git.sesse.net Git - casparcg/blob - modules/ffmpeg/consumer/ffmpeg_consumer.cpp

   1 /*
   2 * Copyright (c) 2011 Sveriges Television AB <info@casparcg.com>
   3 *
   4 * This file is part of CasparCG (www.casparcg.com).
   5 *
   6 * CasparCG 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 3 of the License, or
   9 * (at your option) any later version.
  10 *
  11 * CasparCG 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 CasparCG. If not, see <http://www.gnu.org/licenses/>.
  18 *
  19 * Author: Robert Nagy, ronag89@gmail.com
  20 */
  21
  22 #include "../StdAfx.h"
  23
  24 #include "../ffmpeg_error.h"
  25
  26 #include "ffmpeg_consumer.h"
  27
  28 #include "../producer/tbb_avcodec.h"
  29
  30 #include <core/frame/frame.h>
  31 #include <core/frame/audio_channel_layout.h>
  32 #include <core/mixer/audio/audio_util.h>
  33 #include <core/consumer/frame_consumer.h>
  34 #include <core/video_format.h>
  35 #include <core/help/help_repository.h>
  36 #include <core/help/help_sink.h>
  37
  38 #include <common/array.h>
  39 #include <common/env.h>
  40 #include <common/except.h>
  41 #include <common/executor.h>
  42 #include <common/future.h>
  43 #include <common/diagnostics/graph.h>
  44 #include <common/lock.h>
  45 #include <common/memory.h>
  46 #include <common/param.h>
  47 #include <common/utf.h>
  48 #include <common/assert.h>
  49 #include <common/memshfl.h>
  50 #include <common/timer.h>
  51 #include <common/ptree.h>
  52
  53 #include <boost/algorithm/string.hpp>
  54 #include <boost/property_tree/ptree.hpp>
  55 #include <boost/filesystem.hpp>
  56 #include <boost/range/algorithm.hpp>
  57 #include <boost/range/algorithm_ext.hpp>
  58 #include <boost/lexical_cast.hpp>
  59
  60 #include <tbb/spin_mutex.h>
  61
  62 #include <numeric>
  63 #include <cstring>
  64
  65 #if defined(_MSC_VER)
  66 #pragma warning (push)
  67 #pragma warning (disable : 4244)
  68 #endif
  69 extern "C"
  70 {
  71         #define __STDC_CONSTANT_MACROS
  72         #define __STDC_LIMIT_MACROS
  73         #include <libavformat/avformat.h>
  74         #include <libswscale/swscale.h>
  75         #include <libavutil/opt.h>
  76         #include <libavutil/pixdesc.h>
  77         #include <libavutil/parseutils.h>
  78         #include <libavutil/samplefmt.h>
  79         #include <libswresample/swresample.h>
  80 }
  81 #if defined(_MSC_VER)
  82 #pragma warning (pop)
  83 #endif
  84
  85 namespace caspar { namespace ffmpeg {
  86
  87 int av_opt_set(void *obj, const char *name, const char *val, int search_flags)
  88 {
  89         AVClass* av_class = *(AVClass**)obj;
  90
  91         if((strcmp(name, "pix_fmt") == 0 || strcmp(name, "pixel_format") == 0) && strcmp(av_class->class_name, "AVCodecContext") == 0)
  92         {
  93                 AVCodecContext* c = (AVCodecContext*)obj;
  94                 auto pix_fmt = av_get_pix_fmt(val);
  95                 if(pix_fmt == PIX_FMT_NONE)
  96                         return -1;
  97                 c->pix_fmt = pix_fmt;
  98                 return 0;
  99         }
 100         //if((strcmp(name, "r") == 0 || strcmp(name, "frame_rate") == 0) && strcmp(av_class->class_name, "AVCodecContext") == 0)
 101         //{
 102         //      AVCodecContext* c = (AVCodecContext*)obj;
 103
 104         //      if(c->codec_type != AVMEDIA_TYPE_VIDEO)
 105         //              return -1;
 106
 107         //      AVRational rate;
 108         //      int ret = av_parse_video_rate(&rate, val);
 109         //      if(ret < 0)
 110         //              return ret;
 111
 112         //      c->time_base.num = rate.den;
 113         //      c->time_base.den = rate.num;
 114         //      return 0;
 115         //}
 116
 117         return ::av_opt_set(obj, name, val, search_flags);
 118 }
 119
 120 struct option
 121 {
 122         std::string name;
 123         std::string value;
 124
 125         option(std::string name, std::string value)
 126                 : name(std::move(name))
 127                 , value(std::move(value))
 128         {
 129         }
 130 };
 131
 132 struct output_format
 133 {
 134         AVOutputFormat* format;
 135         int                             width;
 136         int                             height;
 137         AVCodecID               vcodec;
 138         AVCodecID               acodec;
 139         int                             croptop;
 140         int                             cropbot;
 141
 142         output_format(const core::video_format_desc& format_desc, const std::string& filename, std::vector<option>& options)
 143                 : format(av_guess_format(nullptr, filename.c_str(), nullptr))
 144                 , width(format_desc.width)
 145                 , height(format_desc.height)
 146                 , vcodec(CODEC_ID_NONE)
 147                 , acodec(CODEC_ID_NONE)
 148                 , croptop(0)
 149                 , cropbot(0)
 150         {
 151                 if(boost::iequals(boost::filesystem::path(filename).extension().string(), ".dv"))
 152                         set_opt("f", "dv");
 153
 154                 boost::range::remove_erase_if(options, [&](const option& o)
 155                 {
 156                         return set_opt(o.name, o.value);
 157                 });
 158
 159                 if(vcodec == CODEC_ID_NONE && format)
