src/lib/ffmpeg_file_encoder.cc

   1 /*
   2     Copyright (C) 2017-2021 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
  22 #include "compose.hpp"
  23 #include "cross.h"
  24 #include "ffmpeg_encoder.h"
  25 #include "ffmpeg_wrapper.h"
  26 #include "film.h"
  27 #include "image.h"
  28 #include "job.h"
  29 #include "log.h"
  30 #include "player.h"
  31 #include "player_video.h"
  32 extern "C" {
  33 #include <libavutil/channel_layout.h>
  34 }
  35 #include <iostream>
  36
  37 #include "i18n.h"
  38
  39
  40 using std::cout;
  41 using std::make_shared;
  42 using std::shared_ptr;
  43 using std::string;
  44 using boost::bind;
  45 using namespace dcpomatic;
  46 #if BOOST_VERSION >= 106100
  47 using namespace boost::placeholders;
  48 #endif
  49
  50
  51 int FFmpegFileEncoder::_video_stream_index = 0;
  52 int FFmpegFileEncoder::_audio_stream_index_base = 1;
  53
  54
  55 class ExportAudioStream
  56 {
  57 public:
  58         ExportAudioStream (string codec_name, int channels, int frame_rate, AVSampleFormat sample_format, AVFormatContext* format_context, int stream_index)
  59                 : _format_context (format_context)
  60                 , _stream_index (stream_index)
  61         {
  62                 _codec = avcodec_find_encoder_by_name (codec_name.c_str());
  63                 if (!_codec) {
  64                         throw EncodeError (String::compose("avcodec_find_encoder_by_name failed for %1", codec_name));
  65                 }
  66
  67                 _codec_context = avcodec_alloc_context3 (_codec);
  68                 if (!_codec_context) {
  69                         throw std::bad_alloc ();
  70                 }
  71
  72                 /* XXX: configurable */
  73                 _codec_context->bit_rate = channels * 128 * 1024;
  74                 _codec_context->sample_fmt = sample_format;
  75                 _codec_context->sample_rate = frame_rate;
  76                 _codec_context->channel_layout = av_get_default_channel_layout (channels);
  77                 _codec_context->channels = channels;
  78
  79                 int r = avcodec_open2 (_codec_context, _codec, 0);
  80                 if (r < 0) {
  81                         throw EncodeError (N_("avcodec_open2"), N_("ExportAudioStream::ExportAudioStream"), r);
  82                 }
  83
  84                 _stream = avformat_new_stream (format_context, _codec);
  85                 if (!_stream) {
  86                         throw EncodeError (N_("avformat_new_stream"), N_("ExportAudioStream::ExportAudioStream"));
  87                 }
  88
  89                 _stream->id = stream_index;
  90                 _stream->disposition |= AV_DISPOSITION_DEFAULT;
  91                 r = avcodec_parameters_from_context (_stream->codecpar, _codec_context);
  92                 if (r < 0) {
  93                         throw EncodeError (N_("avcodec_parameters_from_context"), N_("ExportAudioStream::ExportAudioStream"), r);
  94                 }
  95         }
  96
  97         ~ExportAudioStream ()
  98         {
  99                 avcodec_close (_codec_context);
 100         }
 101
 102         ExportAudioStream (ExportAudioStream const&) = delete;
 103         ExportAudioStream& operator= (ExportAudioStream const&) = delete;
 104
 105         int frame_size () const {
 106                 return _codec_context->frame_size;
 107         }
 108
 109         bool flush ()
 110         {
 111                 int r = avcodec_send_frame (_codec_context, nullptr);
 112                 if (r < 0 && r != AVERROR_EOF) {
 113                         /* We get EOF if we've already flushed the stream once */
 114                         throw EncodeError (N_("avcodec_send_frame"), N_("ExportAudioStream::flush"), r);
 115                 }
 116
 117                 ffmpeg::Packet packet;
 118                 r = avcodec_receive_packet (_codec_context, packet.get());
 119                 if (r == AVERROR_EOF) {
 120                         return true;
 121                 } else if (r < 0) {
 122                         throw EncodeError (N_("avcodec_receive_packet"), N_("ExportAudioStream::flush"), r);
 123                 }
 124
 125                 packet->stream_index = _stream_index;
 126                 av_interleaved_write_frame (_format_context, packet.get());
 127                 return false;
 128         }
 129
 130         void write (int size, int channel_offset, int channels, float** data, int64_t sample_offset)
 131         {
 132                 DCPOMATIC_ASSERT (size);
 133
 134                 auto frame = av_frame_alloc ();
 135                 DCPOMATIC_ASSERT (frame);
 136
 137                 int const buffer_size = av_samples_get_buffer_size (0, channels, size, _codec_context->sample_fmt, 0);
 138                 DCPOMATIC_ASSERT (buffer_size >= 0);
