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 DCPOMATIC_DISABLE_WARNINGS
  84                 _stream->codec = _codec_context;
  85 DCPOMATIC_ENABLE_WARNINGS
  86
  87                 int r = avcodec_open2 (_codec_context, _codec, 0);
  88                 if (r < 0) {
  89                         char buffer[256];
  90                         av_strerror (r, buffer, sizeof(buffer));
  91                         throw runtime_error (String::compose("could not open FFmpeg audio codec (%1)", buffer));
  92                 }
  93         }
  94
  95         ~ExportAudioStream ()
  96         {
  97                 avcodec_close (_codec_context);
  98         }
  99
 100         int frame_size () const {
 101                 return _codec_context->frame_size;
 102         }
 103
 104         bool flush ()
 105         {
 106                 AVPacket packet;
 107                 av_init_packet (&packet);
 108                 packet.data = 0;
 109                 packet.size = 0;
 110                 bool flushed = false;
 111
 112                 int got_packet;
 113 DCPOMATIC_DISABLE_WARNINGS
 114                 avcodec_encode_audio2 (_codec_context, &packet, 0, &got_packet);
 115 DCPOMATIC_ENABLE_WARNINGS
 116                 if (got_packet) {
 117                         packet.stream_index = 0;
 118                         av_interleaved_write_frame (_format_context, &packet);
 119                 } else {
 120                         flushed = true;
 121                 }
 122                 av_packet_unref (&packet);
 123                 return flushed;
 124         }
 125
 126         void write (int size, int channel_offset, int channels, float** data, int64_t sample_offset)
 127         {
 128                 DCPOMATIC_ASSERT (size);
 129
 130                 AVFrame* frame = av_frame_alloc ();
 131                 DCPOMATIC_ASSERT (frame);
 132
 133                 int const buffer_size = av_samples_get_buffer_size (0, channels, size, _codec_context->sample_fmt, 0);
 134                 DCPOMATIC_ASSERT (buffer_size >= 0);
 135
 136                 void* samples = av_malloc (buffer_size);
 137                 DCPOMATIC_ASSERT (samples);
 138
 139                 frame->nb_samples = size;
 140                 int r = avcodec_fill_audio_frame (frame, channels, _codec_context->sample_fmt, (const uint8_t *) samples, buffer_size, 0);
 141                 DCPOMATIC_ASSERT (r >= 0);
 142
 143                 switch (_codec_context->sample_fmt) {
 144                 case AV_SAMPLE_FMT_S16:
 145                 {
 146                         int16_t* q = reinterpret_cast<int16_t*> (samples);
 147                         for (int i = 0; i < size; ++i) {
 148                                 for (int j = 0; j < channels; ++j) {
 149                                         *q++ = data[j + channel_offset][i] * 32767;
 150                                 }
 151                         }
 152                         break;
 153                 }
 154                 case AV_SAMPLE_FMT_S32:
 155                 {
 156                         int32_t* q = reinterpret_cast<int32_t*> (samples);
 157                         for (int i = 0; i < size; ++i) {
 158                                 for (int j = 0; j < channels; ++j) {
 159                                         *q++ = data[j + channel_offset][i] * 2147483647;
 160                                 }
 161                         }
 162                         break;
 163                 }
 164                 case AV_SAMPLE_FMT_FLTP:
 165                 {
 166                         float* q = reinterpret_cast<float*> (samples);
 167                         for (int i = 0; i < channels; ++i) {
 168                                 memcpy (q, data[i + channel_offset], sizeof(float) * size);
 169                                 q += size;
 170                         }
 171                         break;
 172                 }
 173                 default:
 174                         DCPOMATIC_ASSERT (false);
 175                 }
 176
 177                 DCPOMATIC_ASSERT (_codec_context->time_base.num == 1);
 178                 frame->pts = sample_offset * _codec_context->time_base.den / _codec_context->sample_rate;
 179
 180                 AVPacket packet;
 181                 av_init_packet (&packet);
 182                 packet.data = 0;