 160                         vcodec = format->video_codec;
 161
 162                 if(acodec == CODEC_ID_NONE && format)
 163                         acodec = format->audio_codec;
 164
 165                 if(vcodec == CODEC_ID_NONE)
 166                         vcodec = CODEC_ID_H264;
 167
 168                 if(acodec == CODEC_ID_NONE)
 169                         acodec = CODEC_ID_PCM_S16LE;
 170         }
 171
 172         bool set_opt(const std::string& name, const std::string& value)
 173         {
 174                 //if(name == "target")
 175                 //{
 176                 //      enum { PAL, NTSC, FILM, UNKNOWN } norm = UNKNOWN;
 177                 //
 178                 //      if(name.find("pal-") != std::string::npos)
 179                 //              norm = PAL;
 180                 //      else if(name.find("ntsc-") != std::string::npos)
 181                 //              norm = NTSC;
 182
 183                 //      if(norm == UNKNOWN)
 184                 //              CASPAR_THROW_EXCEPTION(invalid_argument() << arg_name_info("target"));
 185                 //
 186                 //      if (name.find("-dv") != std::string::npos)
 187                 //      {
 188                 //              set_opt("f", "dv");
 189                 //              if(norm == PAL)
 190                 //              {
 191                 //                      set_opt("s", "720x576");
 192                 //              }
 193                 //              else
 194                 //              {
 195                 //                      set_opt("s", "720x480");
 196                 //                      if(height == 486)
 197                 //                      {
 198                 //                              set_opt("croptop", "2");
 199                 //                              set_opt("cropbot", "4");
 200                 //                      }
 201                 //              }
 202                 //              set_opt("s", norm == PAL ? "720x576" : "720x480");
 203                 //      }
 204
 205                 //      return true;
 206                 //}
 207                 //else
 208                 if(name == "f")
 209                 {
 210                         format = av_guess_format(value.c_str(), nullptr, nullptr);
 211
 212                         if(format == nullptr)
 213                                 CASPAR_THROW_EXCEPTION(user_error() << msg_info("Unknown format " + value));
 214
 215                         return true;
 216                 }
 217                 else if(name == "vcodec" || name == "v:codec")
 218                 {
 219                         auto c = avcodec_find_encoder_by_name(value.c_str());
 220                         if(c == nullptr)
 221                                 CASPAR_THROW_EXCEPTION(user_error() << msg_info("Unknown video codec " + value));
 222
 223                         vcodec = avcodec_find_encoder_by_name(value.c_str())->id;
 224                         return true;
 225
 226                 }
 227                 else if(name == "acodec" || name == "a:codec")
 228                 {
 229                         auto c = avcodec_find_encoder_by_name(value.c_str());
 230                         if(c == nullptr)
 231                                 CASPAR_THROW_EXCEPTION(user_error() << msg_info("Unknown audio codec " + value));
 232
 233                         acodec = avcodec_find_encoder_by_name(value.c_str())->id;
 234
 235                         return true;
 236                 }
 237                 else if(name == "s")
 238                 {
 239                         if(av_parse_video_size(&width, &height, value.c_str()) < 0)
 240                                 CASPAR_THROW_EXCEPTION(user_error() << msg_info("Unknown video size " + value));
 241
 242                         return true;
 243                 }
 244                 else if(name == "croptop")
 245                 {
 246                         croptop = boost::lexical_cast<int>(value);
 247
 248                         return true;
 249                 }
 250                 else if(name == "cropbot")
 251                 {
 252                         cropbot = boost::lexical_cast<int>(value);
 253
 254                         return true;
 255                 }
 256
 257                 return false;
 258         }
 259 };
 260
 261 typedef cache_aligned_vector<uint8_t> byte_vector;
 262
 263 struct ffmpeg_consumer : boost::noncopyable
 264 {
 265         const spl::shared_ptr<diagnostics::graph>       graph_;
 266         const std::string                                                       filename_;
 267         const std::string                                                       full_filename_          = u8(env::media_folder()) + filename_;
 268         const std::shared_ptr<AVFormatContext>          oc_                                     { avformat_alloc_context(), avformat_free_context };
 269         const core::video_format_desc                           format_desc_;
 270         const core::audio_channel_layout                        channel_layout_;
 271
 272         core::monitor::subject                                          monitor_subject_;
 273
 274         tbb::spin_mutex                                                         exception_mutex_;
