/*\r
-* copyright (c) 2010 Sveriges Television AB <info@casparcg.com>\r
+* Copyright (c) 2011 Sveriges Television AB <info@casparcg.com>\r
*\r
-* This ffmpeg is part of CasparCG.\r
+* This file is part of CasparCG (www.casparcg.com).\r
*\r
-* CasparCG is free software: you can redistribute it and/or modify\r
-* it under the terms of the GNU General Public License as published by\r
-* the Free Software Foundation, either version 3 of the License, or\r
-* (at your option) any later version.\r
+* CasparCG is free software: you can redistribute it and/or modify\r
+* it under the terms of the GNU General Public License as published by\r
+* the Free Software Foundation, either version 3 of the License, or\r
+* (at your option) any later version.\r
*\r
-* CasparCG is distributed in the hope that it will be useful,\r
-* but WITHOUT ANY WARRANTY; without even the implied warranty of\r
-* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the\r
-* GNU General Public License for more details.\r
-\r
-* You should have received a copy of the GNU General Public License\r
-* along with CasparCG. If not, see <http://www.gnu.org/licenses/>.\r
+* CasparCG is distributed in the hope that it will be useful,\r
+* but WITHOUT ANY WARRANTY; without even the implied warranty of\r
+* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the\r
+* GNU General Public License for more details.\r
+*\r
+* You should have received a copy of the GNU General Public License\r
+* along with CasparCG. If not, see <http://www.gnu.org/licenses/>.\r
*\r
+* Author: Robert Nagy, ronag89@gmail.com\r
*/\r
\r
#include "../StdAfx.h"\r
\r
-#include "ffmpeg_consumer.h"\r
-\r
-#include <core/mixer/read_frame.h>\r
+#include "../ffmpeg_error.h"\r
\r
-#include <common/concurrency/executor.h>\r
-#include <common/utility/string.h>\r
-#include <common/env.h>\r
+#include "ffmpeg_consumer.h"\r
\r
-#include <boost/thread/once.hpp>\r
+#include "../producer/audio/audio_resampler.h"\r
+#include "../producer/tbb_avcodec.h"\r
\r
-#include <tbb/cache_aligned_allocator.h>\r
-#include <tbb/parallel_invoke.h>\r
+#include <core/frame/frame.h>\r
+#include <core/mixer/audio/audio_util.h>\r
+#include <core/consumer/frame_consumer.h>\r
+#include <core/video_format.h>\r
\r
-#include <cstdio>\r
+#include <common/env.h>\r
+#include <common/utf.h>\r
+#include <common/param.h>\r
+#include <common/executor.h>\r
+#include <common/diagnostics/graph.h>\r
+#include <common/array.h>\r
+#include <common/memory.h>\r
+\r
+#include <boost/algorithm/string.hpp>\r
+#include <boost/timer.hpp>\r
+#include <boost/property_tree/ptree.hpp>\r
+#include <boost/filesystem.hpp>\r
+#include <boost/range/algorithm.hpp>\r
+#include <boost/range/algorithm_ext.hpp>\r
+#include <boost/lexical_cast.hpp>\r
\r
#if defined(_MSC_VER)\r
#pragma warning (push)\r
#define __STDC_LIMIT_MACROS\r
#include <libavformat/avformat.h>\r
#include <libswscale/swscale.h>\r
+ #include <libavutil/opt.h>\r
+ #include <libavutil/pixdesc.h>\r
+ #include <libavutil/parseutils.h>\r
+ #include <libavutil/samplefmt.h>\r
}\r
#if defined(_MSC_VER)\r
#pragma warning (pop)\r
#endif\r
\r
-namespace caspar { \r
+namespace caspar { namespace ffmpeg {\r
\r
-struct ffmpeg_consumer::implementation : boost::noncopyable\r
-{ \r
- std::string filename_;\r
+int av_opt_set(void *obj, const char *name, const char *val, int search_flags)\r
+{\r
+ AVClass* av_class = *(AVClass**)obj;\r
\r
- // Audio\r
- AVStream* audio_st_;\r
- std::vector<unsigned char, tbb::cache_aligned_allocator<unsigned char>> audio_outbuf_;\r
+ if((strcmp(name, "pix_fmt") == 0 || strcmp(name, "pixel_format") == 0) && strcmp(av_class->class_name, "AVCodecContext") == 0)\r
+ {\r
+ AVCodecContext* c = (AVCodecContext*)obj; \r
+ auto pix_fmt = av_get_pix_fmt(val);\r
+ if(pix_fmt == PIX_FMT_NONE)\r
+ return -1; \r
+ c->pix_fmt = pix_fmt;\r
+ return 0;\r
+ }\r
+ if((strcmp(name, "r") == 0 || strcmp(name, "frame_rate") == 0) && strcmp(av_class->class_name, "AVCodecContext") == 0)\r
+ {\r
+ AVCodecContext* c = (AVCodecContext*)obj; \r
\r
- // Video\r
- AVStream* video_st_;\r
- std::vector<uint8_t, tbb::cache_aligned_allocator<unsigned char>> picture_buf_;\r
+ if(c->codec_type != AVMEDIA_TYPE_VIDEO)\r
+ return -1;\r
