FFmpeg: libavcodec/libopusenc.c Source File

00001 /*
00002  * Opus encoder using libopus
00003  * Copyright (c) 2012 Nathan Caldwell
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 <opus.h>
00023 #include <opus_multistream.h>
00024 
00025 #include "libavutil/opt.h"
00026 #include "avcodec.h"
00027 #include "bytestream.h"
00028 #include "internal.h"
00029 #include "libopus.h"
00030 #include "vorbis.h"
00031 #include "audio_frame_queue.h"
00032 
00033 typedef struct LibopusEncOpts {
00034     int vbr;
00035     int application;
00036     int packet_loss;
00037     int complexity;
00038     float frame_duration;
00039     int packet_size;
00040     int max_bandwidth;
00041 } LibopusEncOpts;
00042 
00043 typedef struct LibopusEncContext {
00044     AVClass *class;
00045     OpusMSEncoder *enc;
00046     int stream_count;
00047     uint8_t *samples;
00048     LibopusEncOpts opts;
00049     AudioFrameQueue afq;
00050 } LibopusEncContext;
00051 
00052 static const uint8_t opus_coupled_streams[8] = {
00053     0, 1, 1, 2, 2, 2, 2, 3
00054 };
00055 
00056 /* Opus internal to Vorbis channel order mapping written in the header */
00057 static const uint8_t opus_vorbis_channel_map[8][8] = {
00058     { 0 },
00059     { 0, 1 },
00060     { 0, 2, 1 },
00061     { 0, 1, 2, 3 },
00062     { 0, 4, 1, 2, 3 },
00063     { 0, 4, 1, 2, 3, 5 },
00064     { 0, 4, 1, 2, 3, 5, 6 },
00065     { 0, 6, 1, 2, 3, 4, 5, 7 },
00066 };
00067 
00068 /* libavcodec to libopus channel order mapping, passed to libopus */
00069 static const uint8_t libavcodec_libopus_channel_map[8][8] = {
00070     { 0 },
00071     { 0, 1 },
00072     { 0, 1, 2 },
00073     { 0, 1, 2, 3 },
00074     { 0, 1, 3, 4, 2 },
00075     { 0, 1, 4, 5, 2, 3 },
00076     { 0, 1, 5, 6, 2, 4, 3 },
00077     { 0, 1, 6, 7, 4, 5, 2, 3 },
00078 };
00079 
00080 static void libopus_write_header(AVCodecContext *avctx, int stream_count,
00081                                  int coupled_stream_count,
00082                                  const uint8_t *channel_mapping)
00083 {
00084     uint8_t *p   = avctx->extradata;
00085     int channels = avctx->channels;
00086 
00087     bytestream_put_buffer(&p, "OpusHead", 8);
00088     bytestream_put_byte(&p, 1); /* Version */
00089     bytestream_put_byte(&p, channels);
00090     bytestream_put_le16(&p, avctx->delay); /* Lookahead samples at 48kHz */
00091     bytestream_put_le32(&p, avctx->sample_rate); /* Original sample rate */
00092     bytestream_put_le16(&p, 0); /* Gain of 0dB is recommended. */
00093 
00094     /* Channel mapping */
00095     if (channels > 2) {
00096         bytestream_put_byte(&p, channels <= 8 ? 1 : 255);
00097         bytestream_put_byte(&p, stream_count);
00098         bytestream_put_byte(&p, coupled_stream_count);
00099         bytestream_put_buffer(&p, channel_mapping, channels);
00100     } else {
00101         bytestream_put_byte(&p, 0);
00102     }
00103 }
00104 
00105 static int libopus_configure_encoder(AVCodecContext *avctx, OpusMSEncoder *enc,
00106                                      LibopusEncOpts *opts)
00107 {
00108     int ret;
00109 
00110     ret = opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(avctx->bit_rate));
00111     if (ret != OPUS_OK) {
00112         av_log(avctx, AV_LOG_ERROR,
00113                "Failed to set bitrate: %s\n", opus_strerror(ret));
00114         return ret;
00115     }
00116 
00117     ret = opus_multistream_encoder_ctl(enc,
00118                                        OPUS_SET_COMPLEXITY(opts->complexity));
00119     if (ret != OPUS_OK)
00120         av_log(avctx, AV_LOG_WARNING,
00121                "Unable to set complexity: %s\n", opus_strerror(ret));
00122 
00123     ret = opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(!!opts->vbr));
00124     if (ret != OPUS_OK)
00125         av_log(avctx, AV_LOG_WARNING,
00126                "Unable to set VBR: %s\n", opus_strerror(ret));
00127 
00128     ret = opus_multistream_encoder_ctl(enc,
00129                                        OPUS_SET_VBR_CONSTRAINT(opts->vbr == 2));
00130     if (ret != OPUS_OK)
00131         av_log(avctx, AV_LOG_WARNING,
00132                "Unable to set constrained VBR: %s\n", opus_strerror(ret));
00133 
00134     ret = opus_multistream_encoder_ctl(enc,
00135                                        OPUS_SET_PACKET_LOSS_PERC(opts->packet_loss));
00136     if (ret != OPUS_OK)
