src/lib/ffmpeg_file_encoder.cc

   1 /*
   2     Copyright (C) 2017-2018 Carl Hetherington <cth@carlh.net>
   3
   4     This file is part of DCP-o-matic.
   5
   6     DCP-o-matic is free software; you can redistribute it and/or modify
   7     it under the terms of the GNU General Public License as published by
   8     the Free Software Foundation; either version 2 of the License, or
   9     (at your option) any later version.
  10
  11     DCP-o-matic is distributed in the hope that it will be useful,
  12     but WITHOUT ANY WARRANTY; without even the implied warranty of
  13     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14     GNU General Public License for more details.
  15
  16     You should have received a copy of the GNU General Public License
  17     along with DCP-o-matic.  If not, see <http://www.gnu.org/licenses/>.
  18
  19 */
  20
  21 #include "ffmpeg_encoder.h"
  22 #include "film.h"
  23 #include "job.h"
  24 #include "player.h"
  25 #include "player_video.h"
  26 #include "log.h"
  27 #include "image.h"
  28 #include "cross.h"
  29 #include "compose.hpp"
  30 #include <iostream>
  31
  32 #include "i18n.h"
  33
  34 using std::string;
  35 using std::runtime_error;
  36 using std::cout;
  37 using std::pair;
  38 using boost::shared_ptr;
  39 using boost::bind;
  40 using boost::weak_ptr;
  41 using boost::optional;
  42 using namespace dcpomatic;
  43 #if BOOST_VERSION >= 106100
  44 using namespace boost::placeholders;
  45 #endif
  46
  47 int FFmpegFileEncoder::_video_stream_index = 0;
  48 int FFmpegFileEncoder::_audio_stream_index_base = 1;
  49
  50
  51 class ExportAudioStream : public boost::noncopyable
  52 {
  53 public:
  54         ExportAudioStream (string codec_name, int channels, int frame_rate, AVSampleFormat sample_format, AVFormatContext* format_context, int stream_index)
  55                 : _format_context (format_context)
  56                 , _stream_index (stream_index)
  57         {
  58                 _codec = avcodec_find_encoder_by_name (codec_name.c_str());
  59                 if (!_codec) {
  60                         throw runtime_error (String::compose("could not find FFmpeg encoder %1", codec_name));
  61                 }
  62
  63                 _codec_context = avcodec_alloc_context3 (_codec);
  64                 if (!_codec_context) {
  65                         throw runtime_error ("could not allocate FFmpeg audio context");
  66                 }
  67
  68                 avcodec_get_context_defaults3 (_codec_context, _codec);
  69
  70                 /* XXX: configurable */
  71                 _codec_context->bit_rate = channels * 128 * 1024;
  72                 _codec_context->sample_fmt = sample_format;
  73                 _codec_context->sample_rate = frame_rate;
  74                 _codec_context->channel_layout = av_get_default_channel_layout (channels);
  75                 _codec_context->channels = channels;
  76
  77                 _stream = avformat_new_stream (format_context, _codec);
  78                 if (!_stream) {
  79                         throw runtime_error ("could not create FFmpeg output audio stream");
  80                 }
  81
  82                 _stream->id = stream_index;
  83                 _stream->disposition |= AV_DISPOSITION_DEFAULT;
  84 DCPOMATIC_DISABLE_WARNINGS
  85                 _stream->codec = _codec_context;
  86 DCPOMATIC_ENABLE_WARNINGS
  87
  88                 int r = avcodec_open2 (_codec_context, _codec, 0);
  89                 if (r < 0) {
  90                         char buffer[256];
  91                         av_strerror (r, buffer, sizeof(buffer));
  92                         throw runtime_error (String::compose("could not open FFmpeg audio codec (%1)", buffer));
  93                 }
  94         }
  95
  96         ~ExportAudioStream ()
  97         {
  98                 avcodec_close (_codec_context);
  99         }
 100
 101         int frame_size () const {
 102                 return _codec_context->frame_size;
 103         }
 104
 105         bool flush ()
 106         {
 107                 AVPacket packet;
 108                 av_init_packet (&packet);
