension for ELD stereo signals)", offsetof(AACContext, eld_v2), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
  56 #endif
  57     { "signaling", "SBR/PS signaling style", offsetof(AACContext, signaling), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 2, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
  58     { "default", "Choose signaling implicitly (explicit hierarchical by default, implicit if global header is disabled)", 0, AV_OPT_TYPE_CONST, { .i64 = -1 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
  59     { "implicit", "Implicit backwards compatible signaling", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
  60     { "explicit_sbr", "Explicit SBR, implicit PS signaling", 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
  61     { "explicit_hierarchical", "Explicit hierarchical signaling", 0, AV_OPT_TYPE_CONST, { .i64 = 2 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
  62     { "latm", "Output LATM/LOAS encapsulated data", offsetof(AACContext, latm), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
  63     { "header_period", "StreamMuxConfig and PCE repetition period (in frames)", offsetof(AACContext, header_period), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 0xffff, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
  64     { "vbr", "VBR mode (1-5)", offsetof(AACContext, vbr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 5, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
  65     { NULL }
  66 };
  67
  68 static const AVClass aac_enc_class = {
  69     .class_name = "libfdk_aac",
  70     .item_name  = av_default_item_name,
  71     .option     = aac_enc_options,
  72     .version    = LIBAVUTIL_VERSION_INT,
  73 };
  74
  75 static const char *aac_get_error(AACENC_ERROR err)
  76 {
  77     switch (err) {
  78     case AACENC_OK:
  79         return "No error";
  80     case AACENC_INVALID_HANDLE:
  81         return "Invalid handle";
  82     case AACENC_MEMORY_ERROR:
  83         return "Memory allocation error";
  84     case AACENC_UNSUPPORTED_PARAMETER:
  85         return "Unsupported parameter";
  86     case AACENC_INVALID_CONFIG:
  87         return "Invalid config";
  88     case AACENC_INIT_ERROR:
  89         return "Initialization error";
  90     case AACENC_INIT_AAC_ERROR:
  91         return "AAC library initialization error";
  92     case AACENC_INIT_SBR_ERROR:
  93         return "SBR library initialization error";
  94     case AACENC_INIT_TP_ERROR:
  95         return "Transport library initialization error";
  96     case AACENC_INIT_META_ERROR:
  97         return "Metadata library initialization error";
  98     case AACENC_ENCODE_ERROR:
  99         return "Encoding error";
 100     case AACENC_ENCODE_EOF:
 101         return "End of file";
 102     default:
 103         return "Unknown error";
 104     }
 105 }
 106
 107 static int aac_encode_close(AVCodecContext *avctx)
 108 {
 109     AACContext *s = avctx->priv_data;
 110
 111     if (s->handle)
 112         aacEncClose(&s->handle);
 113     av_freep(&avctx->extradata);
 114     ff_af_queue_close(&s->afq);
 115
 116     return 0;
 117 }
 118
 119 static av_cold int aac_encode_init(AVCodecContext *avctx)
 120 {
 121     AACContext *s = avctx->priv_data;
 122     int ret = AVERROR(EINVAL);
 123     AACENC_InfoStruct info = { 0 };
 124     CHANNEL_MODE mode;
 125     AACENC_ERROR err;
 126     int aot = FF_PROFILE_AAC_LOW + 1;
 127     int sce = 0, cpe = 0;
 128
 129     if ((err = aacEncOpen(&s->handle, 0, avctx->channels)) != AACENC_OK) {
 130         av_log(avctx, AV_LOG_ERROR, "Unable to open the encoder: %s\n",
 131                aac_get_error(err));
 132         goto error;
 133     }
 134
 135     if (avctx->profile != FF_PROFILE_UNKNOWN)
 136         aot = avctx->profile + 1;
 137
 138     if ((err = aacEncoder_SetParam(s->handle, AACENC_AOT, aot)) != AACENC_OK) {
 139         av_log(avctx, AV_LOG_ERROR, "Unable to set the AOT %d: %s\n",
 140                aot, aac_get_error(err));
 141         goto error;
 142     }
 143
 144     if (aot == FF_PROFILE_AAC_ELD + 1 && s->eld_sbr) {
 145         if ((err = aacEncoder_SetParam(s->handle, AACENC_SBR_MODE,
 146                                        1)) != AACENC_OK) {
 147             av_log(avctx, AV_LOG_ERROR, "Unable to enable SBR for ELD: %s\n",
 148                    aac_get_error(err));
 149             goto error;
 150         }
 151     }
 152
 153     if ((err = aacEncoder_SetParam(s->handle, AACENC_SAMPLERATE,