 275         std::exception_ptr                                                      exception_;
 276
 277         std::shared_ptr<AVStream>                                       audio_st_;
 278         std::shared_ptr<AVStream>                                       video_st_;
 279
 280         byte_vector                                                                     picture_buffer_;
 281         byte_vector                                                                     key_picture_buf_;
 282         byte_vector                                                                     audio_buffer_;
 283         std::shared_ptr<SwrContext>                                     swr_;
 284         std::shared_ptr<SwsContext>                                     sws_;
 285
 286         int64_t                                                                         frame_number_           = 0;
 287
 288         output_format                                                           output_format_;
 289         bool                                                                            key_only_;
 290         tbb::atomic<int64_t>                                            current_encoding_delay_;
 291
 292         executor                                                                        executor_;
 293 public:
 294         ffmpeg_consumer(
 295                         const std::string& filename,
 296                         const core::video_format_desc& format_desc,
 297                         const core::audio_channel_layout& channel_layout,
 298                         std::vector<option> options,
 299                         bool key_only)
 300                 : filename_(filename)
 301                 , format_desc_(format_desc)
 302                 , channel_layout_(channel_layout)
 303                 , output_format_(format_desc, full_filename_, options)
 304                 , key_only_(key_only)
 305                 , executor_(print())
 306         {
 307                 current_encoding_delay_ = 0;
 308                 check_space();
 309
 310                 // TODO: Ask stakeholders about case where file already exists.
 311                 boost::filesystem::remove(boost::filesystem::path(full_filename_)); // Delete the file if it exists
 312
 313                 graph_->set_color("frame-time", diagnostics::color(0.1f, 1.0f, 0.1f));
 314                 graph_->set_color("dropped-frame", diagnostics::color(0.3f, 0.6f, 0.3f));
 315                 graph_->set_text(print());
 316                 diagnostics::register_graph(graph_);
 317
 318                 executor_.set_capacity(8);
 319
 320                 oc_->oformat = output_format_.format;
 321
 322                 std::strcpy(oc_->filename, full_filename_.c_str());
 323
 324                 //  Add the audio and video streams using the default format codecs     and initialize the codecs.
 325                 video_st_ = add_video_stream(options);
 326
 327                 if (!key_only)
 328                         audio_st_ = add_audio_stream(options);
 329
 330                 av_dump_format(oc_.get(), 0, full_filename_.c_str(), 1);
 331
 332                 // Open the output ffmpeg, if needed.
 333                 if (!(oc_->oformat->flags & AVFMT_NOFILE))
 334                         THROW_ON_ERROR2(avio_open(&oc_->pb, full_filename_.c_str(), AVIO_FLAG_WRITE), "[ffmpeg_consumer]");
 335
 336                 THROW_ON_ERROR2(avformat_write_header(oc_.get(), nullptr), "[ffmpeg_consumer]");
 337
 338                 if(options.size() > 0)
 339                 {
 340                         for (auto& option : options)
 341                                 CASPAR_LOG(warning) << L"Invalid option: -" << u16(option.name) << L" " << u16(option.value);
 342                 }
 343         }
 344
 345         ~ffmpeg_consumer()
 346         {
 347                 try
 348                 {
 349                         executor_.wait();
 350                 }
 351                 catch(...)
 352                 {
 353                         CASPAR_LOG_CURRENT_EXCEPTION();
 354                 }
 355
 356                 LOG_ON_ERROR2(av_write_trailer(oc_.get()), "[ffmpeg_consumer]");
 357
 358                 if (!key_only_)
 359                         audio_st_.reset();
 360
 361                 video_st_.reset();
 362
 363                 if (!(oc_->oformat->flags & AVFMT_NOFILE))
 364                         LOG_ON_ERROR2(avio_close(oc_->pb), "[ffmpeg_consumer]");
 365         }
 366
 367         // frame_consumer
 368
 369         void send(core::const_frame& frame)
 370         {
 371                 auto exception = lock(exception_mutex_, [&]
 372                 {
 373                         return exception_;
 374                 });
 375
 376                 if(exception != nullptr)
 377                         std::rethrow_exception(exception);
 378
 379                 executor_.begin_invoke([=]
 380                 {
 381                         encode(frame);
 382                         current_encoding_delay_ = frame.get_age_millis();
 383                 });
 384         }
 385
 386         bool ready_for_frame() const
 387         {
 388                 return !executor_.is_full();
 389         }
 390
 391         void mark_dropped()
 392         {
