FFmpeg: libavcodec/libfdk-aacenc.c Source File

00001 /*
00002  * AAC encoder wrapper
00003  * Copyright (c) 2012 Martin Storsjo
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00022 #include <fdk-aac/aacenc_lib.h>
00023 
00024 #include "avcodec.h"
00025 #include "audio_frame_queue.h"
00026 #include "internal.h"
00027 #include "libavutil/audioconvert.h"
00028 #include "libavutil/common.h"
00029 #include "libavutil/opt.h"
00030 
00031 typedef struct AACContext {
00032     const AVClass *class;
00033     HANDLE_AACENCODER handle;
00034     int afterburner;
00035     int eld_sbr;
00036     int signaling;
00037     int latm;
00038     int header_period;
00039     int vbr;
00040 
00041     AudioFrameQueue afq;
00042 } AACContext;
00043 
00044 static const AVOption aac_enc_options[] = {
00045     { "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 },
00046     { "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 },
00047     { "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" },
00048     { "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" },
00049     { "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" },
00050     { "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" },
00051     { "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" },
00052     { "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 },
00053     { "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 },
00054     { "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 },
00055     { NULL }
00056 };
00057 
00058 static const AVClass aac_enc_class = {
00059     "libfdk_aac", av_default_item_name, aac_enc_options, LIBAVUTIL_VERSION_INT
00060 };
00061 
00062 static const char *aac_get_error(AACENC_ERROR err)
00063 {
00064     switch (err) {
00065     case AACENC_OK:
00066         return "No error";
00067     case AACENC_INVALID_HANDLE:
00068         return "Invalid handle";
00069     case AACENC_MEMORY_ERROR:
00070         return "Memory allocation error";
00071     case AACENC_UNSUPPORTED_PARAMETER:
00072         return "Unsupported parameter";
00073     case AACENC_INVALID_CONFIG:
00074         return "Invalid config";
00075     case AACENC_INIT_ERROR:
00076         return "Initialization error";
00077     case AACENC_INIT_AAC_ERROR:
00078         return "AAC library initialization error";
00079     case AACENC_INIT_SBR_ERROR:
00080         return "SBR library initialization error";
00081     case AACENC_INIT_TP_ERROR:
00082         return "Transport library initialization error";
00083     case AACENC_INIT_META_ERROR:
00084         return "Metadata library initialization error";
00085     case AACENC_ENCODE_ERROR:
00086         return "Encoding error";
00087     case AACENC_ENCODE_EOF:
00088         return "End of file";
00089     default:
00090         return "Unknown error";
00091     }
00092 }
00093 
00094 static int aac_encode_close(AVCodecContext *avctx)
00095 {
00096     AACContext *s = avctx->priv_data;
00097 
00098     if (s->handle)
00099         aacEncClose(&s->handle);
00100 #if FF_API_OLD_ENCODE_AUDIO
00101     av_freep(&avctx->coded_frame);
00102 #endif
00103     av_freep(&avctx->extradata);
00104     ff_af_queue_close(&s->afq);
00105 
00106     return 0;
00107 }
00108 
00109 static av_cold int aac_encode_init(AVCodecContext *avctx)
00110 {
00111     AACContext *s = avctx->priv_data;
00112     int ret = AVERROR(EINVAL);
00113     AACENC_InfoStruct info = { 0 };
00114     CHANNEL_MODE mode;
00115     AACENC_ERROR err;
00116     int aot = FF_PROFILE_AAC_LOW + 1;
00117     int sce = 0, cpe = 0;
00118 
00119     if ((err = aacEncOpen(&s->handle, 0, avctx->channels)) != AACENC_OK) {
00120         av_log(avctx, AV_LOG_ERROR, "Unable to open the encoder: %s\n",
00121                aac_get_error(err));
00122         goto error;
00123     }
00124 
00125     if (avctx->profile != FF_PROFILE_UNKNOWN)
00126         aot = avctx->profile + 1;
00127 
00128     if ((err = aacEncoder_SetParam(s->handle, AACENC_AOT, aot)) != AACENC_OK) {
00129         av_log(avctx, AV_LOG_ERROR, "Unable to set the AOT %d: %s\n",
00130                aot, aac_get_error(err));
00131         goto error;
00132     }
00133 