 109                 packet.data = 0;
 110                 packet.size = 0;
 111                 bool flushed = false;
 112
 113                 int got_packet;
 114 DCPOMATIC_DISABLE_WARNINGS
 115                 avcodec_encode_audio2 (_codec_context, &packet, 0, &got_packet);
 116 DCPOMATIC_ENABLE_WARNINGS
 117                 if (got_packet) {
 118                         packet.stream_index = 0;
 119                         av_interleaved_write_frame (_format_context, &packet);
 120                 } else {
 121                         flushed = true;
 122                 }
 123                 av_packet_unref (&packet);
 124                 return flushed;
 125         }
 126
 127         void write (int size, int channel_offset, int channels, float** data, int64_t sample_offset)
 128         {
 129                 DCPOMATIC_ASSERT (size);
 130
 131                 AVFrame* frame = av_frame_alloc ();
 132                 DCPOMATIC_ASSERT (frame);
 133
 134                 int const buffer_size = av_samples_get_buffer_size (0, channels, size, _codec_context->sample_fmt, 0);
 135                 DCPOMATIC_ASSERT (buffer_size >= 0);
 136
 137                 void* samples = av_malloc (buffer_size);
 138                 DCPOMATIC_ASSERT (samples);
 139
 140                 frame->nb_samples = size;
 141                 int r = avcodec_fill_audio_frame (frame, channels, _codec_context->sample_fmt, (const uint8_t *) samples, buffer_size, 0);
 142                 DCPOMATIC_ASSERT (r >= 0);
 143
 144                 switch (_codec_context->sample_fmt) {
 145                 case AV_SAMPLE_FMT_S16:
 146                 {
 147                         int16_t* q = reinterpret_cast<int16_t*> (samples);
 148                         for (int i = 0; i < size; ++i) {
 149                                 for (int j = 0; j < channels; ++j) {
 150                                         *q++ = data[j + channel_offset][i] * 32767;
 151                                 }
 152                         }
 153                         break;
 154                 }
 155                 case AV_SAMPLE_FMT_S32:
 156                 {
 157                         int32_t* q = reinterpret_cast<int32_t*> (samples);
 158                         for (int i = 0; i < size; ++i) {
 159                                 for (int j = 0; j < channels; ++j) {
 160                                         *q++ = data[j + channel_offset][i] * 2147483647;
 161                                 }
 162                         }
 163                         break;
 164                 }
 165                 case AV_SAMPLE_FMT_FLTP:
 166                 {
 167                         float* q = reinterpret_cast<float*> (samples);
 168                         for (int i = 0; i < channels; ++i) {
 169                                 memcpy (q, data[i + channel_offset], sizeof(float) * size);
 170                                 q += size;
 171                         }
 172                         break;
 173                 }
 174                 default:
 175                         DCPOMATIC_ASSERT (false);
 176                 }
 177
 178                 DCPOMATIC_ASSERT (_codec_context->time_base.num == 1);
 179                 frame->pts = sample_offset * _codec_context->time_base.den / _codec_context->sample_rate;
 180
 181                 AVPacket packet;
 182                 av_init_packet (&packet);
 183                 packet.data = 0;
 184                 packet.size = 0;
 185                 int got_packet;
 186
 187                 DCPOMATIC_DISABLE_WARNINGS
 188                 if (avcodec_encode_audio2 (_codec_context, &packet, frame, &got_packet) < 0) {
 189                         throw EncodeError ("FFmpeg audio encode failed");
 190                 }
 191                 DCPOMATIC_ENABLE_WARNINGS
 192
 193                 if (got_packet && packet.size) {
 194                         packet.stream_index = _stream_index;
 195                         av_interleaved_write_frame (_format_context, &packet);
 196                         av_packet_unref (&packet);
 197                 }
 198
 199                 av_free (samples);