 393                 graph_->set_tag(diagnostics::tag_severity::WARNING, "dropped-frame");
 394         }
 395
 396         std::wstring print() const
 397         {
 398                 return L"ffmpeg[" + u16(filename_) + L"]";
 399         }
 400
 401         core::monitor::subject& monitor_output()
 402         {
 403                 return monitor_subject_;
 404         }
 405
 406 private:
 407         std::shared_ptr<AVStream> add_video_stream(std::vector<option>& options)
 408         {
 409                 if(output_format_.vcodec == CODEC_ID_NONE)
 410                         return nullptr;
 411
 412                 auto st = avformat_new_stream(oc_.get(), 0);
 413                 if (!st)
 414                         CASPAR_THROW_EXCEPTION(caspar_exception() << msg_info("Could not allocate video-stream.") << boost::errinfo_api_function("av_new_stream"));
 415
 416                 auto encoder = avcodec_find_encoder(output_format_.vcodec);
 417                 if (!encoder)
 418                         CASPAR_THROW_EXCEPTION(caspar_exception() << msg_info("Codec not found."));
 419
 420                 auto c = st->codec;
 421
 422                 avcodec_get_context_defaults3(c, encoder);
 423
 424                 c->codec_id                     = output_format_.vcodec;
 425                 c->codec_type           = AVMEDIA_TYPE_VIDEO;
 426                 c->width                        = output_format_.width;
 427                 c->height                       = output_format_.height - output_format_.croptop - output_format_.cropbot;
 428                 c->time_base.den        = format_desc_.time_scale;
 429                 c->time_base.num        = format_desc_.duration;
 430                 c->gop_size                     = 25;
 431                 c->flags                   |= format_desc_.field_mode == core::field_mode::progressive ? 0 : (CODEC_FLAG_INTERLACED_ME | CODEC_FLAG_INTERLACED_DCT);
 432                 c->pix_fmt                      = c->pix_fmt != PIX_FMT_NONE ? c->pix_fmt : PIX_FMT_YUV420P;
 433
 434                 if(c->codec_id == CODEC_ID_PRORES)
 435                 {
 436                         c->bit_rate     = output_format_.width < 1280 ? 63*1000000 : 220*1000000;
 437                         c->pix_fmt      = PIX_FMT_YUV422P10;
 438                 }
 439                 else if(c->codec_id == CODEC_ID_DNXHD)
 440                 {
 441                         if(c->width < 1280 || c->height < 720)
 442                                 CASPAR_THROW_EXCEPTION(caspar_exception() << msg_info("Unsupported video dimensions."));
 443
 444                         c->bit_rate     = 220*1000000;
 445                         c->pix_fmt      = PIX_FMT_YUV422P;
 446                 }
 447                 else if(c->codec_id == CODEC_ID_DVVIDEO)
 448                 {
 449                         c->width = c->height == 1280 ? 960  : c->width;
 450
 451                         if(format_desc_.format == core::video_format::ntsc)
 452                         {
 453                                 c->pix_fmt = PIX_FMT_YUV411P;
 454                                 output_format_.croptop = 2;
 455                                 output_format_.cropbot = 4;
 456                                 c->height                          = output_format_.height - output_format_.croptop - output_format_.cropbot;
 457                         }
 458                         else if(format_desc_.format == core::video_format::pal)
 459                                 c->pix_fmt = PIX_FMT_YUV420P;
 460                         else // dv50
 461                                 c->pix_fmt = PIX_FMT_YUV422P;
 462
 463                         if(format_desc_.duration == 1001)
 464                                 c->width = c->height == 1080 ? 1280 : c->width;
 465                         else
 466                                 c->width = c->height == 1080 ? 1440 : c->width;
 467                 }
 468                 else if(c->codec_id == CODEC_ID_H264)
 469                 {
 470                         c->pix_fmt = PIX_FMT_YUV420P;
 471                         av_opt_set(c->priv_data, "preset", "ultrafast", 0);
 472                         av_opt_set(c->priv_data, "tune",   "fastdecode",   0);
 473                         av_opt_set(c->priv_data, "crf",    "5",     0);
 474                 }
 475                 else if(c->codec_id == CODEC_ID_QTRLE)
 476                 {
 477                         c->pix_fmt = PIX_FMT_ARGB;
 478                 }
 479
 480                 boost::range::remove_erase_if(options, [&](const option& o)
 481                 {
 482                         return o.name.at(0) != 'a' && ffmpeg::av_opt_set(c, o.name.c_str(), o.value.c_str(), AV_OPT_SEARCH_CHILDREN) > -1;
 483                 });
 484
 485                 if(output_format_.format->flags & AVFMT_GLOBALHEADER)
 486                         c->flags |= CODEC_FLAG_GLOBAL_HEADER;
 487
 488                 THROW_ON_ERROR2(tbb_avcodec_open(c, encoder, false), "[ffmpeg_consumer]");
 489
 490                 return std::shared_ptr<AVStream>(st, [](AVStream* st)
 491                 {
 492                         LOG_ON_ERROR2(tbb_avcodec_close(st->codec), "[ffmpeg_consumer]");
 493                 });
 494         }
 495
 496         std::shared_ptr<AVStream> add_audio_stream(std::vector<option>& options)