 154                                    avctx->sample_rate)) != AACENC_OK) {
 155         av_log(avctx, AV_LOG_ERROR, "Unable to set the sample rate %d: %s\n",
 156                avctx->sample_rate, aac_get_error(err));
 157         goto error;
 158     }
 159
 160     switch (avctx->channels) {
 161     case 1: mode = MODE_1;       sce = 1; cpe = 0; break;
 162     case 2:
 163 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
 164       // (profile + 1) to map from profile range to AOT range
 165       if (aot == FF_PROFILE_AAC_ELD + 1 && s->eld_v2) {
 166           if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELMODE,
 167                                          128)) != AACENC_OK) {
 168               av_log(avctx, AV_LOG_ERROR, "Unable to enable ELDv2: %s\n",
 169                      aac_get_error(err));
 170               goto error;
 171           } else {
 172             mode = MODE_212;
 173             sce = 1;
 174             cpe = 0;
 175           }
 176       } else
 177 #endif
 178       {
 179         mode = MODE_2;
 180         sce = 0;
 181         cpe = 1;
 182       }
 183       break;
 184     case 3: mode = MODE_1_2;     sce = 1; cpe = 1; break;
 185     case 4: mode = MODE_1_2_1;   sce = 2; cpe = 1; break;
 186     case 5: mode = MODE_1_2_2;   sce = 1; cpe = 2; break;
 187     case 6: mode = MODE_1_2_2_1; sce = 2; cpe = 2; break;
 188 /* The version macro is introduced the same time as the 7.1 support, so this
 189    should suffice. */
 190 #if FDKENC_VER_AT_LEAST(3, 4) // 3.4.12
 191     case 8:
 192         sce = 2;
 193         cpe = 3;
 194         if (avctx->channel_layout == AV_CH_LAYOUT_7POINT1) {
 195             mode = MODE_7_1_REAR_SURROUND;
 196         } else {
 197             // MODE_1_2_2_2_1 and MODE_7_1_FRONT_CENTER use the same channel layout
 198             mode = MODE_7_1_FRONT_CENTER;
 199         }
 200         break;
 201 #endif
 202     default:
 203         av_log(avctx, AV_LOG_ERROR,
 204                "Unsupported number of channels %d\n", avctx->channels);
 205         goto error;
 206     }
 207
 208     if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELMODE,
 209                                    mode)) != AACENC_OK) {
 210         av_log(avctx, AV_LOG_ERROR,
 211                "Unable to set channel mode %d: %s\n", mode, aac_get_error(err));
 212         goto error;
 213     }
 214
 215     if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELORDER,
 216                                    1)) != AACENC_OK) {
 217         av_log(avctx, AV_LOG_ERROR,
 218                "Unable to set wav channel order %d: %s\n",
 219                mode, aac_get_error(err));
 220         goto error;
 221     }
 222
 223     if (avctx->flags & AV_CODEC_FLAG_QSCALE || s->vbr) {
 224         int mode = s->vbr ? s->vbr : avctx->global_quality;
 225         if (mode <  1 || mode > 5) {
 226             av_log(avctx, AV_LOG_WARNING,
 227                    "VBR quality %d out of range, should be 1-5\n", mode);
 228             mode = av_clip(mode, 1, 5);
 229         }
 230         av_log(avctx, AV_LOG_WARNING,
 231                "Note, the VBR setting is unsupported and only works with "
 232                "some parameter combinations\n");
 233         if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATEMODE,
 234                                        mode)) != AACENC_OK) {
 235             av_log(avctx, AV_LOG_ERROR, "Unable to set the VBR bitrate mode %d: %s\n",
 236                    mode, aac_get_error(err));
 237             goto error;
 238         }
 239     } else {
 240         if (avctx->bit_rate <= 0) {
 241             if (avctx->profile == FF_PROFILE_AAC_HE_V2) {
 242                 sce = 1;
 243                 cpe = 0;
 244             }
 245             avctx->bit_rate = (96*sce + 128*cpe) * avctx->sample_rate / 44;
 246             if (avctx->profile == FF_PROFILE_AAC_HE ||
 247                 avctx->profile == FF_PROFILE_AAC_HE_V2 ||
 248                 avctx->profile == FF_PROFILE_MPEG2_AAC_HE ||
 249                 s->eld_sbr)
 250                 avctx->bit_rate /= 2;
 251         }
 252         if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATE,
 253                                        avctx->bit_rate)) != AACENC_OK) {
 254             av_log(avctx, AV_LOG_ERROR, "Unable to set the bitrate %d: %s\n",
 255                    avctx->bit_rate, aac_get_error(err));
 256             goto error;
 257         }
 258     }
 259
 260     /* Choose bitstream format - if global header is requested, use