 200                 av_frame_free (&frame);
 201         }
 202
 203 private:
 204         AVFormatContext* _format_context;
 205         AVCodec* _codec;
 206         AVCodecContext* _codec_context;
 207         AVStream* _stream;
 208         int _stream_index;
 209 };
 210
 211
 212 FFmpegFileEncoder::FFmpegFileEncoder (
 213         dcp::Size video_frame_size,
 214         int video_frame_rate,
 215         int audio_frame_rate,
 216         int channels,
 217         ExportFormat format,
 218         bool audio_stream_per_channel,
 219         int x264_crf,
 220         boost::filesystem::path output
 221 #ifdef DCPOMATIC_VARIANT_SWAROOP
 222         , optional<dcp::Key> key
 223         , optional<string> id
 224 #endif
 225         )
 226         : _audio_stream_per_channel (audio_stream_per_channel)
 227         , _video_options (0)
 228         , _audio_channels (channels)
 229         , _output (output)
 230         , _video_frame_size (video_frame_size)
 231         , _video_frame_rate (video_frame_rate)
 232         , _audio_frame_rate (audio_frame_rate)
 233         , _audio_frames (0)
 234 {
 235         _pixel_format = pixel_format (format);
 236
 237         switch (format) {
 238         case EXPORT_FORMAT_PRORES:
 239                 _sample_format = AV_SAMPLE_FMT_S16;
 240                 _video_codec_name = "prores_ks";
 241                 _audio_codec_name = "pcm_s16le";
 242                 av_dict_set (&_video_options, "profile", "3", 0);
 243                 av_dict_set (&_video_options, "threads", "auto", 0);
 244                 break;
 245         case EXPORT_FORMAT_H264_AAC:
 246                 _sample_format = AV_SAMPLE_FMT_FLTP;
 247                 _video_codec_name = "libx264";
 248                 _audio_codec_name = "aac";
 249                 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
 250                 break;
 251         case EXPORT_FORMAT_H264_PCM:
 252                 _sample_format = AV_SAMPLE_FMT_S32;
 253                 _video_codec_name = "libx264";
 254                 _audio_codec_name = "pcm_s24le";
 255                 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
 256                 break;
 257         default:
 258                 DCPOMATIC_ASSERT (false);
 259         }
 260
 261 #ifdef DCPOMATIC_VARIANT_SWAROOP
 262         int r = avformat_alloc_output_context2 (&_format_context, av_guess_format("mov", 0, 0), 0, 0);
 263 #else
 264         int r = avformat_alloc_output_context2 (&_format_context, 0, 0, _output.string().c_str());
 265 #endif
 266         if (!_format_context) {
 267                 throw runtime_error (String::compose("could not allocate FFmpeg format context (%1)", r));
 268         }
 269
 270         setup_video ();
 271         setup_audio ();
 272
 273         r = avio_open_boost (&_format_context->pb, _output, AVIO_FLAG_WRITE);
 274         if (r < 0) {
 275                 throw runtime_error (String::compose("could not open FFmpeg output file %1 (%2)", _output.string(), r));
 276         }
 277
 278         AVDictionary* options = 0;
 279
 280 #ifdef DCPOMATIC_VARIANT_SWAROOP
 281         if (key) {
 282                 av_dict_set (&options, "encryption_key", key->hex().c_str(), 0);
 283                 /* XXX: is this OK? */
 284                 av_dict_set (&options, "encryption_kid", "00000000000000000000000000000000", 0);
 285                 av_dict_set (&options, "encryption_scheme", "cenc-aes-ctr", 0);
 286         }
 287
 288         if (id) {
 289                 if (av_dict_set(&_format_context->metadata, SWAROOP_ID_TAG, id->c_str(), 0) < 0) {
 290                         throw runtime_error ("Could not write ID to output");
 291                 }
 292         }
 293 #endif
 294
 295         if (avformat_write_header (_format_context, &options) < 0) {
 296                 throw runtime_error ("could not write header to FFmpeg output file");
 297         }
 298
 299         _pending_audio.reset (new AudioBuffers(channels, 0));
 300 }
 301
 302
 303 FFmpegFileEncoder::~FFmpegFileEncoder ()
 304 {
 305         _audio_streams.clear ();
 306         avcodec_close (_video_codec_context);
 307         avformat_free_context (_format_context);
