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