00137         av_log(avctx, AV_LOG_WARNING,
00138                "Unable to set expected packet loss percentage: %s\n",
00139                opus_strerror(ret));
00140 
00141     if (avctx->cutoff) {
00142         ret = opus_multistream_encoder_ctl(enc,
00143                                            OPUS_SET_MAX_BANDWIDTH(opts->max_bandwidth));
00144         if (ret != OPUS_OK)
00145             av_log(avctx, AV_LOG_WARNING,
00146                    "Unable to set maximum bandwidth: %s\n", opus_strerror(ret));
00147     }
00148 
00149     return OPUS_OK;
00150 }
00151 
00152 static int av_cold libopus_encode_init(AVCodecContext *avctx)
00153 {
00154     LibopusEncContext *opus = avctx->priv_data;
00155     const uint8_t *channel_mapping;
00156     OpusMSEncoder *enc;
00157     int ret = OPUS_OK;
00158     int coupled_stream_count, header_size, frame_size;
00159 
00160     coupled_stream_count = opus_coupled_streams[avctx->channels - 1];
00161     opus->stream_count   = avctx->channels - coupled_stream_count;
00162     channel_mapping      = libavcodec_libopus_channel_map[avctx->channels - 1];
00163 
00164     /* FIXME: Opus can handle up to 255 channels. However, the mapping for
00165      * anything greater than 8 is undefined. */
00166     if (avctx->channels > 8)
00167         av_log(avctx, AV_LOG_WARNING,
00168                "Channel layout undefined for %d channels.\n", avctx->channels);
00169 
00170     if (!avctx->bit_rate) {
00171         /* Sane default copied from opusenc */
00172         avctx->bit_rate = 64000 * opus->stream_count +
00173                           32000 * coupled_stream_count;
00174         av_log(avctx, AV_LOG_WARNING,
00175                "No bit rate set. Defaulting to %d bps.\n", avctx->bit_rate);
00176     }
00177 
00178     if (avctx->bit_rate < 500 || avctx->bit_rate > 256000 * avctx->channels) {
00179         av_log(avctx, AV_LOG_ERROR, "The bit rate %d bps is unsupported. "
00180                "Please choose a value between 500 and %d.\n", avctx->bit_rate,
00181                256000 * avctx->channels);
00182         return AVERROR(EINVAL);
00183     }
00184 
00185     frame_size = opus->opts.frame_duration * 48000 / 1000;
00186     switch (frame_size) {
00187     case 120:
00188     case 240:
00189         if (opus->opts.application != OPUS_APPLICATION_RESTRICTED_LOWDELAY)
00190             av_log(avctx, AV_LOG_WARNING,
00191                    "LPC mode cannot be used with a frame duration of less "
00192                    "than 10ms. Enabling restricted low-delay mode.\n"
00193                    "Use a longer frame duration if this is not what you want.\n");
00194         /* Frame sizes less than 10 ms can only use MDCT mode, so switching to
00195          * RESTRICTED_LOWDELAY avoids an unnecessary extra 2.5ms lookahead. */
00196         opus->opts.application = OPUS_APPLICATION_RESTRICTED_LOWDELAY;
00197     case 480:
00198     case 960:
00199     case 1920:
00200     case 2880:
00201         opus->opts.packet_size =
00202         avctx->frame_size      = frame_size * avctx->sample_rate / 48000;
00203         break;
00204     default:
00205         av_log(avctx, AV_LOG_ERROR, "Invalid frame duration: %g.\n"
00206                "Frame duration must be exactly one of: 2.5, 5, 10, 20, 40 or 60.\n",
00207                opus->opts.frame_duration);
00208         return AVERROR(EINVAL);
00209     }
00210 
00211     if (avctx->compression_level < 0 || avctx->compression_level > 10) {
00212         av_log(avctx, AV_LOG_WARNING,
00213                "Compression level must be in the range 0 to 10. "
00214                "Defaulting to 10.\n");
00215         opus->opts.complexity = 10;
00216     } else {
00217         opus->opts.complexity = avctx->compression_level;
00218     }
00219 
00220     if (avctx->cutoff) {
00221         switch (avctx->cutoff) {
00222         case  4000:
00223             opus->opts.max_bandwidth = OPUS_BANDWIDTH_NARROWBAND;
00224             break;
00225         case  6000:
00226             opus->opts.max_bandwidth = OPUS_BANDWIDTH_MEDIUMBAND;
00227             break;
00228         case  8000:
00229             opus->opts.max_bandwidth = OPUS_BANDWIDTH_WIDEBAND;
00230             break;
00231         case 12000:
00232             opus->opts.max_bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND;
00233             break;