\r
- std::vector<unsigned char, tbb::cache_aligned_allocator<unsigned char>> video_outbuf_;\r
- SwsContext* img_convert_ctx_;\r
- \r
- AVOutputFormat* fmt_;\r
- std::shared_ptr<AVFormatContext> oc_;\r
- core::video_format_desc format_desc_;\r
+ AVRational rate;\r
+ int ret = av_parse_video_rate(&rate, val);\r
+ if(ret < 0)\r
+ return ret;\r
\r
- std::vector<short, tbb::cache_aligned_allocator<short>> audio_input_buffer_;\r
+ c->time_base.num = rate.den;\r
+ c->time_base.den = rate.num;\r
+ return 0;\r
+ }\r
\r
- boost::unique_future<void> active_;\r
+ return ::av_opt_set(obj, name, val, search_flags);\r
+}\r
\r
- executor executor_;\r
-public:\r
- implementation(const std::string& filename)\r
- : filename_(filename)\r
- , audio_st_(nullptr)\r
- , video_st_(nullptr)\r
- , fmt_(nullptr)\r
- , img_convert_ctx_(nullptr)\r
- , video_outbuf_(1920*1080*4)\r
- , audio_outbuf_(48000)\r
- , executor_(L"ffmpeg_consumer")\r
- {}\r
-\r
- ~implementation()\r
- { \r
- executor_.invoke([]{});\r
- executor_.stop();\r
+struct option\r
+{\r
+ std::string name;\r
+ std::string value;\r
\r
- av_write_trailer(oc_.get());\r
+ option(std::string name, std::string value)\r
+ : name(std::move(name))\r
+ , value(std::move(value))\r
+ {\r
+ }\r
+};\r
+ \r
+struct output_format\r
+{\r
+ AVOutputFormat* format;\r
+ int width;\r
+ int height;\r
+ CodecID vcodec;\r
+ CodecID acodec;\r
+ int croptop;\r
+ int cropbot;\r
+\r
+ output_format(const core::video_format_desc& format_desc, const std::string& filename, std::vector<option>& options)\r
+ : format(av_guess_format(nullptr, filename.c_str(), nullptr))\r
+ , width(format_desc.width)\r
+ , height(format_desc.height)\r
+ , vcodec(CODEC_ID_NONE)\r
+ , acodec(CODEC_ID_NONE)\r
+ , croptop(0)\r
+ , cropbot(0)\r
+ {\r
+ boost::range::remove_erase_if(options, [&](const option& o)\r
+ {\r
+ return set_opt(o.name, o.value);\r
+ });\r
+ \r
+ if(vcodec == CODEC_ID_NONE)\r
+ vcodec = format->video_codec;\r
\r
- // Close each codec.\r
- if (video_st_) \r
- avcodec_close(video_st_->codec);\r
+ if(acodec == CODEC_ID_NONE)\r
+ acodec = format->audio_codec;\r
\r
- if (audio_st_)\r
- avcodec_close(audio_st_->codec);\r
- \r
- // Free the streams.\r
- for(size_t i = 0; i < oc_->nb_streams; ++i) \r
+ if(vcodec == CODEC_ID_NONE)\r
+ vcodec = CODEC_ID_H264;\r
+ \r
+ if(acodec == CODEC_ID_NONE)\r
+ acodec = CODEC_ID_PCM_S16LE;\r
+ }\r
+ \r
+ bool set_opt(const std::string& name, const std::string& value)\r
+ {\r
+ //if(name == "target")\r
+ //{ \r
+ // enum { PAL, NTSC, FILM, UNKNOWN } norm = UNKNOWN;\r
+ // \r
+ // if(name.find("pal-") != std::string::npos)\r
+ // norm = PAL;\r
+ // else if(name.find("ntsc-") != std::string::npos)\r
+ // norm = NTSC;\r
+\r
+ // if(norm == UNKNOWN)\r
+ // BOOST_THROW_EXCEPTION(invalid_argument() << arg_name_info("target"));\r
+ // \r
+ // if (name.find("-dv") != std::string::npos) \r
+ // {\r
+ // set_opt("f", "dv");\r
+ // set_opt("s", norm == PAL ? "720x576" : "720x480");\r
+ // //set_opt("pix_fmt", name.find("-dv50") != std::string::npos ? "yuv422p" : norm == PAL ? "yuv420p" : "yuv411p");\r
+ // //set_opt("ar", "48000");\r
+ // //set_opt("ac", "2");\r
+ // } \r
+ //}\r
+ if(name == "f")\r
{\r
- av_freep(&oc_->streams[i]->codec);\r
- av_freep(&oc_->streams[i]);\r
- }\r
+ format = av_guess_format(value.c_str(), nullptr, nullptr);\r
\r
- if (!(fmt_->flags & AVFMT_NOFILE)) \r
- url_fclose(oc_->pb); // Close the output ffmpeg.\r
- }\r
+ if(format == nullptr)\r
+ BOOST_THROW_EXCEPTION(invalid_argument() << arg_name_info("f"));\r
\r
- void initialize(const core::video_format_desc& format_desc)\r
- {\r
- format_desc_ = format_desc;\r
- executor_.start();\r
- active_ = executor_.begin_invoke([]{});\r
+ return true;\r
+ }\r
+ else if(name == "vcodec")\r
+ {\r
+ auto c = avcodec_find_encoder_by_name(value.c_str());\r
+ if(c == nullptr)\r
+ BOOST_THROW_EXCEPTION(invalid_argument() << arg_name_info("vcodec"));\r
\r
- fmt_ = av_guess_format(nullptr, filename_.c_str(), nullptr);\r
- if (!fmt_) \r
+ vcodec = avcodec_find_encoder_by_name(value.c_str())->id;\r
+ return true;\r
+\r
+ }\r
+ else if(name == "acodec")\r
{\r
- CASPAR_LOG(warning) << "Could not deduce output format from ffmpeg extension: using MPEG.";\r