 139
 140                 auto samples = av_malloc (buffer_size);
 141                 DCPOMATIC_ASSERT (samples);
 142
 143                 frame->nb_samples = size;
 144                 frame->format = _codec_context->sample_fmt;
 145                 frame->channels = channels;
 146                 int r = avcodec_fill_audio_frame (frame, channels, _codec_context->sample_fmt, (const uint8_t *) samples, buffer_size, 0);
 147                 DCPOMATIC_ASSERT (r >= 0);
 148
 149                 switch (_codec_context->sample_fmt) {
 150                 case AV_SAMPLE_FMT_S16:
 151                 {
 152                         int16_t* q = reinterpret_cast<int16_t*> (samples);
 153                         for (int i = 0; i < size; ++i) {
 154                                 for (int j = 0; j < channels; ++j) {
 155                                         *q++ = data[j + channel_offset][i] * 32767;
 156                                 }
 157                         }
 158                         break;
 159                 }
 160                 case AV_SAMPLE_FMT_S32:
 161                 {
 162                         int32_t* q = reinterpret_cast<int32_t*> (samples);
 163                         for (int i = 0; i < size; ++i) {
 164                                 for (int j = 0; j < channels; ++j) {
 165                                         *q++ = data[j + channel_offset][i] * 2147483647;
 166                                 }
 167                         }
 168                         break;
 169                 }
 170                 case AV_SAMPLE_FMT_FLTP:
 171                 {
 172                         float* q = reinterpret_cast<float*> (samples);
 173                         for (int i = 0; i < channels; ++i) {
 174                                 memcpy (q, data[i + channel_offset], sizeof(float) * size);
 175                                 q += size;
 176                         }
 177                         break;
 178                 }
 179                 default:
 180                         DCPOMATIC_ASSERT (false);
 181                 }
 182
 183                 DCPOMATIC_ASSERT (_codec_context->time_base.num == 1);
 184                 frame->pts = sample_offset * _codec_context->time_base.den / _codec_context->sample_rate;
 185
 186                 r = avcodec_send_frame (_codec_context, frame);
 187                 av_free (samples);
 188                 av_frame_free (&frame);
 189                 if (r < 0) {
 190                         throw EncodeError (N_("avcodec_send_frame"), N_("ExportAudioStream::write"), r);
 191                 }
 192
 193                 ffmpeg::Packet packet;
 194                 r = avcodec_receive_packet (_codec_context, packet.get());
 195                 if (r < 0 && r != AVERROR(EAGAIN)) {
 196                         throw EncodeError (N_("avcodec_receive_packet"), N_("ExportAudioStream::write"), r);
 197                 } else if (r >= 0) {
 198                         packet->stream_index = _stream_index;
 199                         av_interleaved_write_frame (_format_context, packet.get());
 200                 }
 201         }
 202
 203 private:
 204         AVFormatContext* _format_context;
 205         AVCodec const * _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         )
 222         : _audio_stream_per_channel (audio_stream_per_channel)
 223         , _audio_channels (channels)
 224         , _output (output)
 225         , _video_frame_size (video_frame_size)
 226         , _video_frame_rate (video_frame_rate)
 227         , _audio_frame_rate (audio_frame_rate)
 228 {
 229         _pixel_format = pixel_format (format);
 230
 231         switch (format) {
 232         case ExportFormat::PRORES:
 233                 _sample_format = AV_SAMPLE_FMT_S16;
 234                 _video_codec_name = "prores_ks";
 235                 _audio_codec_name = "pcm_s16le";
 236                 av_dict_set (&_video_options, "profile", "3", 0);
 237                 av_dict_set (&_video_options, "threads", "auto", 0);
 238                 break;
 239         case ExportFormat::H264_AAC:
 240                 _sample_format = AV_SAMPLE_FMT_FLTP;
 241                 _video_codec_name = "libx264";
 242                 _audio_codec_name = "aac";
 243                 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
 244                 break;
 245         case ExportFormat::H264_PCM:
 246                 _sample_format = AV_SAMPLE_FMT_S32;
 247                 _video_codec_name = "libx264";
 248                 _audio_codec_name = "pcm_s24le";
 249                 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
 250                 break;
 251         default:
 252                 DCPOMATIC_ASSERT (false);
 253         }
 254
 255         int r = avformat_alloc_output_context2 (&_format_context, 0, 0, _output.string().c_str());