 497         {
 498                 if(output_format_.acodec == CODEC_ID_NONE)
 499                         return nullptr;
 500
 501                 auto st = avformat_new_stream(oc_.get(), nullptr);
 502                 if(!st)
 503                         CASPAR_THROW_EXCEPTION(caspar_exception() << msg_info("Could not allocate audio-stream") << boost::errinfo_api_function("av_new_stream"));
 504
 505                 auto encoder = avcodec_find_encoder(output_format_.acodec);
 506                 if (!encoder)
 507                         CASPAR_THROW_EXCEPTION(caspar_exception() << msg_info("codec not found"));
 508
 509                 auto c = st->codec;
 510
 511                 avcodec_get_context_defaults3(c, encoder);
 512
 513                 c->codec_id                     = output_format_.acodec;
 514                 c->codec_type           = AVMEDIA_TYPE_AUDIO;
 515                 c->sample_rate          = 48000;
 516                 c->channels                     = 2;
 517                 c->sample_fmt           = AV_SAMPLE_FMT_S16;
 518                 c->time_base.num        = 1;
 519                 c->time_base.den        = c->sample_rate;
 520
 521                 if(output_format_.vcodec == CODEC_ID_FLV1)
 522                         c->sample_rate  = 44100;
 523
 524                 if(output_format_.format->flags & AVFMT_GLOBALHEADER)
 525                         c->flags |= CODEC_FLAG_GLOBAL_HEADER;
 526
 527                 boost::range::remove_erase_if(options, [&](const option& o)
 528                 {
 529                         return ffmpeg::av_opt_set(c, o.name.c_str(), o.value.c_str(), AV_OPT_SEARCH_CHILDREN) > -1;
 530                 });
 531
 532                 THROW_ON_ERROR2(avcodec_open2(c, encoder, nullptr), "[ffmpeg_consumer]");
 533
 534                 return std::shared_ptr<AVStream>(st, [](AVStream* st)
 535                 {
 536                         LOG_ON_ERROR2(avcodec_close(st->codec), "[ffmpeg_consumer]");
 537                 });
 538         }
 539
 540         void encode_video_frame(core::const_frame frame)
 541         {
 542                 if(!video_st_)
 543                         return;
 544
 545                 auto enc = video_st_->codec;
 546
 547                 auto av_frame                           = convert_video(frame, enc);
 548                 av_frame->interlaced_frame      = format_desc_.field_mode != core::field_mode::progressive;
 549                 av_frame->top_field_first       = format_desc_.field_mode == core::field_mode::upper;
 550                 av_frame->pts = frame_number_++;
 551
 552                 monitor_subject_
 553                         << core::monitor::message("/frame") % static_cast<int64_t>(frame_number_)
 554                         << core::monitor::message("/path") % filename_
 555                         << core::monitor::message("/fps") % format_desc_.fps;
 556
 557                 AVPacket pkt;
 558                 av_init_packet(&pkt);
 559                 pkt.data = nullptr;
 560                 pkt.size = 0;
 561
 562                 int got_packet = 0;
 563                 THROW_ON_ERROR2(avcodec_encode_video2(enc, &pkt, av_frame.get(), &got_packet), "[ffmpeg_consumer]");
 564                 std::shared_ptr<AVPacket> guard(&pkt, av_free_packet);
 565
 566                 if(!got_packet)
 567                         return;
 568
 569                 if (pkt.pts != AV_NOPTS_VALUE)
 570                         pkt.pts = av_rescale_q(pkt.pts, enc->time_base, video_st_->time_base);
 571                 if (pkt.dts != AV_NOPTS_VALUE)
 572                         pkt.dts = av_rescale_q(pkt.dts, enc->time_base, video_st_->time_base);
 573
 574                 pkt.stream_index = video_st_->index;
 575
 576                 THROW_ON_ERROR2(av_interleaved_write_frame(oc_.get(), &pkt), "[ffmpeg_consumer]");
 577         }
 578
 579         uint64_t get_channel_layout(AVCodecContext* dec)
 580         {
 581                 auto layout = (dec->channel_layout && dec->channels == av_get_channel_layout_nb_channels(dec->channel_layout)) ? dec->channel_layout : av_get_default_channel_layout(dec->channels);
 582                 return layout;
 583         }
 584
 585         void encode_audio_frame(core::const_frame frame)
 586         {
 587                 if(!audio_st_)
 588                         return;
 589
 590                 auto enc = audio_st_->codec;
 591
 592                 boost::push_back(audio_buffer_, convert_audio(frame, enc));
 593
 594                 auto frame_size = enc->frame_size != 0 ? enc->frame_size * enc->channels * av_get_bytes_per_sample(enc->sample_fmt) : static_cast<int>(audio_buffer_.size());
 595
 596                 while(audio_buffer_.size() >= frame_size)
 597                 {
 598                         std::shared_ptr<AVFrame> av_frame(av_frame_alloc(), [=](AVFrame* p) { av_frame_free(&p); });
 599                         avcodec_get_frame_defaults(av_frame.get());
 600                         av_frame->nb_samples = frame_size / (enc->channels * av_get_bytes_per_sample(enc->sample_fmt));
 601
 602                         AVPacket pkt;
 603                         av_init_packet(&pkt);
 604                         pkt.data = nullptr;
 605                         pkt.size = 0;
 606