 183                 packet.size = 0;
 184                 int got_packet;
 185
 186                 DCPOMATIC_DISABLE_WARNINGS
 187                 if (avcodec_encode_audio2 (_codec_context, &packet, frame, &got_packet) < 0) {
 188                         throw EncodeError ("FFmpeg audio encode failed");
 189                 }
 190                 DCPOMATIC_ENABLE_WARNINGS
 191
 192                 if (got_packet && packet.size) {
 193                         packet.stream_index = _stream_index;
 194                         av_interleaved_write_frame (_format_context, &packet);
 195                         av_packet_unref (&packet);
 196                 }
 197
 198                 av_free (samples);
 199                 av_frame_free (&frame);
 200         }
 201
 202 private:
 203         AVFormatContext* _format_context;
 204         AVCodec* _codec;
 205         AVCodecContext* _codec_context;
 206         AVStream* _stream;
 207         int _stream_index;
 208 };
 209
 210
 211 FFmpegFileEncoder::FFmpegFileEncoder (
 212         dcp::Size video_frame_size,
 213         int video_frame_rate,
 214         int audio_frame_rate,
 215         int channels,
 216         ExportFormat format,
 217         bool audio_stream_per_channel,
 218         int x264_crf,
 219         boost::filesystem::path output
 220 #ifdef DCPOMATIC_VARIANT_SWAROOP
 221         , optional<dcp::Key> key
 222         , optional<string> id
 223 #endif
 224         )
 225         : _audio_stream_per_channel (audio_stream_per_channel)
 226         , _video_options (0)
 227         , _audio_channels (channels)
 228         , _output (output)
 229         , _video_frame_size (video_frame_size)
 230         , _video_frame_rate (video_frame_rate)
 231         , _audio_frame_rate (audio_frame_rate)
 232         , _audio_frames (0)
 233 {
 234         _pixel_format = pixel_format (format);
 235
 236         switch (format) {
 237         case EXPORT_FORMAT_PRORES:
 238                 _sample_format = AV_SAMPLE_FMT_S16;
 239                 _video_codec_name = "prores_ks";
 240                 _audio_codec_name = "pcm_s16le";
 241                 av_dict_set (&_video_options, "profile", "3", 0);
 242                 av_dict_set (&_video_options, "threads", "auto", 0);
 243                 break;
 244         case EXPORT_FORMAT_H264_AAC:
 245                 _sample_format = AV_SAMPLE_FMT_FLTP;
 246                 _video_codec_name = "libx264";
 247                 _audio_codec_name = "aac";
 248                 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
 249                 break;
 250         case EXPORT_FORMAT_H264_PCM:
 251                 _sample_format = AV_SAMPLE_FMT_S32;
 252                 _video_codec_name = "libx264";
 253                 _audio_codec_name = "pcm_s24le";
 254                 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
 255                 break;
 256         default:
 257                 DCPOMATIC_ASSERT (false);
 258         }
 259
 260 #ifdef DCPOMATIC_VARIANT_SWAROOP
 261         int r = avformat_alloc_output_context2 (&_format_context, av_guess_format("mov", 0, 0), 0, 0);
 262 #else
 263         int r = avformat_alloc_output_context2 (&_format_context, 0, 0, _output.string().c_str());
 264 #endif
 265         if (!_format_context) {
 266                 throw runtime_error (String::compose("could not allocate FFmpeg format context (%1)", r));
 267         }
 268
 269         setup_video ();
 270         setup_audio ();
 271
 272         r = avio_open_boost (&_format_context->pb, _output, AVIO_FLAG_WRITE);
 273         if (r < 0) {
 274                 throw runtime_error (String::compose("could not open FFmpeg output file %1 (%2)", _output.string(), r));
 275         }
 276
 277         AVDictionary* options = 0;
 278
 279 #ifdef DCPOMATIC_VARIANT_SWAROOP
 280         if (key) {
 281                 av_dict_set (&options, "encryption_key", key->hex().c_str(), 0);
 282                 /* XXX: is this OK? */
 283                 av_dict_set (&options, "encryption_kid", "00000000000000000000000000000000", 0);
 284                 av_dict_set (&options, "encryption_scheme", "cenc-aes-ctr", 0);
 285         }
 286
 287         if (id) {
 288                 if (av_dict_set(&_format_context->metadata, SWAROOP_ID_TAG, id->c_str(), 0) < 0) {
 289                         throw runtime_error ("Could not write ID to output");