 261      * raw access units, otherwise use ADTS. */
 262     if ((err = aacEncoder_SetParam(s->handle, AACENC_TRANSMUX,
 263                                    avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER ? TT_MP4_RAW :
 264                                    s->latm ? TT_MP4_LOAS : TT_MP4_ADTS)) != AACENC_OK) {
 265         av_log(avctx, AV_LOG_ERROR, "Unable to set the transmux format: %s\n",
 266                aac_get_error(err));
 267         goto error;
 268     }
 269
 270     if (s->latm && s->header_period) {
 271         if ((err = aacEncoder_SetParam(s->handle, AACENC_HEADER_PERIOD,
 272                                        s->header_period)) != AACENC_OK) {
 273              av_log(avctx, AV_LOG_ERROR, "Unable to set header period: %s\n",
 274                     aac_get_error(err));
 275              goto error;
 276         }
 277     }
 278
 279     /* If no signaling mode is chosen, use explicit hierarchical signaling
 280      * if using mp4 mode (raw access units, with global header) and
 281      * implicit signaling if using ADTS. */
 282     if (s->signaling < 0)
 283         s->signaling = avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER ? 2 : 0;
 284
 285     if ((err = aacEncoder_SetParam(s->handle, AACENC_SIGNALING_MODE,
 286                                    s->signaling)) != AACENC_OK) {
 287         av_log(avctx, AV_LOG_ERROR, "Unable to set signaling mode %d: %s\n",
 288                s->signaling, aac_get_error(err));
 289         goto error;
 290     }
 291
 292     if ((err = aacEncoder_SetParam(s->handle, AACENC_AFTERBURNER,
 293                                    s->afterburner)) != AACENC_OK) {
 294         av_log(avctx, AV_LOG_ERROR, "Unable to set afterburner to %d: %s\n",
 295                s->afterburner, aac_get_error(err));
 296         goto error;
 297     }
 298
 299     if (avctx->cutoff > 0) {
 300         if (avctx->cutoff < (avctx->sample_rate + 255) >> 8 || avctx->cutoff > 20000) {
 301             av_log(avctx, AV_LOG_ERROR, "cutoff valid range is %d-20000\n",
 302                    (avctx->sample_rate + 255) >> 8);
 303             goto error;
 304         }
 305         if ((err = aacEncoder_SetParam(s->handle, AACENC_BANDWIDTH,
 306                                        avctx->cutoff)) != AACENC_OK) {
 307             av_log(avctx, AV_LOG_ERROR, "Unable to set the encoder bandwidth to %d: %s\n",
 308                    avctx->cutoff, aac_get_error(err));
 309             goto error;
 310         }
 311     }
 312
 313     if ((err = aacEncEncode(s->handle, NULL, NULL, NULL, NULL)) != AACENC_OK) {
 314         av_log(avctx, AV_LOG_ERROR, "Unable to initialize the encoder: %s\n",
 315                aac_get_error(err));
 316         return AVERROR(EINVAL);
 317     }
 318
 319     if ((err = aacEncInfo(s->handle, &info)) != AACENC_OK) {
 320         av_log(avctx, AV_LOG_ERROR, "Unable to get encoder info: %s\n",
 321                aac_get_error(err));
 322         goto error;
 323     }
 324
 325     avctx->frame_size = info.frameLength;
 326 #if FDKENC_VER_AT_LEAST(4, 0) // 4.0.0
 327     avctx->initial_padding = info.nDelay;
 328 #else
 329     avctx->initial_padding = info.encoderDelay;
 330 #endif
 331     ff_af_queue_init(avctx, &s->afq);
 332
 333     if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
 334         avctx->extradata_size = info.confSize;
 335         avctx->extradata      = av_mallocz(avctx->extradata_size +
 336                                            AV_INPUT_BUFFER_PADDING_SIZE);
 337         if (!avctx->extradata) {
 338             ret = AVERROR(ENOMEM);
 339             goto error;
 340         }
 341
 342         memcpy(avctx->extradata, info.confBuf, info.confSize);
 343     }
 344     return 0;
 345 error:
 346     aac_encode_close(avctx);
 347     return ret;
 348 }
 349
 350 static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
 351                             const AVFrame *frame, int *got_packet_ptr)
 352 {
 353     AACContext    *s        = avctx->priv_data;
 354     AACENC_BufDesc in_buf   = { 0 }, out_buf = { 0 };
 355     AACENC_InArgs  in_args  = { 0 };
 356     AACENC_OutArgs out_args = { 0 };
 357     int in_buffer_identifier = IN_AUDIO_DATA;
 358     int in_buffer_size, in_buffer_element_size;
 359     int out_buffer_identifier = OUT_BITSTREAM_DATA;
 360     int out_buffer_size, out_buffer_element_size;
 361     void *in_ptr, *out_ptr;
 362     int ret;
 363     uint8_t dummy_buf[1];