 308 }
 309
 310
 311 AVPixelFormat
 312 FFmpegFileEncoder::pixel_format (ExportFormat format)
 313 {
 314         switch (format) {
 315         case EXPORT_FORMAT_PRORES:
 316                 return AV_PIX_FMT_YUV422P10;
 317         case EXPORT_FORMAT_H264_AAC:
 318         case EXPORT_FORMAT_H264_PCM:
 319                 return AV_PIX_FMT_YUV420P;
 320         default:
 321                 DCPOMATIC_ASSERT (false);
 322         }
 323
 324         return AV_PIX_FMT_YUV422P10;
 325 }
 326
 327 void
 328 FFmpegFileEncoder::setup_video ()
 329 {
 330         _video_codec = avcodec_find_encoder_by_name (_video_codec_name.c_str());
 331         if (!_video_codec) {
 332                 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _video_codec_name));
 333         }
 334
 335         _video_codec_context = avcodec_alloc_context3 (_video_codec);
 336         if (!_video_codec_context) {
 337                 throw runtime_error ("could not allocate FFmpeg video context");
 338         }
 339
 340         avcodec_get_context_defaults3 (_video_codec_context, _video_codec);
 341
 342         /* Variable quantisation */
 343         _video_codec_context->global_quality = 0;
 344         _video_codec_context->width = _video_frame_size.width;
 345         _video_codec_context->height = _video_frame_size.height;
 346         _video_codec_context->time_base = (AVRational) { 1, _video_frame_rate };
 347         _video_codec_context->pix_fmt = _pixel_format;
 348         _video_codec_context->flags |= AV_CODEC_FLAG_QSCALE | AV_CODEC_FLAG_GLOBAL_HEADER;
 349
 350         _video_stream = avformat_new_stream (_format_context, _video_codec);
 351         if (!_video_stream) {
 352                 throw runtime_error ("could not create FFmpeg output video stream");
 353         }
 354
 355 DCPOMATIC_DISABLE_WARNINGS
 356         _video_stream->id = _video_stream_index;
 357         _video_stream->codec = _video_codec_context;
 358 DCPOMATIC_ENABLE_WARNINGS
 359
 360         if (avcodec_open2 (_video_codec_context, _video_codec, &_video_options) < 0) {
 361                 throw runtime_error ("could not open FFmpeg video codec");
 362         }
 363 }
 364
 365
 366 void
 367 FFmpegFileEncoder::setup_audio ()
 368 {
 369         int const streams = _audio_stream_per_channel ? _audio_channels : 1;
 370         int const channels_per_stream = _audio_stream_per_channel ? 1 : _audio_channels;
 371
 372         for (int i = 0; i < streams; ++i) {
 373                 _audio_streams.push_back(
 374                         shared_ptr<ExportAudioStream>(
 375                                 new ExportAudioStream(_audio_codec_name, channels_per_stream, _audio_frame_rate, _sample_format, _format_context, _audio_stream_index_base + i)
 376                                 )
 377                         );
 378         }
 379 }
 380
 381
 382 void
 383 FFmpegFileEncoder::flush ()
 384 {
 385         if (_pending_audio->frames() > 0) {
 386                 audio_frame (_pending_audio->frames ());
 387         }
 388
 389         bool flushed_video = false;
 390         bool flushed_audio = false;
 391
 392         while (!flushed_video || !flushed_audio) {
 393                 AVPacket packet;
 394                 av_init_packet (&packet);
 395                 packet.data = 0;
 396                 packet.size = 0;
 397
 398                 int got_packet;
 399 DCPOMATIC_DISABLE_WARNINGS
 400                 avcodec_encode_video2 (_video_codec_context, &packet, 0, &got_packet);
 401 DCPOMATIC_ENABLE_WARNINGS
 402                 if (got_packet) {
 403                         packet.stream_index = 0;
 404                         av_interleaved_write_frame (_format_context, &packet);
 405                 } else {
 406                         flushed_video = true;
 407                 }
 408                 av_packet_unref (&packet);
 409
 410                 flushed_audio = true;
 411                 BOOST_FOREACH (shared_ptr<ExportAudioStream> i, _audio_streams) {
 412                         if (!i->flush()) {
 413                                 flushed_audio = false;
 414                         }
 415                 }