- fmt_ = av_guess_format("mpeg", nullptr, nullptr);\r
- filename_ = filename_ + ".avi"; \r
+ auto c = avcodec_find_encoder_by_name(value.c_str());\r
+ if(c == nullptr)\r
+ BOOST_THROW_EXCEPTION(invalid_argument() << arg_name_info("acodec"));\r
+\r
+ acodec = avcodec_find_encoder_by_name(value.c_str())->id;\r
+\r
+ return true;\r
}\r
- if (!fmt_)\r
- BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Could not find suitable output format"));\r
- \r
- oc_.reset(avformat_alloc_context(), av_free);\r
- if (!oc_)\r
- BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Memory error"));\r
- std::copy_n(filename_.c_str(), filename_.size(), oc_->filename);\r
-\r
- oc_->oformat = fmt_;\r
- // To avoid mpeg buffer underflow (http://www.mail-archive.com/libav-user@mplayerhq.hu/msg00194.html)\r
- oc_->preload = static_cast<int>(0.5*AV_TIME_BASE);\r
- oc_->max_delay = static_cast<int>(0.7*AV_TIME_BASE);\r
+ else if(name == "s")\r
+ {\r
+ if(av_parse_video_size(&width, &height, value.c_str()) < 0)\r
+ BOOST_THROW_EXCEPTION(invalid_argument() << arg_name_info("s"));\r
\r
- // Add the audio and video streams using the default format codecs and initialize the codecs .\r
- if (fmt_->video_codec != CODEC_ID_NONE) \r
- video_st_ = add_video_stream(fmt_->video_codec);\r
- \r
- if (fmt_->audio_codec != CODEC_ID_NONE) \r
- audio_st_ = add_audio_stream(fmt_->audio_codec); \r
-\r
- // Set the output parameters (must be done even if no parameters). \r
- int errn = 0;\r
- if ((errn = -av_set_parameters(oc_.get(), nullptr)) > 0)\r
- BOOST_THROW_EXCEPTION(\r
- file_read_error() << \r
- msg_info("Invalid output format parameters") <<\r
- boost::errinfo_api_function("avcodec_open") <<\r
- boost::errinfo_errno(errn) <<\r
- boost::errinfo_file_name(filename_));\r
- \r
- dump_format(oc_.get(), 0, filename_.c_str(), 1);\r
-\r
- // Now that all the parameters are set, we can open the audio and\r
- // video codecs and allocate the necessary encode buffers.\r
- if (video_st_)\r
- open_video(video_st_);\r
- \r
- try\r
+ return true;\r
+ }\r
+ else if(name == "croptop")\r
{\r
- if (audio_st_)\r
- open_audio(audio_st_);\r
+ croptop = boost::lexical_cast<int>(value);\r
+\r
+ return true;\r
}\r
- catch(...)\r
+ else if(name == "cropbot")\r
{\r
- CASPAR_LOG_CURRENT_EXCEPTION();\r
- audio_st_ = nullptr;\r
+ cropbot = boost::lexical_cast<int>(value);\r
+\r
+ return true;\r
}\r
- \r
+ \r
+ return false;\r
+ }\r
+};\r
+\r
+typedef std::vector<uint8_t, tbb::cache_aligned_allocator<uint8_t>> byte_vector;\r
+\r
+struct ffmpeg_consumer : boost::noncopyable\r
+{ \r
+ const std::string filename_;\r
+ \r
+ const std::shared_ptr<AVFormatContext> oc_;\r
+ const core::video_format_desc format_desc_;\r
+ \r
+ const spl::shared_ptr<diagnostics::graph> graph_;\r
+\r
+ executor encode_executor_;\r
+ executor write_executor_;\r
+ \r
+ std::shared_ptr<AVStream> audio_st_;\r
+ std::shared_ptr<AVStream> video_st_;\r
+ \r
+ byte_vector audio_outbuf_;\r
+ byte_vector audio_buf_;\r
+ byte_vector video_outbuf_;\r
+ byte_vector picture_buf_;\r
+ std::shared_ptr<audio_resampler> swr_;\r
+ std::shared_ptr<SwsContext> sws_;\r
+\r
+ int64_t in_frame_number_;\r
+ int64_t out_frame_number_;\r
+\r
+ output_format output_format_;\r
+ \r
+public:\r
+ ffmpeg_consumer(const std::string& filename, const core::video_format_desc& format_desc, std::vector<option> options)\r
+ : filename_(filename)\r
+ , video_outbuf_(1920*1080*8)\r
+ , audio_outbuf_(10000)\r
+ , oc_(avformat_alloc_context(), av_free)\r
+ , format_desc_(format_desc)\r
+ , encode_executor_(print())\r
+ , write_executor_(print() + L"/output")\r
+ , in_frame_number_(0)\r
+ , out_frame_number_(0)\r
+ , output_format_(format_desc, filename, options)\r
+ {\r
+ // TODO: Ask stakeholders about case where file already exists.\r
+ boost::filesystem::remove(boost::filesystem::path(env::media_folder() + u16(filename))); // Delete the file if it exists\r
+\r
+ graph_->set_color("frame-time", diagnostics::color(0.1f, 1.0f, 0.1f));\r
+ graph_->set_text(print());\r
+ diagnostics::register_graph(graph_);\r
+\r
+ encode_executor_.set_capacity(8);\r
+ write_executor_.set_capacity(8);\r
+\r
+ oc_->oformat = output_format_.format;\r
+ \r