 364     AACENC_ERROR err;
 365
 366     /* handle end-of-stream small frame and flushing */
 367     if (!frame) {
 368         /* Must be a non-null pointer, even if it's a dummy. We could use
 369          * the address of anything else on the stack as well. */
 370         in_ptr               = dummy_buf;
 371         in_buffer_size       = 0;
 372
 373         in_args.numInSamples = -1;
 374     } else {
 375         in_ptr               = frame->data[0];
 376         in_buffer_size       = 2 * avctx->channels * frame->nb_samples;
 377
 378         in_args.numInSamples = avctx->channels * frame->nb_samples;
 379
 380         /* add current frame to the queue */
 381         if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
 382             return ret;
 383     }
 384
 385     in_buffer_element_size   = 2;
 386     in_buf.numBufs           = 1;
 387     in_buf.bufs              = &in_ptr;
 388     in_buf.bufferIdentifiers = &in_buffer_identifier;
 389     in_buf.bufSizes          = &in_buffer_size;
 390     in_buf.bufElSizes        = &in_buffer_element_size;
 391
 392     /* The maximum packet size is 6144 bits aka 768 bytes per channel. */
 393     if ((ret = ff_alloc_packet(avpkt, FFMAX(8192, 768 * avctx->channels)))) {
 394         av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
 395         return ret;
 396     }
 397
 398     out_ptr                   = avpkt->data;
 399     out_buffer_size           = avpkt->size;
 400     out_buffer_element_size   = 1;
 401     out_buf.numBufs           = 1;
 402     out_buf.bufs              = &out_ptr;
 403     out_buf.bufferIdentifiers = &out_buffer_identifier;
 404     out_buf.bufSizes          = &out_buffer_size;
 405     out_buf.bufElSizes        = &out_buffer_element_size;
 406
 407     if ((err = aacEncEncode(s->handle, &in_buf, &out_buf, &in_args,
 408                             &out_args)) != AACENC_OK) {
 409         if (!frame && err == AACENC_ENCODE_EOF)
 410             return 0;
 411         av_log(avctx, AV_LOG_ERROR, "Unable to encode frame: %s\n",
 412                aac_get_error(err));
 413         return AVERROR(EINVAL);
 414     }
 415
 416     if (!out_args.numOutBytes)
 417         return 0;
 418
 419     /* Get the next frame pts & duration */
 420     ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
 421                        &avpkt->duration);
 422
 423     avpkt->size     = out_args.numOutBytes;
 424     *got_packet_ptr = 1;
 425     return 0;
 426 }
 427
 428 static const AVProfile profiles[] = {
 429     { FF_PROFILE_AAC_LOW,   "LC"       },
 430     { FF_PROFILE_AAC_HE,    "HE-AAC"   },
 431     { FF_PROFILE_AAC_HE_V2, "HE-AACv2" },
 432     { FF_PROFILE_AAC_LD,    "LD"       },
 433     { FF_PROFILE_AAC_ELD,   "ELD"      },
 434     { FF_PROFILE_UNKNOWN },
 435 };
 436
 437 static const AVCodecDefault aac_encode_defaults[] = {
 438     { "b", "0" },
 439     { NULL }
 440 };
 441
 442 static const uint64_t aac_channel_layout[] = {
 443     AV_CH_LAYOUT_MONO,
 444     AV_CH_LAYOUT_STEREO,
 445     AV_CH_LAYOUT_SURROUND,
 446     AV_CH_LAYOUT_4POINT0,
 447     AV_CH_LAYOUT_5POINT0_BACK,
 448     AV_CH_LAYOUT_5POINT1_BACK,
 449 #if FDKENC_VER_AT_LEAST(3, 4) // 3.4.12
 450     AV_CH_LAYOUT_7POINT1_WIDE_BACK,
 451     AV_CH_LAYOUT_7POINT1,
 452 #endif
 453     0,
 454 };
 455
 456 static const int aac_sample_rates[] = {
 457     96000, 88200, 64000, 48000, 44100, 32000,
 458     24000, 22050, 16000, 12000, 11025, 8000, 0
 459 };
 460
 461 AVCodec ff_libfdk_aac_encoder = {
 462     .name                  = "libfdk_aac",
 463     .long_name             = NULL_IF_CONFIG_SMALL("Fraunhofer FDK AAC"),
 464     .type                  = AVMEDIA_TYPE_AUDIO,
 465     .id                    = AV_CODEC_ID_AAC,
 466     .priv_data_size        = sizeof(AACContext),
 467     .init                  = aac_encode_init,
 468     .encode2               = aac_encode_frame,
 469     .close                 = aac_encode_close,
 470     .capabilities          = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY,
 471     .sample_fmts           = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
 472                                                             AV_SAMPLE_FMT_NONE },
 473     .priv_class            = &aac_enc_class,
 474     .defaults              = aac_encode_defaults,
 475     .profiles              = profiles,
 476     .supported_samplerates = aac_sample_rates,
 477     .channel_layouts       = aac_channel_layout,
 478     .wrapper_name          = "libfdk",
 479 };