 290                 }
 291         }
 292 #endif
 293
 294         if (avformat_write_header (_format_context, &options) < 0) {
 295                 throw runtime_error ("could not write header to FFmpeg output file");
 296         }
 297
 298         _pending_audio.reset (new AudioBuffers(channels, 0));
 299 }
 300
 301
 302 FFmpegFileEncoder::~FFmpegFileEncoder ()
 303 {
 304         _audio_streams.clear ();
 305         avcodec_close (_video_codec_context);
 306         avformat_free_context (_format_context);
 307 }
 308
 309
 310 AVPixelFormat
 311 FFmpegFileEncoder::pixel_format (ExportFormat format)
 312 {
 313         switch (format) {
 314         case EXPORT_FORMAT_PRORES:
 315                 return AV_PIX_FMT_YUV422P10;
 316         case EXPORT_FORMAT_H264_AAC:
 317         case EXPORT_FORMAT_H264_PCM:
 318                 return AV_PIX_FMT_YUV420P;
 319         default:
 320                 DCPOMATIC_ASSERT (false);
 321         }
 322
 323         return AV_PIX_FMT_YUV422P10;
 324 }
 325
 326 void
 327 FFmpegFileEncoder::setup_video ()
 328 {
 329         _video_codec = avcodec_find_encoder_by_name (_video_codec_name.c_str());
 330         if (!_video_codec) {
 331                 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _video_codec_name));
 332         }
 333
 334         _video_codec_context = avcodec_alloc_context3 (_video_codec);
 335         if (!_video_codec_context) {
 336                 throw runtime_error ("could not allocate FFmpeg video context");
 337         }
 338
 339         avcodec_get_context_defaults3 (_video_codec_context, _video_codec);
 340
 341         /* Variable quantisation */
 342         _video_codec_context->global_quality = 0;
 343         _video_codec_context->width = _video_frame_size.width;
 344         _video_codec_context->height = _video_frame_size.height;
 345         _video_codec_context->time_base = (AVRational) { 1, _video_frame_rate };
 346         _video_codec_context->pix_fmt = _pixel_format;
 347         _video_codec_context->flags |= AV_CODEC_FLAG_QSCALE | AV_CODEC_FLAG_GLOBAL_HEADER;
 348
 349         _video_stream = avformat_new_stream (_format_context, _video_codec);
 350         if (!_video_stream) {
 351                 throw runtime_error ("could not create FFmpeg output video stream");
 352         }
 353
 354 DCPOMATIC_DISABLE_WARNINGS
 355         _video_stream->id = _video_stream_index;
 356         _video_stream->codec = _video_codec_context;
 357 DCPOMATIC_ENABLE_WARNINGS
 358
 359         if (avcodec_open2 (_video_codec_context, _video_codec, &_video_options) < 0) {
 360                 throw runtime_error ("could not open FFmpeg video codec");
 361         }
 362 }
 363
 364
 365 void
 366 FFmpegFileEncoder::setup_audio ()
 367 {
 368         int const streams = _audio_stream_per_channel ? _audio_channels : 1;
 369         int const channels_per_stream = _audio_stream_per_channel ? 1 : _audio_channels;
 370
 371         for (int i = 0; i < streams; ++i) {
 372                 _audio_streams.push_back(
 373                         shared_ptr<ExportAudioStream>(
 374                                 new ExportAudioStream(_audio_codec_name, channels_per_stream, _audio_frame_rate, _sample_format, _format_context, _audio_stream_index_base + i)
 375                                 )
 376                         );
 377         }
 378 }
 379
 380
 381 void
 382 FFmpegFileEncoder::flush ()
 383 {
 384         if (_pending_audio->frames() > 0) {
 385                 audio_frame (_pending_audio->frames ());
 386         }
 387
 388         bool flushed_video = false;
 389         bool flushed_audio = false;
 390
 391         while (!flushed_video || !flushed_audio) {
 392                 AVPacket packet;
 393                 av_init_packet (&packet);
 394                 packet.data = 0;
 395                 packet.size = 0;
 396
 397                 int got_packet;
 398 DCPOMATIC_DISABLE_WARNINGS
 399                 avcodec_encode_video2 (_video_codec_context, &packet, 0, &got_packet);
 400 DCPOMATIC_ENABLE_WARNINGS
 401                 if (got_packet) {
 402                         packet.stream_index = 0;
 403                         av_interleaved_write_frame (_format_context, &packet);
 404                 } else {
 405                         flushed_video = true;
 406                 }
 407                 av_packet_unref (&packet);