+ strcpy_s(oc_->filename, filename_.c_str());\r
+ \r
+ // Add the audio and video streams using the default format codecs and initialize the codecs.\r
+ video_st_ = add_video_stream(options);\r
+ audio_st_ = add_audio_stream(options);\r
+ \r
+ av_dump_format(oc_.get(), 0, filename_.c_str(), 1);\r
+ \r
// Open the output ffmpeg, if needed.\r
- if (!(fmt_->flags & AVFMT_NOFILE)) \r
+ if (!(oc_->oformat->flags & AVFMT_NOFILE)) \r
+ THROW_ON_ERROR2(avio_open(&oc_->pb, filename.c_str(), AVIO_FLAG_WRITE), "[ffmpeg_consumer]");\r
+ \r
+ THROW_ON_ERROR2(avformat_write_header(oc_.get(), nullptr), "[ffmpeg_consumer]");\r
+\r
+ if(options.size() > 0)\r
{\r
- int errn = 0;\r
- if ((errn = -url_fopen(&oc_->pb, filename_.c_str(), URL_WRONLY)) > 0) \r
- BOOST_THROW_EXCEPTION(\r
- file_not_found() << \r
- msg_info("Could not open file") <<\r
- boost::errinfo_api_function("url_fopen") <<\r
- boost::errinfo_errno(errn) <<\r
- boost::errinfo_file_name(filename_));\r
+ BOOST_FOREACH(auto& option, options)\r
+ CASPAR_LOG(warning) << L"Invalid option: -" << u16(option.name) << L" " << u16(option.value);\r
}\r
+\r
+ CASPAR_LOG(info) << print() << L" Successfully Initialized."; \r
+ }\r
+\r
+ ~ffmpeg_consumer()\r
+ { \r
+ encode_executor_.wait();\r
+ write_executor_.wait();\r
\r
- av_write_header(oc_.get()); // write the stream header, if any \r
+ LOG_ON_ERROR2(av_write_trailer(oc_.get()), "[ffmpeg_consumer]");\r
+ \r
+ audio_st_.reset();\r
+ video_st_.reset();\r
+ \r
+ if (!(oc_->oformat->flags & AVFMT_NOFILE)) \r
+ LOG_ON_ERROR2(avio_close(oc_->pb), "[ffmpeg_consumer]"); // Close the output ffmpeg.\r
\r
- CASPAR_LOG(info) << print() << L" Successfully initialized.";\r
+ CASPAR_LOG(info) << print() << L" Successfully Uninitialized."; \r
}\r
- \r
+ \r
std::wstring print() const\r
{\r
- return L"ffmpeg[" + widen(filename_) + L"]";\r
+ return L"ffmpeg[" + u16(filename_) + L"]";\r
}\r
\r
- AVStream* add_video_stream(enum CodecID codec_id)\r
+ std::shared_ptr<AVStream> add_video_stream(std::vector<option>& options)\r
{ \r
+ if(output_format_.vcodec == CODEC_ID_NONE)\r
+ return nullptr;\r
+\r
auto st = av_new_stream(oc_.get(), 0);\r
- if (!st) \r
- BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Could not alloc stream"));\r
- \r
+ if (!st) \r
+ BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Could not allocate video-stream.") << boost::errinfo_api_function("av_new_stream")); \r
+\r
+ auto encoder = avcodec_find_encoder(output_format_.vcodec);\r
+ if (!encoder)\r
+ BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Codec not found."));\r
+\r
auto c = st->codec;\r
- c->codec_id = codec_id;\r
- c->codec_type = AVMEDIA_TYPE_VIDEO;\r
- \r
- // Put sample parameters.\r
- c->bit_rate = static_cast<int>(static_cast<double>(format_desc_.size)*format_desc_.fps*0.1326);\r
- c->width = format_desc_.width;\r
- c->height = format_desc_.height;\r
- c->time_base.den = static_cast<int>(format_desc_.fps);\r
- c->time_base.num = 1;\r
- c->pix_fmt = c->pix_fmt == -1 ? PIX_FMT_YUV420P : c->pix_fmt;\r
-\r
- // Some formats want stream headers to be separate.\r
- if(oc_->oformat->flags & AVFMT_GLOBALHEADER)\r
+\r
+ avcodec_get_context_defaults3(c, encoder);\r
+ \r
+ c->codec_id = output_format_.vcodec;\r
+ c->codec_type = AVMEDIA_TYPE_VIDEO;\r
+ c->width = output_format_.width;\r
+ c->height = output_format_.height - output_format_.croptop - output_format_.cropbot;\r
+ c->time_base.den = format_desc_.time_scale;\r
+ c->time_base.num = format_desc_.duration;\r
+ c->gop_size = 25;\r
+ c->flags |= format_desc_.field_mode == core::field_mode::progressive ? 0 : (CODEC_FLAG_INTERLACED_ME | CODEC_FLAG_INTERLACED_DCT);\r
+ if(c->pix_fmt == PIX_FMT_NONE)\r
+ c->pix_fmt = PIX_FMT_YUV420P;\r
+\r
+ if(c->codec_id == CODEC_ID_PRORES)\r
+ { \r
+ c->bit_rate = c->width < 1280 ? 63*1000000 : 220*1000000;\r
+ c->pix_fmt = PIX_FMT_YUV422P10;\r
+ }\r
+ else if(c->codec_id == CODEC_ID_DNXHD)\r
+ {\r
+ if(c->width < 1280 || c->height < 720)\r
+ BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Unsupported video dimensions."));\r
+\r
+ c->bit_rate = 220*1000000;\r
+ c->pix_fmt = PIX_FMT_YUV422P;\r
+ }\r
+ else if(c->codec_id == CODEC_ID_DVVIDEO)\r
+ {\r
+ c->width = c->height == 1280 ? 960 : c->width;\r