00134     if (aot == FF_PROFILE_AAC_ELD + 1 && s->eld_sbr) {
00135         if ((err = aacEncoder_SetParam(s->handle, AACENC_SBR_MODE,
00136                                        1)) != AACENC_OK) {
00137             av_log(avctx, AV_LOG_ERROR, "Unable to enable SBR for ELD: %s\n",
00138                    aac_get_error(err));
00139             goto error;
00140         }
00141     }
00142 
00143     if ((err = aacEncoder_SetParam(s->handle, AACENC_SAMPLERATE,
00144                                    avctx->sample_rate)) != AACENC_OK) {
00145         av_log(avctx, AV_LOG_ERROR, "Unable to set the sample rate %d: %s\n",
00146                avctx->sample_rate, aac_get_error(err));
00147         goto error;
00148     }
00149 
00150     switch (avctx->channels) {
00151     case 1: mode = MODE_1;       sce = 1; cpe = 0; break;
00152     case 2: mode = MODE_2;       sce = 0; cpe = 1; break;
00153     case 3: mode = MODE_1_2;     sce = 1; cpe = 1; break;
00154     case 4: mode = MODE_1_2_1;   sce = 2; cpe = 1; break;
00155     case 5: mode = MODE_1_2_2;   sce = 1; cpe = 2; break;
00156     case 6: mode = MODE_1_2_2_1; sce = 2; cpe = 2; break;
00157     default:
00158         av_log(avctx, AV_LOG_ERROR,
00159                "Unsupported number of channels %d\n", avctx->channels);
00160         goto error;
00161     }
00162 
00163     if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELMODE,
00164                                    mode)) != AACENC_OK) {
00165         av_log(avctx, AV_LOG_ERROR,
00166                "Unable to set channel mode %d: %s\n", mode, aac_get_error(err));
00167         goto error;
00168     }
00169 
00170     if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELORDER,
00171                                    1)) != AACENC_OK) {
00172         av_log(avctx, AV_LOG_ERROR,
00173                "Unable to set wav channel order %d: %s\n",
00174                mode, aac_get_error(err));
00175         goto error;
00176     }
00177 
00178     if (avctx->flags & CODEC_FLAG_QSCALE || s->vbr) {
00179         int mode = s->vbr ? s->vbr : avctx->global_quality;
00180         if (mode <  1 || mode > 5) {
00181             av_log(avctx, AV_LOG_WARNING,
00182                    "VBR quality %d out of range, should be 1-5\n", mode);
00183             mode = av_clip(mode, 1, 5);
00184         }
00185         av_log(avctx, AV_LOG_WARNING,
00186                "Note, the VBR setting is unsupported and only works with "
00187                "some parameter combinations\n");
00188         if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATEMODE,
00189                                        mode)) != AACENC_OK) {
00190             av_log(avctx, AV_LOG_ERROR, "Unable to set the VBR bitrate mode %d: %s\n",
00191                    mode, aac_get_error(err));
00192             goto error;
00193         }
00194     } else {
00195         if (avctx->bit_rate <= 0) {
00196             if (avctx->profile == FF_PROFILE_AAC_HE_V2) {
00197                 sce = 1;
00198                 cpe = 0;
00199             }
00200             avctx->bit_rate = (96*sce + 128*cpe) * avctx->sample_rate / 44;
00201             if (avctx->profile == FF_PROFILE_AAC_HE ||
00202                 avctx->profile == FF_PROFILE_AAC_HE_V2 ||
00203                 s->eld_sbr)
00204                 avctx->bit_rate /= 2;
00205         }
00206         if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATE,
00207                                        avctx->bit_rate)) != AACENC_OK) {
00208             av_log(avctx, AV_LOG_ERROR, "Unable to set the bitrate %d: %s\n",
00209                    avctx->bit_rate, aac_get_error(err));
00210             goto error;
00211         }
00212     }
00213 
00214     /* Choose bitstream format - if global header is requested, use
00215      * raw access units, otherwise use ADTS. */
00216     if ((err = aacEncoder_SetParam(s->handle, AACENC_TRANSMUX,
00217                                    avctx->flags & CODEC_FLAG_GLOBAL_HEADER ? 0 : s->latm ? 10 : 2)) != AACENC_OK) {
00218         av_log(avctx, AV_LOG_ERROR, "Unable to set the transmux format: %s\n",
00219                aac_get_error(err));