 256         if (!_format_context) {
 257                 throw EncodeError (N_("avformat_alloc_output_context2"), "FFmpegFileEncoder::FFmpegFileEncoder", r);
 258         }
 259
 260         setup_video ();
 261         setup_audio ();
 262
 263         r = avio_open_boost (&_format_context->pb, _output, AVIO_FLAG_WRITE);
 264         if (r < 0) {
 265                 throw EncodeError (String::compose(_("Could not open output file %1 (%2)"), _output.string(), r));
 266         }
 267
 268         AVDictionary* options = nullptr;
 269
 270         r = avformat_write_header (_format_context, &options);
 271         if (r < 0) {
 272                 throw EncodeError (N_("avformat_write_header"), N_("FFmpegFileEncoder::FFmpegFileEncoder"), r);
 273         }
 274
 275         _pending_audio.reset (new AudioBuffers(channels, 0));
 276 }
 277
 278
 279 FFmpegFileEncoder::~FFmpegFileEncoder ()
 280 {
 281         _audio_streams.clear ();
 282         avcodec_close (_video_codec_context);
 283         avformat_free_context (_format_context);
 284 }
 285
 286
 287 AVPixelFormat
 288 FFmpegFileEncoder::pixel_format (ExportFormat format)
 289 {
 290         switch (format) {
 291         case ExportFormat::PRORES:
 292                 return AV_PIX_FMT_YUV422P10;
 293         case ExportFormat::H264_AAC:
 294         case ExportFormat::H264_PCM:
 295                 return AV_PIX_FMT_YUV420P;
 296         default:
 297                 DCPOMATIC_ASSERT (false);
 298         }
 299
 300         return AV_PIX_FMT_YUV422P10;
 301 }
 302
 303
 304 void
 305 FFmpegFileEncoder::setup_video ()
 306 {
 307         _video_codec = avcodec_find_encoder_by_name (_video_codec_name.c_str());
 308         if (!_video_codec) {
 309                 throw EncodeError (String::compose("avcodec_find_encoder_by_name failed for %1", _video_codec_name));
 310         }
 311
 312         _video_codec_context = avcodec_alloc_context3 (_video_codec);
 313         if (!_video_codec_context) {
 314                 throw std::bad_alloc ();
 315         }
 316
 317         /* Variable quantisation */
 318         _video_codec_context->global_quality = 0;
 319         _video_codec_context->width = _video_frame_size.width;
 320         _video_codec_context->height = _video_frame_size.height;
 321         _video_codec_context->time_base = (AVRational) { 1, _video_frame_rate };
 322         _video_codec_context->pix_fmt = _pixel_format;
 323         _video_codec_context->flags |= AV_CODEC_FLAG_QSCALE | AV_CODEC_FLAG_GLOBAL_HEADER;
 324
 325         if (avcodec_open2 (_video_codec_context, _video_codec, &_video_options) < 0) {
 326                 throw EncodeError (N_("avcodec_open"), N_("FFmpegFileEncoder::setup_video"));
 327         }
 328
 329         _video_stream = avformat_new_stream (_format_context, _video_codec);
 330         if (!_video_stream) {
 331                 throw EncodeError (N_("avformat_new_stream"), N_("FFmpegFileEncoder::setup_video"));
 332         }
 333
 334         _video_stream->id = _video_stream_index;
 335         int r = avcodec_parameters_from_context (_video_stream->codecpar, _video_codec_context);
 336         if (r < 0) {
 337                 throw EncodeError (N_("avcodec_parameters_from_context"), N_("FFmpegFileEncoder::setup_video"), r);
 338         }
 339 }
 340
 341
 342 void
 343 FFmpegFileEncoder::setup_audio ()
 344 {
 345         int const streams = _audio_stream_per_channel ? _audio_channels : 1;
 346         int const channels_per_stream = _audio_stream_per_channel ? 1 : _audio_channels;
 347
 348         for (int i = 0; i < streams; ++i) {
 349                 _audio_streams.push_back(
 350                         make_shared<ExportAudioStream>(
 351                                 _audio_codec_name, channels_per_stream, _audio_frame_rate, _sample_format, _format_context, _audio_stream_index_base + i
 352                                 )
 353                         );
 354         }
 355 }
 356
 357
 358 void
 359 FFmpegFileEncoder::flush ()
 360 {
 361         if (_pending_audio->frames() > 0) {
 362                 audio_frame (_pending_audio->frames ());
 363         }
 364
 365         bool flushed_video = false;
 366         bool flushed_audio = false;
 367
 368         while (!flushed_video || !flushed_audio) {
 369                 int r = avcodec_send_frame (_video_codec_context, nullptr);
 370                 if (r < 0 && r != AVERROR_EOF) {
 371                         /* We get EOF if we've already flushed the stream once */
 372                         throw EncodeError (N_("avcodec_send_frame"), N_("FFmpegFileEncoder::flush"), r);
 373                 }
 374
 375                 ffmpeg::Packet packet;
 376                 r = avcodec_receive_packet (_video_codec_context, packet.get());