+ \r
+ if(format_desc_.format == core::video_format::ntsc)\r
+ c->pix_fmt = PIX_FMT_YUV411P;\r
+ else if(format_desc_.format == core::video_format::pal)\r
+ c->pix_fmt = PIX_FMT_YUV420P;\r
+ else // dv50\r
+ c->pix_fmt = PIX_FMT_YUV422P;\r
+ \r
+ if(format_desc_.duration == 1001) \r
+ c->width = c->height == 1080 ? 1280 : c->width; \r
+ else\r
+ c->width = c->height == 1080 ? 1440 : c->width; \r
+ }\r
+ else if(c->codec_id == CODEC_ID_H264)\r
+ { \r
+ c->pix_fmt = PIX_FMT_YUV420P; \r
+ if(options.empty())\r
+ {\r
+ av_opt_set(c->priv_data, "preset", "ultrafast", 0);\r
+ av_opt_set(c->priv_data, "tune", "fastdecode", 0);\r
+ av_opt_set(c->priv_data, "crf", "5", 0);\r
+ }\r
+ }\r
+ else if(c->codec_id == CODEC_ID_QTRLE)\r
+ {\r
+ c->pix_fmt = PIX_FMT_ARGB;\r
+ }\r
+ \r
+ c->max_b_frames = 0; // b-frames not supported.\r
+ \r
+ boost::range::remove_erase_if(options, [&](const option& o)\r
+ {\r
+ return o.name.at(0) != 'a' && ffmpeg::av_opt_set(c, o.name.c_str(), o.value.c_str(), AV_OPT_SEARCH_CHILDREN) > -1;\r
+ });\r
+ \r
+ if(output_format_.format->flags & AVFMT_GLOBALHEADER)\r
c->flags |= CODEC_FLAG_GLOBAL_HEADER;\r
- \r
- return st;\r
+ \r
+ //c->thread_count = boost::thread::hardware_concurrency();\r
+ THROW_ON_ERROR2(tbb_avcodec_open(c, encoder), "[ffmpeg_consumer]");\r
+\r
+ return std::shared_ptr<AVStream>(st, [](AVStream* st)\r
+ {\r
+ LOG_ON_ERROR2(tbb_avcodec_close(st->codec), "[ffmpeg_consumer]");\r
+ av_freep(&st->codec);\r
+ av_freep(&st);\r
+ });\r
}\r
- \r
- void open_video(AVStream* st)\r
- { \r
- auto c = st->codec;\r
- \r
- auto codec = avcodec_find_encoder(c->codec_id);\r
- if (!codec)\r
+ \r
+ std::shared_ptr<AVStream> add_audio_stream(std::vector<option>& options)\r
+ {\r
+ if(output_format_.acodec == CODEC_ID_NONE)\r
+ return nullptr;\r
+\r
+ auto st = av_new_stream(oc_.get(), 1);\r
+ if(!st)\r
+ BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Could not allocate audio-stream") << boost::errinfo_api_function("av_new_stream")); \r
+ \r
+ auto encoder = avcodec_find_encoder(output_format_.acodec);\r
+ if (!encoder)\r
BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("codec not found"));\r
\r
- int errn = 0;\r
- if ((errn = -avcodec_open(c, codec)) > 0)\r
- BOOST_THROW_EXCEPTION(\r
- file_read_error() << \r
- msg_info("Could not open video codec.") <<\r
- boost::errinfo_api_function("avcodec_open") <<\r
- boost::errinfo_errno(errn) <<\r
- boost::errinfo_file_name(filename_)); \r
- }\r
- \r
- void encode_video_frame(const safe_ptr<const core::read_frame>& frame)\r
- { \r
- AVCodecContext* c = video_st_->codec;\r
- \r
- if (img_convert_ctx_ == nullptr) \r
- {\r
- img_convert_ctx_ = sws_getContext(format_desc_.width, format_desc_.height, PIX_FMT_BGRA, c->width, c->height, c->pix_fmt, SWS_BICUBIC, nullptr, nullptr, nullptr);\r
- if (img_convert_ctx_ == nullptr) \r
- BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Cannot initialize the conversion context"));\r
- }\r
+ auto c = st->codec;\r
\r
- std::shared_ptr<AVFrame> av_frame(avcodec_alloc_frame(), av_free);\r
- avpicture_fill(reinterpret_cast<AVPicture*>(av_frame.get()), const_cast<uint8_t*>(frame->image_data().begin()), PIX_FMT_BGRA, format_desc_.width, format_desc_.height);\r
- \r
- std::shared_ptr<AVFrame> local_av_frame(avcodec_alloc_frame(), av_free);\r
- picture_buf_.resize(avpicture_get_size(c->pix_fmt, format_desc_.width, format_desc_.height));\r
- avpicture_fill(reinterpret_cast<AVPicture*>(local_av_frame.get()), picture_buf_.data(), c->pix_fmt, format_desc_.width, format_desc_.height);\r
+ avcodec_get_context_defaults3(c, encoder);\r
\r
- sws_scale(img_convert_ctx_, av_frame->data, av_frame->linesize, 0, c->height, local_av_frame->data, local_av_frame->linesize);\r
- \r
- int ret = avcodec_encode_video(c, video_outbuf_.data(), video_outbuf_.size(), local_av_frame.get());\r
+ c->codec_id = output_format_.acodec;\r
+ c->codec_type = AVMEDIA_TYPE_AUDIO;\r
+ c->sample_rate = 48000;\r
+ c->channels = 2;\r
+ c->sample_fmt = AV_SAMPLE_FMT_S16;\r
+\r
+ if(output_format_.vcodec == CODEC_ID_FLV1) \r
+ c->sample_rate = 44100; \r
+\r
+ if(output_format_.format->flags & AVFMT_GLOBALHEADER)\r
+ c->flags |= CODEC_FLAG_GLOBAL_HEADER;\r