00220         goto error;
00221     }
00222 
00223     if (s->latm && s->header_period) {
00224         if ((err = aacEncoder_SetParam(s->handle, AACENC_HEADER_PERIOD,
00225                                        s->header_period)) != AACENC_OK) {
00226              av_log(avctx, AV_LOG_ERROR, "Unable to set header period: %s\n",
00227                     aac_get_error(err));
00228              goto error;
00229         }
00230     }
00231 
00232     /* If no signaling mode is chosen, use explicit hierarchical signaling
00233      * if using mp4 mode (raw access units, with global header) and
00234      * implicit signaling if using ADTS. */
00235     if (s->signaling < 0)
00236         s->signaling = avctx->flags & CODEC_FLAG_GLOBAL_HEADER ? 2 : 0;
00237 
00238     if ((err = aacEncoder_SetParam(s->handle, AACENC_SIGNALING_MODE,
00239                                    s->signaling)) != AACENC_OK) {
00240         av_log(avctx, AV_LOG_ERROR, "Unable to set signaling mode %d: %s\n",
00241                s->signaling, aac_get_error(err));
00242         goto error;
00243     }
00244 
00245     if ((err = aacEncoder_SetParam(s->handle, AACENC_AFTERBURNER,
00246                                    s->afterburner)) != AACENC_OK) {
00247         av_log(avctx, AV_LOG_ERROR, "Unable to set afterburner to %d: %s\n",
00248                s->afterburner, aac_get_error(err));
00249         goto error;
00250     }
00251 
00252     if (avctx->cutoff > 0) {
00253         if (avctx->cutoff < (avctx->sample_rate + 255) >> 8) {
00254             av_log(avctx, AV_LOG_ERROR, "cutoff valid range is %d-20000\n",
00255                    (avctx->sample_rate + 255) >> 8);
00256             goto error;
00257         }
00258         if ((err = aacEncoder_SetParam(s->handle, AACENC_BANDWIDTH,
00259                                        avctx->cutoff)) != AACENC_OK) {
00260             av_log(avctx, AV_LOG_ERROR, "Unable to set the encoder bandwith to %d: %s\n",
00261                    avctx->cutoff, aac_get_error(err));
00262             goto error;
00263         }
00264     }
00265 
00266     if ((err = aacEncEncode(s->handle, NULL, NULL, NULL, NULL)) != AACENC_OK) {
00267         av_log(avctx, AV_LOG_ERROR, "Unable to initialize the encoder: %s\n",
00268                aac_get_error(err));
00269         return AVERROR(EINVAL);
00270     }
00271 
00272     if ((err = aacEncInfo(s->handle, &info)) != AACENC_OK) {
00273         av_log(avctx, AV_LOG_ERROR, "Unable to get encoder info: %s\n",
00274                aac_get_error(err));
00275         goto error;
00276     }
00277 
00278 #if FF_API_OLD_ENCODE_AUDIO
00279     avctx->coded_frame = avcodec_alloc_frame();
00280     if (!avctx->coded_frame) {
00281         ret = AVERROR(ENOMEM);
00282         goto error;
00283     }
00284 #endif
00285     avctx->frame_size = info.frameLength;
00286     avctx->delay      = info.encoderDelay;
00287     ff_af_queue_init(avctx, &s->afq);
00288 
00289     if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
00290         avctx->extradata_size = info.confSize;
00291         avctx->extradata      = av_mallocz(avctx->extradata_size +
00292                                            FF_INPUT_BUFFER_PADDING_SIZE);
00293         if (!avctx->extradata) {
00294             ret = AVERROR(ENOMEM);
00295             goto error;
00296         }
00297 
00298         memcpy(avctx->extradata, info.confBuf, info.confSize);
00299     }
00300     return 0;
00301 error:
00302     aac_encode_close(avctx);
00303     return ret;
00304 }
00305 
00306 static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
00307                             const AVFrame *frame, int *got_packet_ptr)
00308 {
00309     AACContext    *s        = avctx->priv_data;
00310     AACENC_BufDesc in_buf   = { 0 }, out_buf = { 0 };
00311     AACENC_InArgs  in_args  = { 0 };
00312     AACENC_OutArgs out_args = { 0 };
00313     int in_buffer_identifier = IN_AUDIO_DATA;
00314     int in_buffer_size, in_buffer_element_size;
00315     int out_buffer_identifier = OUT_BITSTREAM_DATA;
00316     int out_buffer_size, out_buffer_element_size;
00317     void *in_ptr, *out_ptr;
00318     int ret;