00234         case 20000:
00235             opus->opts.max_bandwidth = OPUS_BANDWIDTH_FULLBAND;
00236             break;
00237         default:
00238             av_log(avctx, AV_LOG_WARNING,
00239                    "Invalid frequency cutoff: %d. Using default maximum bandwidth.\n"
00240                    "Cutoff frequency must be exactly one of: 4000, 6000, 8000, 12000 or 20000.\n",
00241                    avctx->cutoff);
00242             avctx->cutoff = 0;
00243         }
00244     }
00245 
00246     enc = opus_multistream_encoder_create(avctx->sample_rate, avctx->channels,
00247                                           opus->stream_count,
00248                                           coupled_stream_count,
00249                                           channel_mapping,
00250                                           opus->opts.application, &ret);
00251     if (ret != OPUS_OK) {
00252         av_log(avctx, AV_LOG_ERROR,
00253                "Failed to create encoder: %s\n", opus_strerror(ret));
00254         return ff_opus_error_to_averror(ret);
00255     }
00256 
00257     ret = libopus_configure_encoder(avctx, enc, &opus->opts);
00258     if (ret != OPUS_OK) {
00259         ret = ff_opus_error_to_averror(ret);
00260         goto fail;
00261     }
00262 
00263     header_size = 19 + (avctx->channels > 2 ? 2 + avctx->channels : 0);
00264     avctx->extradata = av_malloc(header_size + FF_INPUT_BUFFER_PADDING_SIZE);
00265     if (!avctx->extradata) {
00266         av_log(avctx, AV_LOG_ERROR, "Failed to allocate extradata.\n");
00267         ret = AVERROR(ENOMEM);
00268         goto fail;
00269     }
00270     avctx->extradata_size = header_size;
00271 
00272     opus->samples = av_mallocz(frame_size * avctx->channels *
00273                                av_get_bytes_per_sample(avctx->sample_fmt));
00274     if (!opus->samples) {
00275         av_log(avctx, AV_LOG_ERROR, "Failed to allocate samples buffer.\n");
00276         ret = AVERROR(ENOMEM);
00277         goto fail;
00278     }
00279 
00280     ret = opus_multistream_encoder_ctl(enc, OPUS_GET_LOOKAHEAD(&avctx->delay));
00281     if (ret != OPUS_OK)
00282         av_log(avctx, AV_LOG_WARNING,
00283                "Unable to get number of lookahead samples: %s\n",
00284                opus_strerror(ret));
00285 
00286     libopus_write_header(avctx, opus->stream_count, coupled_stream_count,
00287                          opus_vorbis_channel_map[avctx->channels - 1]);
00288 
00289     ff_af_queue_init(avctx, &opus->afq);
00290 
00291     opus->enc = enc;
00292 
00293     return 0;
00294 
00295 fail:
00296     opus_multistream_encoder_destroy(enc);
00297     av_freep(&avctx->extradata);
00298     return ret;
00299 }
00300 
00301 static int libopus_encode(AVCodecContext *avctx, AVPacket *avpkt,
00302                           const AVFrame *frame, int *got_packet_ptr)
00303 {
00304     LibopusEncContext *opus = avctx->priv_data;
00305     const int sample_size   = avctx->channels *
00306                               av_get_bytes_per_sample(avctx->sample_fmt);
00307     uint8_t *audio;
00308     int ret;
00309 
00310     if (frame) {
00311         ff_af_queue_add(&opus->afq, frame);
00312         if (frame->nb_samples < opus->opts.packet_size) {
00313             audio = opus->samples;
00314             memcpy(audio, frame->data[0], frame->nb_samples * sample_size);
00315         } else
00316             audio = frame->data[0];
00317     } else {
00318         if (!opus->afq.remaining_samples)
00319             return 0;
00320         audio = opus->samples;
00321         memset(audio, 0, opus->opts.packet_size * sample_size);
00322     }
00323 
00324     /* Maximum packet size taken from opusenc in opus-tools. 60ms packets
00325      * consist of 3 frames in one packet. The maximum frame size is 1275
00326      * bytes along with the largest possible packet header of 7 bytes. */
00327     if (ret = ff_alloc_packet(avpkt, (1275 * 3 + 7) * opus->stream_count)) {
00328         av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
00329         return ret;
00330     }
00331 
00332     if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT)
00333         ret = opus_multistream_encode_float(opus->enc, (float *)audio,
00334                                             opus->opts.packet_size,
00335                                             avpkt->data, avpkt->size);