\r
- int errn = -ret;\r
- if (errn > 0) \r
- BOOST_THROW_EXCEPTION(\r
- invalid_operation() << \r
- msg_info("Could not encode video frame.") <<\r
- boost::errinfo_api_function("avcodec_encode_video") <<\r
- boost::errinfo_errno(errn) <<\r
- boost::errinfo_file_name(filename_));\r
-\r
- auto out_size = ret;\r
- AVPacket pkt;\r
- av_init_packet(&pkt);\r
- pkt.size = out_size;\r
-\r
- // If zero size, it means the image was buffered.\r
- if (out_size > 0) \r
- { \r
- if (c->coded_frame->pts != AV_NOPTS_VALUE)\r
- pkt.pts = av_rescale_q(c->coded_frame->pts, c->time_base, video_st_->time_base);\r
- if(c->coded_frame->key_frame)\r
- pkt.flags |= AV_PKT_FLAG_KEY;\r
+ boost::range::remove_erase_if(options, [&](const option& o)\r
+ {\r
+ return ffmpeg::av_opt_set(c, o.name.c_str(), o.value.c_str(), AV_OPT_SEARCH_CHILDREN) > -1;\r
+ });\r
\r
- pkt.stream_index = video_st_->index;\r
- pkt.data = video_outbuf_.data();\r
+ THROW_ON_ERROR2(avcodec_open(c, encoder), "[ffmpeg_consumer]");\r
\r
- if (av_interleaved_write_frame(oc_.get(), &pkt) != 0)\r
- BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Error while writing video frame"));\r
- } \r
+ return std::shared_ptr<AVStream>(st, [](AVStream* st)\r
+ {\r
+ LOG_ON_ERROR2(avcodec_close(st->codec), "[ffmpeg_consumer]");;\r
+ av_freep(&st->codec);\r
+ av_freep(&st);\r
+ });\r
}\r
-\r
- AVStream* add_audio_stream(enum CodecID codec_id)\r
+ \r
+ std::shared_ptr<AVFrame> convert_video(core::const_frame frame, AVCodecContext* c)\r
{\r
- audio_st_ = av_new_stream(oc_.get(), 1);\r
- if (!audio_st_)\r
- BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Could not alloc stream"));\r
+ if(!sws_) \r
+ {\r
+ sws_.reset(sws_getContext(format_desc_.width, \r
+ format_desc_.height - output_format_.croptop - output_format_.cropbot, \r
+ PIX_FMT_BGRA,\r
+ c->width,\r
+ c->height, \r
+ c->pix_fmt, \r
+ SWS_BICUBIC, nullptr, nullptr, nullptr), \r
+ sws_freeContext);\r
+ if (sws_ == nullptr) \r
+ BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Cannot initialize the conversion context"));\r
+ }\r
\r
- auto c = audio_st_->codec;\r
- c->codec_id = codec_id;\r
- c->codec_type = AVMEDIA_TYPE_AUDIO;\r
+ std::shared_ptr<AVFrame> in_frame(avcodec_alloc_frame(), av_free);\r
+\r
+ avpicture_fill(reinterpret_cast<AVPicture*>(in_frame.get()), \r
+ const_cast<uint8_t*>(frame.image_data().begin()),\r
+ PIX_FMT_BGRA, \r
+ format_desc_.width,\r
+ format_desc_.height - output_format_.croptop - output_format_.cropbot);\r
+\r
+ for(int n = 0; n < 4; ++n) \r
+ in_frame->data[n] += in_frame->linesize[n] * output_format_.croptop; \r
+ \r
+ std::shared_ptr<AVFrame> out_frame(avcodec_alloc_frame(), av_free);\r
+\r
+ picture_buf_.resize(avpicture_get_size(c->pix_fmt, \r
+ c->width,\r
+ c->height));\r
+\r
+ avpicture_fill(reinterpret_cast<AVPicture*>(out_frame.get()),\r
+ picture_buf_.data(), \r
+ c->pix_fmt, \r
+ c->width, \r
+ c->height);\r
+\r
+ sws_scale(sws_.get(), \r
+ in_frame->data, \r
+ in_frame->linesize,\r
+ 0, \r
+ format_desc_.height - output_format_.cropbot - output_format_.croptop, \r
+ out_frame->data, \r
+ out_frame->linesize);\r
+\r
+ return out_frame;\r
+ }\r
+ \r
+ std::shared_ptr<AVPacket> encode_video_frame(core::const_frame frame)\r
+ { \r
+ auto c = video_st_->codec;\r
+ \r
+ auto in_time = static_cast<double>(in_frame_number_) / format_desc_.fps;\r
+ auto out_time = static_cast<double>(out_frame_number_) / (static_cast<double>(c->time_base.den) / static_cast<double>(c->time_base.num));\r
+ \r
+ in_frame_number_++;\r
\r
- // Put sample parameters.\r
- c->bit_rate = 192000;\r
- c->sample_rate = 48000;\r
- c->channels = 2;\r
+ if(out_time - in_time > 0.01)\r
+ return nullptr;\r
+ \r
+ auto av_frame = convert_video(frame, c);\r
+ av_frame->interlaced_frame = format_desc_.field_mode != core::field_mode::progressive;\r
+ av_frame->top_field_first = format_desc_.field_mode == core::field_mode::upper;\r
+ av_frame->pts = out_frame_number_++;\r
\r
- // Some formats want stream headers to be separate.\r
- if(oc_->oformat->flags & AVFMT_GLOBALHEADER)\r
- c->flags |= CODEC_FLAG_GLOBAL_HEADER;\r
+ int out_size = THROW_ON_ERROR2(avcodec_encode_video(c, video_outbuf_.data(), static_cast<int>(video_outbuf_.size()), av_frame.get()), "[ffmpeg_consumer]");\r