 377                 if (r == AVERROR_EOF) {
 378                         flushed_video = true;
 379                 } else if (r < 0) {
 380                         throw EncodeError (N_("avcodec_receive_packet"), N_("FFmpegFileEncoder::flush"), r);
 381                 } else {
 382                         packet->stream_index = _video_stream_index;
 383                         av_interleaved_write_frame (_format_context, packet.get());
 384                 }
 385
 386                 flushed_audio = true;
 387                 for (auto i: _audio_streams) {
 388                         if (!i->flush()) {
 389                                 flushed_audio = false;
 390                         }
 391                 }
 392         }
 393
 394         av_write_trailer (_format_context);
 395 }
 396
 397
 398 void
 399 FFmpegFileEncoder::video (shared_ptr<PlayerVideo> video, DCPTime time)
 400 {
 401         /* All our output formats are video range at the moment */
 402         auto image = video->image (
 403                 bind (&PlayerVideo::force, _1, _pixel_format),
 404                 VideoRange::VIDEO,
 405                 false
 406                 );
 407
 408         auto frame = av_frame_alloc ();
 409         DCPOMATIC_ASSERT (frame);
 410
 411         {
 412                 boost::mutex::scoped_lock lm (_pending_images_mutex);
 413                 _pending_images[image->data()[0]] = image;
 414         }
 415
 416         for (int i = 0; i < 3; ++i) {
 417                 auto buffer = av_buffer_create(image->data()[i], image->stride()[i] * image->size().height, &buffer_free, this, 0);
 418                 frame->buf[i] = av_buffer_ref (buffer);
 419                 frame->data[i] = buffer->data;
 420                 frame->linesize[i] = image->stride()[i];
 421                 av_buffer_unref (&buffer);
 422         }
 423
 424         frame->width = image->size().width;
 425         frame->height = image->size().height;
 426         frame->format = _pixel_format;
 427         DCPOMATIC_ASSERT (_video_stream->time_base.num == 1);
 428         frame->pts = time.get() * _video_stream->time_base.den / DCPTime::HZ;
 429
 430         int r = avcodec_send_frame (_video_codec_context, frame);
 431         av_frame_free (&frame);
 432         if (r < 0) {
 433                 throw EncodeError (N_("avcodec_send_frame"), N_("FFmpegFileEncoder::video"), r);
 434         }
 435
 436         ffmpeg::Packet packet;
 437         r = avcodec_receive_packet (_video_codec_context, packet.get());
 438         if (r < 0 && r != AVERROR(EAGAIN)) {
 439                 throw EncodeError (N_("avcodec_receive_packet"), N_("FFmpegFileEncoder::video"), r);
 440         } else if (r >= 0) {
 441                 packet->stream_index = _video_stream_index;
 442                 av_interleaved_write_frame (_format_context, packet.get());
 443         }
 444 }
 445
 446
 447 /** Called when the player gives us some audio */
 448 void
 449 FFmpegFileEncoder::audio (shared_ptr<AudioBuffers> audio)
 450 {
 451         _pending_audio->append (audio);
 452
 453         DCPOMATIC_ASSERT (!_audio_streams.empty());
 454         int frame_size = _audio_streams[0]->frame_size();
 455         if (frame_size == 0) {
 456                 /* codec has AV_CODEC_CAP_VARIABLE_FRAME_SIZE */
 457                 frame_size = _audio_frame_rate / _video_frame_rate;
 458         }
 459
 460         while (_pending_audio->frames() >= frame_size) {
 461                 audio_frame (frame_size);
 462         }
 463 }
 464
 465
 466 void
 467 FFmpegFileEncoder::audio_frame (int size)
 468 {
 469         if (_audio_stream_per_channel) {
 470                 int offset = 0;
 471                 for (auto i: _audio_streams) {
 472                         i->write (size, offset, 1, _pending_audio->data(), _audio_frames);
 473                         ++offset;
 474                 }
 475         } else {
 476                 DCPOMATIC_ASSERT (!_audio_streams.empty());
 477                 DCPOMATIC_ASSERT (_pending_audio->channels());
 478                 _audio_streams[0]->write (size, 0, _pending_audio->channels(), _pending_audio->data(), _audio_frames);
 479         }
 480
 481         _pending_audio->trim_start (size);
 482         _audio_frames += size;
 483 }
 484
 485
 486 void
 487 FFmpegFileEncoder::subtitle (PlayerText, DCPTimePeriod)
 488 {
 489
 490 }
 491
 492
 493 void
 494 FFmpegFileEncoder::buffer_free (void* opaque, uint8_t* data)
 495 {
 496         reinterpret_cast<FFmpegFileEncoder*>(opaque)->buffer_free2(data);
 497 }
 498
 499
 500 void
 501 FFmpegFileEncoder::buffer_free2 (uint8_t* data)
 502 {
 503         boost::mutex::scoped_lock lm (_pending_images_mutex);
 504         if (_pending_images.find(data) != _pending_images.end()) {
 505                 _pending_images.erase (data);
 506         }
 507 }