 607                         THROW_ON_ERROR2(avcodec_fill_audio_frame(av_frame.get(), enc->channels, enc->sample_fmt, audio_buffer_.data(), frame_size, 1), "[ffmpeg_consumer]");
 608
 609                         int got_packet = 0;
 610                         THROW_ON_ERROR2(avcodec_encode_audio2(enc, &pkt, av_frame.get(), &got_packet), "[ffmpeg_consumer]");
 611                         std::shared_ptr<AVPacket> guard(&pkt, av_free_packet);
 612
 613                         audio_buffer_.erase(audio_buffer_.begin(), audio_buffer_.begin() + frame_size);
 614
 615                         if(!got_packet)
 616                                 return;
 617
 618                         if (pkt.pts != AV_NOPTS_VALUE)
 619                                 pkt.pts      = av_rescale_q(pkt.pts, enc->time_base, audio_st_->time_base);
 620                         if (pkt.dts != AV_NOPTS_VALUE)
 621                                 pkt.dts      = av_rescale_q(pkt.dts, enc->time_base, audio_st_->time_base);
 622                         if (pkt.duration > 0)
 623                                 pkt.duration = static_cast<int>(av_rescale_q(pkt.duration, enc->time_base, audio_st_->time_base));
 624
 625                         pkt.stream_index = audio_st_->index;
 626
 627                         THROW_ON_ERROR2(av_interleaved_write_frame(oc_.get(), &pkt), "[ffmpeg_consumer]");
 628                 }
 629         }
 630
 631         std::shared_ptr<AVFrame> convert_video(core::const_frame frame, AVCodecContext* c)
 632         {
 633                 if(!sws_)
 634                 {
 635                         sws_.reset(sws_getContext(format_desc_.width,
 636                                                                           format_desc_.height - output_format_.croptop  - output_format_.cropbot,
 637                                                                           PIX_FMT_BGRA,
 638                                                                           c->width,
 639                                                                           c->height,
 640                                                                           c->pix_fmt,
 641                                                                           SWS_BICUBIC, nullptr, nullptr, nullptr),
 642                                                 sws_freeContext);
 643                         if (sws_ == nullptr)
 644                                 CASPAR_THROW_EXCEPTION(caspar_exception() << msg_info("Cannot initialize the conversion context"));
 645                 }
 646
 647                 // #in_frame
 648
 649                 std::shared_ptr<AVFrame> in_frame(avcodec_alloc_frame(), av_free);
 650
 651                 auto in_picture = reinterpret_cast<AVPicture*>(in_frame.get());
 652
 653                 if (key_only_)
 654                 {
 655                         key_picture_buf_.resize(frame.image_data().size());
 656                         in_picture->linesize[0] = format_desc_.width * 4;
 657                         in_picture->data[0] = key_picture_buf_.data();
 658
 659                         aligned_memshfl(in_picture->data[0], frame.image_data().begin(), frame.image_data().size(), 0x0F0F0F0F, 0x0B0B0B0B, 0x07070707, 0x03030303);
 660                 }
 661                 else
 662                 {
 663                         avpicture_fill(
 664                                         in_picture,
 665                                         const_cast<uint8_t*>(frame.image_data().begin()),
 666                                         PIX_FMT_BGRA,
 667                                         format_desc_.width,
 668                                         format_desc_.height - output_format_.croptop  - output_format_.cropbot);
 669                 }
 670
 671                 // crop-top
 672
 673                 for(int n = 0; n < 4; ++n)
 674                         in_frame->data[n] += in_frame->linesize[n] * output_format_.croptop;
 675
 676                 // #out_frame
 677
 678                 std::shared_ptr<AVFrame> out_frame(avcodec_alloc_frame(), av_free);
 679
 680                 av_image_fill_linesizes(out_frame->linesize, c->pix_fmt, c->width);
 681                 for(int n = 0; n < 4; ++n)
 682                         out_frame->linesize[n] += 32 - (out_frame->linesize[n] % 32); // align
 683
 684                 picture_buffer_.resize(av_image_fill_pointers(out_frame->data, c->pix_fmt, c->height, nullptr, out_frame->linesize));
 685                 av_image_fill_pointers(out_frame->data, c->pix_fmt, c->height, picture_buffer_.data(), out_frame->linesize);
 686
 687                 // #scale
 688
 689                 sws_scale(sws_.get(),
 690                                   in_frame->data,
 691                                   in_frame->linesize,
 692                                   0,
 693                                   format_desc_.height - output_format_.cropbot - output_format_.croptop,
 694                                   out_frame->data,
 695                                   out_frame->linesize);
 696
 697                 out_frame->format       = c->pix_fmt;
 698                 out_frame->width        = c->width;
 699                 out_frame->height       = c->height;
 700
 701                 return out_frame;
 702         }
 703
 704         byte_vector convert_audio(core::const_frame& frame, AVCodecContext* c)