+ if(out_size > 0)\r
+ {\r
+ spl::shared_ptr<AVPacket> pkt(new AVPacket, [](AVPacket* p)\r
+ {\r
+ av_free_packet(p);\r
+ delete p;\r
+ });\r
+ av_init_packet(pkt.get());\r
+ \r
+ if (c->coded_frame->pts != AV_NOPTS_VALUE)\r
+ pkt->pts = av_rescale_q(c->coded_frame->pts, c->time_base, video_st_->time_base);\r
\r
- return audio_st_;\r
- }\r
+ if(c->coded_frame->key_frame)\r
+ pkt->flags |= AV_PKT_FLAG_KEY;\r
\r
- void open_audio(AVStream* st)\r
+ pkt->stream_index = video_st_->index;\r
+ pkt->data = video_outbuf_.data();\r
+ pkt->size = out_size;\r
+ \r
+ av_dup_packet(pkt.get());\r
+ return pkt;\r
+ } \r
+ return nullptr;\r
+ }\r
+ \r
+ byte_vector convert_audio(core::const_frame& frame, AVCodecContext* c)\r
{\r
- auto c = st->codec;\r
+ if(!swr_) \r
+ swr_.reset(new audio_resampler(c->channels, format_desc_.audio_channels, \r
+ c->sample_rate, format_desc_.audio_sample_rate,\r
+ c->sample_fmt, AV_SAMPLE_FMT_S32));\r
+ \r
\r
- // Find the audio encoder.\r
- auto codec = avcodec_find_encoder(c->codec_id);\r
- if (!codec) \r
- BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("codec not found"));\r
+ auto audio_data = frame.audio_data();\r
+\r
+ std::vector<int8_t, tbb::cache_aligned_allocator<int8_t>> audio_resample_buffer;\r
+ std::copy(reinterpret_cast<const uint8_t*>(audio_data.data()), \r
+ reinterpret_cast<const uint8_t*>(audio_data.data()) + audio_data.size()*4, \r
+ std::back_inserter(audio_resample_buffer));\r
+ \r
+ audio_resample_buffer = swr_->resample(std::move(audio_resample_buffer));\r
\r
- // Open it.\r
- int errn = 0;\r
- if ((errn = -avcodec_open(c, codec)) > 0)\r
- BOOST_THROW_EXCEPTION(\r
- file_read_error() << \r
- msg_info("Could not open audio codec") <<\r
- boost::errinfo_api_function("avcodec_open") <<\r
- boost::errinfo_errno(errn) <<\r
- boost::errinfo_file_name(filename_));\r
+ return byte_vector(audio_resample_buffer.begin(), audio_resample_buffer.end());\r
}\r
\r
- void encode_audio_frame(const safe_ptr<const core::read_frame>& frame)\r
- { \r
- if(!frame->audio_data().empty())\r
- audio_input_buffer_.insert(audio_input_buffer_.end(), frame->audio_data().begin(), frame->audio_data().end());\r
- else\r
- audio_input_buffer_.insert(audio_input_buffer_.end(), 3840, 0);\r
-\r
- while(encode_audio_packet()){}\r
- }\r
-\r
- bool encode_audio_packet()\r
- { \r
+ std::shared_ptr<AVPacket> encode_audio_frame(core::const_frame frame)\r
+ { \r
auto c = audio_st_->codec;\r
\r
- auto frame_bytes = c->frame_size * 2 * 2; // samples per frame * 2 channels * 2 bytes per sample\r
- if(static_cast<int>(audio_input_buffer_.size()) < frame_bytes/2)\r
- return false;\r
-\r
- AVPacket pkt;\r
- av_init_packet(&pkt);\r
+ boost::range::push_back(audio_buf_, convert_audio(frame, c));\r
+ \r
+ std::size_t frame_size = c->frame_size;\r
\r
- int ret = avcodec_encode_audio(c, audio_outbuf_.data(), audio_outbuf_.size(), audio_input_buffer_.data());\r
+ spl::shared_ptr<AVPacket> pkt(new AVPacket, [](AVPacket* p)\r
+ {\r
+ av_free_packet(p);\r
+ delete p;\r
+ });\r
+ av_init_packet(pkt.get());\r
+\r
+ if(frame_size > 1)\r
+ { \r
+ auto input_audio_size = frame_size * av_get_bytes_per_sample(c->sample_fmt) * c->channels;\r
+ \r
+ if(audio_buf_.size() < input_audio_size)\r
+ return nullptr;\r
+ \r
+ pkt->size = avcodec_encode_audio(c, audio_outbuf_.data(), static_cast<int>(audio_outbuf_.size()), reinterpret_cast<short*>(audio_buf_.data()));\r
+ audio_buf_.erase(audio_buf_.begin(), audio_buf_.begin() + input_audio_size);\r
+ }\r
+ else\r
+ {\r
+ audio_outbuf_ = std::move(audio_buf_); \r
+ audio_buf_.clear();\r
+ pkt->size = static_cast<int>(audio_outbuf_.size());\r
+ pkt->data = audio_outbuf_.data();\r
+ }\r
\r
- int errn = -ret;\r
- if (errn > 0) \r
- BOOST_THROW_EXCEPTION(\r
- invalid_operation() << \r
- msg_info("Could not encode audio samples.") <<\r
- boost::errinfo_api_function("avcodec_encode_audio") <<\r
- boost::errinfo_errno(errn) <<\r
- boost::errinfo_file_name(filename_));\r
-\r
- pkt.size = ret;\r
- audio_input_buffer_ = std::vector<short, tbb::cache_aligned_allocator<short>>(audio_input_buffer_.begin() + frame_bytes/2, audio_input_buffer_.end());\r
+ if(pkt->size == 0)\r