00336     else
00337         ret = opus_multistream_encode(opus->enc, (opus_int16 *)audio,
00338                                       opus->opts.packet_size,
00339                                       avpkt->data, avpkt->size);
00340 
00341     if (ret < 0) {
00342         av_log(avctx, AV_LOG_ERROR,
00343                "Error encoding frame: %s\n", opus_strerror(ret));
00344         return ff_opus_error_to_averror(ret);
00345     }
00346 
00347     av_shrink_packet(avpkt, ret);
00348 
00349     ff_af_queue_remove(&opus->afq, opus->opts.packet_size,
00350                        &avpkt->pts, &avpkt->duration);
00351 
00352     *got_packet_ptr = 1;
00353 
00354     return 0;
00355 }
00356 
00357 static int av_cold libopus_encode_close(AVCodecContext *avctx)
00358 {
00359     LibopusEncContext *opus = avctx->priv_data;
00360 
00361     opus_multistream_encoder_destroy(opus->enc);
00362 
00363     ff_af_queue_close(&opus->afq);
00364 
00365     av_freep(&opus->samples);
00366     av_freep(&avctx->extradata);
00367 
00368     return 0;
00369 }
00370 
00371 #define OFFSET(x) offsetof(LibopusEncContext, opts.x)
00372 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
00373 static const AVOption libopus_options[] = {
00374     { "application",    "Intended application type",           OFFSET(application),    AV_OPT_TYPE_INT,   { .i64 = OPUS_APPLICATION_AUDIO }, OPUS_APPLICATION_VOIP, OPUS_APPLICATION_RESTRICTED_LOWDELAY, FLAGS, "application" },
00375         { "voip",           "Favor improved speech intelligibility",   0, AV_OPT_TYPE_CONST, { .i64 = OPUS_APPLICATION_VOIP },                0, 0, FLAGS, "application" },
00376         { "audio",          "Favor faithfulness to the input",         0, AV_OPT_TYPE_CONST, { .i64 = OPUS_APPLICATION_AUDIO },               0, 0, FLAGS, "application" },
00377         { "lowdelay",       "Restrict to only the lowest delay modes", 0, AV_OPT_TYPE_CONST, { .i64 = OPUS_APPLICATION_RESTRICTED_LOWDELAY }, 0, 0, FLAGS, "application" },
00378     { "frame_duration", "Duration of a frame in milliseconds", OFFSET(frame_duration), AV_OPT_TYPE_FLOAT, { .dbl = 10.0 }, 2.5, 60.0, FLAGS },
00379     { "packet_loss",    "Expected packet loss percentage",     OFFSET(packet_loss),    AV_OPT_TYPE_INT,   { .i64 = 0 },    0,   100,  FLAGS },
00380     { "vbr",            "Variable bit rate mode",              OFFSET(vbr),            AV_OPT_TYPE_INT,   { .i64 = 1 },    0,   2,    FLAGS, "vbr" },
00381         { "off",            "Use constant bit rate", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, 0, 0, FLAGS, "vbr" },
00382         { "on",             "Use variable bit rate", 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, 0, 0, FLAGS, "vbr" },
00383         { "constrained",    "Use constrained VBR",   0, AV_OPT_TYPE_CONST, { .i64 = 2 }, 0, 0, FLAGS, "vbr" },
00384     { NULL },
00385 };
00386 
00387 static const AVClass libopus_class = {
00388     .class_name = "libopus",
00389     .item_name  = av_default_item_name,
00390     .option     = libopus_options,
00391     .version    = LIBAVUTIL_VERSION_INT,
00392 };
00393 
00394 static const AVCodecDefault libopus_defaults[] = {
00395     { "b",                 "0" },
00396     { "compression_level", "10" },
00397     { NULL },
00398 };
00399 
00400 static const int libopus_sample_rates[] = {
00401     48000, 24000, 16000, 12000, 8000, 0,
00402 };
00403 
00404 AVCodec ff_libopus_encoder = {
00405     .name            = "libopus",
00406     .type            = AVMEDIA_TYPE_AUDIO,
00407     .id              = AV_CODEC_ID_OPUS,
00408     .priv_data_size  = sizeof(LibopusEncContext),
00409     .init            = libopus_encode_init,
00410     .encode2         = libopus_encode,
00411     .close           = libopus_encode_close,
00412     .capabilities    = CODEC_CAP_DELAY | CODEC_CAP_SMALL_LAST_FRAME,
00413     .sample_fmts     = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
00414                                                       AV_SAMPLE_FMT_FLT,
00415                                                       AV_SAMPLE_FMT_NONE },
00416     .channel_layouts = ff_vorbis_channel_layouts,
00417     .supported_samplerates = libopus_sample_rates,
00418     .long_name       = NULL_IF_CONFIG_SMALL("libopus Opus"),
00419     .priv_class      = &libopus_class,
00420     .defaults        = libopus_defaults,
00421 };