 705         {
 706                 if(!swr_)
 707                 {
 708                         swr_ = std::shared_ptr<SwrContext>(swr_alloc_set_opts(nullptr,
 709                                                                                 get_channel_layout(c), c->sample_fmt, c->sample_rate,
 710                                                                                 av_get_default_channel_layout(channel_layout_.num_channels), AV_SAMPLE_FMT_S32, format_desc_.audio_sample_rate,
 711                                                                                 0, nullptr), [](SwrContext* p){swr_free(&p);});
 712
 713                         if(!swr_)
 714                                 CASPAR_THROW_EXCEPTION(bad_alloc());
 715
 716                         THROW_ON_ERROR2(swr_init(swr_.get()), "[audio_decoder]");
 717                 }
 718
 719                 byte_vector buffer(48000);
 720
 721                 const uint8_t* in[]  = {reinterpret_cast<const uint8_t*>(frame.audio_data().data())};
 722                 uint8_t*       out[] = {buffer.data()};
 723
 724                 auto channel_samples = swr_convert(swr_.get(),
 725                                                                                    out, static_cast<int>(buffer.size()) / c->channels / av_get_bytes_per_sample(c->sample_fmt),
 726                                                                                    in, static_cast<int>(frame.audio_data().size()/channel_layout_.num_channels));
 727
 728                 buffer.resize(channel_samples * c->channels * av_get_bytes_per_sample(c->sample_fmt));
 729
 730                 return buffer;
 731         }
 732
 733         void check_space()
 734         {
 735                 auto space = boost::filesystem::space(boost::filesystem::path(full_filename_).parent_path());
 736                 if(space.available < 512*1000000)
 737                         CASPAR_THROW_EXCEPTION(file_write_error() << msg_info("out of space"));
 738         }
 739
 740         void encode(const core::const_frame& frame)
 741         {
 742                 try
 743                 {
 744                         if(frame_number_ % 25 == 0)
 745                                 check_space();
 746
 747                         caspar::timer frame_timer;
 748
 749                         encode_video_frame(frame);
 750                         encode_audio_frame(frame);
 751
 752                         graph_->set_value("frame-time", frame_timer.elapsed()*format_desc_.fps*0.5);
 753                 }
 754                 catch(...)
 755                 {
 756                         lock(exception_mutex_, [&]
 757                         {
 758                                 exception_ = std::current_exception();
 759                         });
 760                 }
 761         }
 762 };
 763
 764 struct ffmpeg_consumer_proxy : public core::frame_consumer
 765 {
 766         const std::wstring                      filename_;
 767         const std::vector<option>       options_;
 768         const bool                                      separate_key_;
 769
 770         std::unique_ptr<ffmpeg_consumer> consumer_;
 771         std::unique_ptr<ffmpeg_consumer> key_only_consumer_;
 772
 773 public:
 774
 775         ffmpeg_consumer_proxy(const std::wstring& filename, const std::vector<option>& options, bool separate_key)
 776                 : filename_(filename)
 777                 , options_(options)
 778                 , separate_key_(separate_key)
 779         {
 780         }
 781
 782         void initialize(const core::video_format_desc& format_desc, const core::audio_channel_layout& channel_layout, int) override
 783         {
 784                 if(consumer_)
 785                         CASPAR_THROW_EXCEPTION(invalid_operation() << msg_info("Cannot reinitialize ffmpeg-consumer."));
 786
 787                 consumer_.reset(new ffmpeg_consumer(u8(filename_), format_desc, channel_layout, options_, false));
 788
 789                 if (separate_key_)
 790                 {
 791                         boost::filesystem::path fill_file(filename_);
 792                         auto without_extension = u16(fill_file.stem().string());
 793                         auto key_file = without_extension + L"_A" + u16(fill_file.extension().string());
 794
 795                         key_only_consumer_.reset(new ffmpeg_consumer(u8(key_file), format_desc, channel_layout, options_, true));
 796                 }
 797         }
 798
 799         int64_t presentation_frame_age_millis() const override
 800         {
 801                 return consumer_ ? static_cast<int64_t>(consumer_->current_encoding_delay_) : 0;
 802         }
 803
 804         std::future<bool> send(core::const_frame frame) override
 805         {
 806                 bool ready_for_frame = consumer_->ready_for_frame();
 807
 808                 if (ready_for_frame && separate_key_)
 809                         ready_for_frame = ready_for_frame && key_only_consumer_->ready_for_frame();
 810
 811                 if (ready_for_frame)
 812                 {
 813                         consumer_->send(frame);
 814
 815                         if (separate_key_)
 816                                 key_only_consumer_->send(frame);
 817                 }
 818                 else