 408
 409                 flushed_audio = true;
 410                 BOOST_FOREACH (shared_ptr<ExportAudioStream> i, _audio_streams) {
 411                         if (!i->flush()) {
 412                                 flushed_audio = false;
 413                         }
 414                 }
 415         }
 416
 417         av_write_trailer (_format_context);
 418 }
 419
 420 void
 421 FFmpegFileEncoder::video (shared_ptr<PlayerVideo> video, DCPTime time)
 422 {
 423         shared_ptr<Image> image = video->image (
 424                 bind (&PlayerVideo::force, _1, _pixel_format),
 425                 true,
 426                 false
 427                 );
 428
 429         AVFrame* frame = av_frame_alloc ();
 430         DCPOMATIC_ASSERT (frame);
 431
 432         {
 433                 boost::mutex::scoped_lock lm (_pending_images_mutex);
 434                 _pending_images[image->data()[0]] = image;
 435         }
 436
 437         for (int i = 0; i < 3; ++i) {
 438                 AVBufferRef* buffer = av_buffer_create(image->data()[i], image->stride()[i] * image->size().height, &buffer_free, this, 0);
 439                 frame->buf[i] = av_buffer_ref (buffer);
 440                 frame->data[i] = buffer->data;
 441                 frame->linesize[i] = image->stride()[i];
 442                 av_buffer_unref (&buffer);
 443         }
 444
 445         frame->width = image->size().width;
 446         frame->height = image->size().height;
 447         frame->format = _pixel_format;
 448         DCPOMATIC_ASSERT (_video_stream->time_base.num == 1);
 449         frame->pts = time.get() * _video_stream->time_base.den / DCPTime::HZ;
 450
 451         AVPacket packet;
 452         av_init_packet (&packet);
 453         packet.data = 0;
 454         packet.size = 0;
 455
 456         int got_packet;
 457 DCPOMATIC_DISABLE_WARNINGS
 458         if (avcodec_encode_video2 (_video_codec_context, &packet, frame, &got_packet) < 0) {
 459                 throw EncodeError ("FFmpeg video encode failed");
 460         }
 461 DCPOMATIC_ENABLE_WARNINGS
 462
 463         if (got_packet && packet.size) {
 464                 packet.stream_index = _video_stream_index;
 465                 av_interleaved_write_frame (_format_context, &packet);
 466                 av_packet_unref (&packet);
 467         }
 468
 469         av_frame_free (&frame);
 470
 471 }
 472
 473 /** Called when the player gives us some audio */
 474 void
 475 FFmpegFileEncoder::audio (shared_ptr<AudioBuffers> audio)
 476 {
 477         _pending_audio->append (audio);
 478
 479         DCPOMATIC_ASSERT (!_audio_streams.empty());
 480         int frame_size = _audio_streams[0]->frame_size();
 481         if (frame_size == 0) {
 482                 /* codec has AV_CODEC_CAP_VARIABLE_FRAME_SIZE */
 483                 frame_size = _audio_frame_rate / _video_frame_rate;
 484         }
 485
 486         while (_pending_audio->frames() >= frame_size) {
 487                 audio_frame (frame_size);
 488         }
 489 }
 490
 491 void
 492 FFmpegFileEncoder::audio_frame (int size)
 493 {
 494         if (_audio_stream_per_channel) {
 495                 int offset = 0;
 496                 BOOST_FOREACH (shared_ptr<ExportAudioStream> i, _audio_streams) {
 497                         i->write (size, offset, 1, _pending_audio->data(), _audio_frames);
 498                         ++offset;
 499                 }
 500         } else {
 501                 DCPOMATIC_ASSERT (!_audio_streams.empty());
 502                 DCPOMATIC_ASSERT (_pending_audio->channels());
 503                 _audio_streams[0]->write (size, 0, _pending_audio->channels(), _pending_audio->data(), _audio_frames);
 504         }
 505
 506         _pending_audio->trim_start (size);
 507         _audio_frames += size;
 508 }
 509
 510 void
 511 FFmpegFileEncoder::subtitle (PlayerText, DCPTimePeriod)
 512 {
 513
 514 }
 515
 516 void
 517 FFmpegFileEncoder::buffer_free (void* opaque, uint8_t* data)
 518 {
 519         reinterpret_cast<FFmpegFileEncoder*>(opaque)->buffer_free2(data);
 520 }
 521
 522 void
 523 FFmpegFileEncoder::buffer_free2 (uint8_t* data)
 524 {
 525         boost::mutex::scoped_lock lm (_pending_images_mutex);
 526         if (_pending_images.find(data) != _pending_images.end()) {
 527                 _pending_images.erase (data);
 528         }
 529 }