 416         }
 417
 418         av_write_trailer (_format_context);
 419 }
 420
 421 void
 422 FFmpegFileEncoder::video (shared_ptr<PlayerVideo> video, DCPTime time)
 423 {
 424         /* All our output formats are video range at the moment */
 425         shared_ptr<Image> image = video->image (
 426                 bind (&PlayerVideo::force, _1, _pixel_format),
 427                 VIDEO_RANGE_VIDEO,
 428                 true,
 429                 false
 430                 );
 431
 432         AVFrame* frame = av_frame_alloc ();
 433         DCPOMATIC_ASSERT (frame);
 434
 435         {
 436                 boost::mutex::scoped_lock lm (_pending_images_mutex);
 437                 _pending_images[image->data()[0]] = image;
 438         }
 439
 440         for (int i = 0; i < 3; ++i) {
 441                 AVBufferRef* buffer = av_buffer_create(image->data()[i], image->stride()[i] * image->size().height, &buffer_free, this, 0);
 442                 frame->buf[i] = av_buffer_ref (buffer);
 443                 frame->data[i] = buffer->data;
 444                 frame->linesize[i] = image->stride()[i];
 445                 av_buffer_unref (&buffer);
 446         }
 447
 448         frame->width = image->size().width;
 449         frame->height = image->size().height;
 450         frame->format = _pixel_format;
 451         DCPOMATIC_ASSERT (_video_stream->time_base.num == 1);
 452         frame->pts = time.get() * _video_stream->time_base.den / DCPTime::HZ;
 453
 454         AVPacket packet;
 455         av_init_packet (&packet);
 456         packet.data = 0;
 457         packet.size = 0;
 458
 459         int got_packet;
 460 DCPOMATIC_DISABLE_WARNINGS
 461         if (avcodec_encode_video2 (_video_codec_context, &packet, frame, &got_packet) < 0) {
 462                 throw EncodeError ("FFmpeg video encode failed");
 463         }
 464 DCPOMATIC_ENABLE_WARNINGS
 465
 466         if (got_packet && packet.size) {
 467                 packet.stream_index = _video_stream_index;
 468                 av_interleaved_write_frame (_format_context, &packet);
 469                 av_packet_unref (&packet);
 470         }
 471
 472         av_frame_free (&frame);
 473
 474 }
 475
 476 /** Called when the player gives us some audio */
 477 void
 478 FFmpegFileEncoder::audio (shared_ptr<AudioBuffers> audio)
 479 {
 480         _pending_audio->append (audio);
 481
 482         DCPOMATIC_ASSERT (!_audio_streams.empty());
 483         int frame_size = _audio_streams[0]->frame_size();
 484         if (frame_size == 0) {
 485                 /* codec has AV_CODEC_CAP_VARIABLE_FRAME_SIZE */
 486                 frame_size = _audio_frame_rate / _video_frame_rate;
 487         }
 488
 489         while (_pending_audio->frames() >= frame_size) {
 490                 audio_frame (frame_size);
 491         }
 492 }
 493
 494 void
 495 FFmpegFileEncoder::audio_frame (int size)
 496 {
 497         if (_audio_stream_per_channel) {
 498                 int offset = 0;
 499                 BOOST_FOREACH (shared_ptr<ExportAudioStream> i, _audio_streams) {
 500                         i->write (size, offset, 1, _pending_audio->data(), _audio_frames);
 501                         ++offset;
 502                 }
 503         } else {
 504                 DCPOMATIC_ASSERT (!_audio_streams.empty());
 505                 DCPOMATIC_ASSERT (_pending_audio->channels());
 506                 _audio_streams[0]->write (size, 0, _pending_audio->channels(), _pending_audio->data(), _audio_frames);
 507         }
 508
 509         _pending_audio->trim_start (size);
 510         _audio_frames += size;
 511 }
 512
 513 void
 514 FFmpegFileEncoder::subtitle (PlayerText, DCPTimePeriod)
 515 {
 516
 517 }
 518
 519 void
 520 FFmpegFileEncoder::buffer_free (void* opaque, uint8_t* data)
 521 {
 522         reinterpret_cast<FFmpegFileEncoder*>(opaque)->buffer_free2(data);
 523 }
 524
 525 void
 526 FFmpegFileEncoder::buffer_free2 (uint8_t* data)
 527 {
 528         boost::mutex::scoped_lock lm (_pending_images_mutex);
 529         if (_pending_images.find(data) != _pending_images.end()) {
 530                 _pending_images.erase (data);
 531         }
 532 }