 819                 {
 820                         consumer_->mark_dropped();
 821
 822                         if (separate_key_)
 823                                 key_only_consumer_->mark_dropped();
 824                 }
 825
 826                 return make_ready_future(true);
 827         }
 828
 829         std::wstring print() const override
 830         {
 831                 return consumer_ ? consumer_->print() : L"[ffmpeg_consumer]";
 832         }
 833
 834         std::wstring name() const override
 835         {
 836                 return L"file";
 837         }
 838
 839         boost::property_tree::wptree info() const override
 840         {
 841                 boost::property_tree::wptree info;
 842                 info.add(L"type", L"file");
 843                 info.add(L"filename", filename_);
 844                 info.add(L"separate_key", separate_key_);
 845                 return info;
 846         }
 847
 848         bool has_synchronization_clock() const override
 849         {
 850                 return false;
 851         }
 852
 853         int buffer_depth() const override
 854         {
 855                 return -1;
 856         }
 857
 858         int index() const override
 859         {
 860                 return 200;
 861         }
 862
 863         core::monitor::subject& monitor_output()
 864         {
 865                 return consumer_->monitor_output();
 866         }
 867 };
 868
 869 void describe_consumer(core::help_sink& sink, const core::help_repository& repo)
 870 {
 871         sink.short_description(L"Can record a channel to a file supported by FFmpeg.");
 872         sink.syntax(L"FILE [filename:string] {-[ffmpeg_param1:string] [value1:string] {-[ffmpeg_param2:string] [value2:string] {...}}} {[separate_key:SEPARATE_KEY]}");
 873         sink.para()->text(L"Can record a channel to a file supported by FFmpeg.");
 874         sink.definitions()
 875                 ->item(L"filename", L"The filename under the media folder including the extension (decides which kind of container format that will be used).")
 876                 ->item(L"ffmpeg_paramX", L"A parameter supported by FFmpeg. For example vcodec or acodec etc.")
 877                 ->item(L"separate_key", L"If defined will create two files simultaneously -- One for fill and one for key (_A will be appended).")
 878                 ;
 879         sink.para()->text(L"Examples:");
 880         sink.example(L">> ADD 1 FILE output.mov -vcodec dnxhd");
 881         sink.example(L">> ADD 1 FILE output.mov -vcodec prores");
 882         sink.example(L">> ADD 1 FILE output.mov -vcodec dvvideo");
 883         sink.example(L">> ADD 1 FILE output.mov -vcodec libx264 -preset ultrafast -tune fastdecode -crf 25");
 884         sink.example(L">> ADD 1 FILE output.mov -vcodec dnxhd SEPARATE_KEY", L"for creating output.mov with fill and output_A.mov with key/alpha");
 885 }
 886
 887 spl::shared_ptr<core::frame_consumer> create_consumer(
 888                 const std::vector<std::wstring>& params, core::interaction_sink*)
 889 {
 890         auto params2 = params;
 891         auto separate_key_it = std::find_if(params2.begin(), params2.end(), param_comparer(L"SEPARATE_KEY"));
 892         bool separate_key = false;
 893
 894         if (separate_key_it != params2.end())
 895         {
 896                 separate_key = true;
 897                 params2.erase(separate_key_it);
 898         }
 899
 900         auto str = std::accumulate(params2.begin(), params2.end(), std::wstring(), [](const std::wstring& lhs, const std::wstring& rhs) {return lhs + L" " + rhs;});
 901
 902         boost::wregex path_exp(LR"(\s*FILE(\s(?<PATH>.+\.[^\s]+))?.*)", boost::regex::icase);
 903
 904         boost::wsmatch path;
 905         if(!boost::regex_match(str, path, path_exp))
 906                 return core::frame_consumer::empty();
 907
 908         boost::wregex opt_exp(LR"(-((?<NAME>[^\s]+)\s+(?<VALUE>[^\s]+)))");
 909
 910         std::vector<option> options;
 911         for(boost::wsregex_iterator it(str.begin(), str.end(), opt_exp); it != boost::wsregex_iterator(); ++it)
 912         {
 913                 auto name  = u8(boost::trim_copy(boost::to_lower_copy((*it)["NAME"].str())));
 914                 auto value = u8(boost::trim_copy(boost::to_lower_copy((*it)["VALUE"].str())));
 915
 916                 if(value == "h264")
 917                         value = "libx264";
 918                 else if(value == "dvcpro")
 919                         value = "dvvideo";
 920
 921                 options.push_back(option(name, value));
 922         }
 923
 924         return spl::make_shared<ffmpeg_consumer_proxy>(path["PATH"].str(), options, separate_key);
 925 }
 926
 927 spl::shared_ptr<core::frame_consumer> create_preconfigured_consumer(
 928                 const boost::property_tree::wptree& ptree, core::interaction_sink*)
 929 {
 930         auto filename           = ptree_get<std::wstring>(ptree, L"path");
 931         auto codec                      = ptree.get(L"vcodec", L"libx264");
 932         auto separate_key       = ptree.get(L"separate-key", false);
 933
 934         std::vector<option> options;
 935         options.push_back(option("vcodec", u8(codec)));
 936
 937         return spl::make_shared<ffmpeg_consumer_proxy>(filename, options, separate_key);
 938 }
 939
 940 }}