00319     AACENC_ERROR err;
00320 
00321     /* handle end-of-stream small frame and flushing */
00322     if (!frame) {
00323         in_args.numInSamples = -1;
00324     } else {
00325         in_ptr                   = frame->data[0];
00326         in_buffer_size           = 2 * avctx->channels * frame->nb_samples;
00327         in_buffer_element_size   = 2;
00328 
00329         in_args.numInSamples     = avctx->channels * frame->nb_samples;
00330         in_buf.numBufs           = 1;
00331         in_buf.bufs              = &in_ptr;
00332         in_buf.bufferIdentifiers = &in_buffer_identifier;
00333         in_buf.bufSizes          = &in_buffer_size;
00334         in_buf.bufElSizes        = &in_buffer_element_size;
00335 
00336         /* add current frame to the queue */
00337         if ((ret = ff_af_queue_add(&s->afq, frame) < 0))
00338             return ret;
00339     }
00340 
00341     /* The maximum packet size is 6144 bits aka 768 bytes per channel. */
00342     if ((ret = ff_alloc_packet2(avctx, avpkt, FFMAX(8192, 768 * avctx->channels))))
00343         return ret;
00344 
00345     out_ptr                   = avpkt->data;
00346     out_buffer_size           = avpkt->size;
00347     out_buffer_element_size   = 1;
00348     out_buf.numBufs           = 1;
00349     out_buf.bufs              = &out_ptr;
00350     out_buf.bufferIdentifiers = &out_buffer_identifier;
00351     out_buf.bufSizes          = &out_buffer_size;
00352     out_buf.bufElSizes        = &out_buffer_element_size;
00353 
00354     if ((err = aacEncEncode(s->handle, &in_buf, &out_buf, &in_args,
00355                             &out_args)) != AACENC_OK) {
00356         if (!frame && err == AACENC_ENCODE_EOF)
00357             return 0;
00358         av_log(avctx, AV_LOG_ERROR, "Unable to encode frame: %s\n",
00359                aac_get_error(err));
00360         return AVERROR(EINVAL);
00361     }
00362 
00363     if (!out_args.numOutBytes)
00364         return 0;
00365 
00366     /* Get the next frame pts & duration */
00367     ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
00368                        &avpkt->duration);
00369 
00370     avpkt->size     = out_args.numOutBytes;
00371     *got_packet_ptr = 1;
00372     return 0;
00373 }
00374 
00375 static const AVProfile profiles[] = {
00376     { FF_PROFILE_AAC_LOW,   "LC"       },
00377     { FF_PROFILE_AAC_HE,    "HE-AAC"   },
00378     { FF_PROFILE_AAC_HE_V2, "HE-AACv2" },
00379     { FF_PROFILE_AAC_LD,    "LD"       },
00380     { FF_PROFILE_AAC_ELD,   "ELD"      },
00381     { FF_PROFILE_UNKNOWN },
00382 };
00383 
00384 static const AVCodecDefault aac_encode_defaults[] = {
00385     { "b", "0" },
00386     { NULL }
00387 };
00388 
00389 static const uint64_t aac_channel_layout[] = {
00390     AV_CH_LAYOUT_MONO,
00391     AV_CH_LAYOUT_STEREO,
00392     AV_CH_LAYOUT_SURROUND,
00393     AV_CH_LAYOUT_4POINT0,
00394     AV_CH_LAYOUT_5POINT0_BACK,
00395     AV_CH_LAYOUT_5POINT1_BACK,
00396     0,
00397 };
00398 
00399 static const int aac_sample_rates[] = {
00400     96000, 88200, 64000, 48000, 44100, 32000,
00401     24000, 22050, 16000, 12000, 11025, 8000, 0
00402 };
00403 
00404 AVCodec ff_libfdk_aac_encoder = {
00405     .name                  = "libfdk_aac",
00406     .type                  = AVMEDIA_TYPE_AUDIO,
00407     .id                    = AV_CODEC_ID_AAC,
00408     .priv_data_size        = sizeof(AACContext),
00409     .init                  = aac_encode_init,
00410     .encode2               = aac_encode_frame,
00411     .close                 = aac_encode_close,
00412     .capabilities          = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY,
00413     .sample_fmts           = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
00414                                                             AV_SAMPLE_FMT_NONE },
00415     .long_name             = NULL_IF_CONFIG_SMALL("Fraunhofer FDK AAC"),
00416     .priv_class            = &aac_enc_class,
00417     .defaults              = aac_encode_defaults,
00418     .profiles              = profiles,
00419     .supported_samplerates = aac_sample_rates,
00420     .channel_layouts       = aac_channel_layout,
00421 };