+ return nullptr;\r
\r
if (c->coded_frame && c->coded_frame->pts != AV_NOPTS_VALUE)\r
- pkt.pts = av_rescale_q(c->coded_frame->pts, c->time_base, audio_st_->time_base);\r
- pkt.flags |= AV_PKT_FLAG_KEY;\r
- pkt.stream_index = audio_st_->index;\r
- pkt.data = audio_outbuf_.data();\r
- \r
- if (av_interleaved_write_frame(oc_.get(), &pkt) != 0)\r
- BOOST_THROW_EXCEPTION(caspar_exception() << msg_info("Error while writing audio frame"));\r
+ pkt->pts = av_rescale_q(c->coded_frame->pts, c->time_base, audio_st_->time_base);\r
\r
- return true;\r
+ pkt->flags |= AV_PKT_FLAG_KEY;\r
+ pkt->stream_index = audio_st_->index;\r
+ pkt->data = reinterpret_cast<uint8_t*>(audio_outbuf_.data());\r
+ \r
+ av_dup_packet(pkt.get());\r
+ return pkt;\r
}\r
- \r
- void send(const safe_ptr<const core::read_frame>& frame)\r
+ \r
+ void send(core::const_frame& frame)\r
{\r
- active_.get();\r
- active_ = executor_.begin_invoke([=]\r
- { \r
- auto my_frame = frame;\r
- encode_video_frame(my_frame);\r
- encode_audio_frame(my_frame);\r
+ encode_executor_.begin_invoke([=]\r
+ { \r
+ boost::timer frame_timer;\r
+\r
+ auto video = encode_video_frame(frame);\r
+ auto audio = encode_audio_frame(frame);\r
+\r
+ graph_->set_value("frame-time", frame_timer.elapsed()*format_desc_.fps*0.5);\r
+ \r
+ write_executor_.begin_invoke([=]\r
+ {\r
+ if(video)\r
+ av_interleaved_write_frame(oc_.get(), video.get());\r
+ if(audio)\r
+ av_interleaved_write_frame(oc_.get(), audio.get());\r
+ });\r
});\r
}\r
-\r
- size_t buffer_depth() const { return 1; }\r
};\r
\r
-ffmpeg_consumer::ffmpeg_consumer(const std::wstring& filename) : impl_(new implementation(narrow(filename))){}\r
-ffmpeg_consumer::ffmpeg_consumer(ffmpeg_consumer&& other) : impl_(std::move(other.impl_)){}\r
-void ffmpeg_consumer::send(const safe_ptr<const core::read_frame>& frame){impl_->send(frame);}\r
-size_t ffmpeg_consumer::buffer_depth() const{return impl_->buffer_depth();}\r
-void ffmpeg_consumer::initialize(const core::video_format_desc& format_desc)\r
+struct ffmpeg_consumer_proxy : public core::frame_consumer\r
{\r
- // TODO: Ugly\r
- impl_.reset(new implementation(impl_->filename_));\r
- impl_->initialize(format_desc);\r
-}\r
-std::wstring ffmpeg_consumer::print() const {return impl_->print();}\r
+ const std::wstring filename_;\r
+ const std::vector<option> options_;\r
+\r
+ std::unique_ptr<ffmpeg_consumer> consumer_;\r
\r
-safe_ptr<core::frame_consumer> create_ffmpeg_consumer(const std::vector<std::wstring>& params)\r
+public:\r
+\r
+ ffmpeg_consumer_proxy(const std::wstring& filename, const std::vector<option>& options)\r
+ : filename_(filename)\r
+ , options_(options)\r
+ {\r
+ }\r
+ \r
+ virtual void initialize(const core::video_format_desc& format_desc, int)\r
+ {\r
+ consumer_.reset();\r
+ consumer_.reset(new ffmpeg_consumer(u8(filename_), format_desc, options_));\r
+ }\r
+ \r
+ bool send(core::const_frame frame) override\r
+ {\r
+ consumer_->send(frame);\r
+ return true;\r
+ }\r
+ \r
+ std::wstring print() const override\r
+ {\r
+ return consumer_ ? consumer_->print() : L"[ffmpeg_consumer]";\r
+ }\r
+\r
+ std::wstring name() const override\r
+ {\r
+ return L"file";\r
+ }\r
+\r
+ boost::property_tree::wptree info() const override\r
+ {\r
+ boost::property_tree::wptree info;\r
+ info.add(L"type", L"file");\r
+ info.add(L"filename", filename_);\r
+ return info;\r
+ }\r
+ \r
+ bool has_synchronization_clock() const override\r
+ {\r
+ return false;\r
+ }\r
+\r
+ int buffer_depth() const override\r
+ {\r
+ return 1;\r
+ }\r
+\r
+ int index() const override\r
+ {\r
+ return 200;\r
+ }\r
+}; \r
+spl::shared_ptr<core::frame_consumer> create_consumer(const std::vector<std::wstring>& params)\r
{\r
- if(params.size() < 2 || params[0] != L"FILE")\r
+ if(params.size() < 1 || params[0] != L"FILE")\r
return core::frame_consumer::empty();\r
\r
- // TODO: Ask stakeholders about case where file already exists.\r
- boost::filesystem::remove(boost::filesystem::wpath(env::media_folder() + params[1])); // Delete the file if it exists\r
- return make_safe<ffmpeg_consumer>(env::media_folder() + params[1]);\r
+ auto filename = (params.size() > 1 ? params[1] : L"");\r
+ \r
+ std::vector<option> options;\r
+ \r
+ if(params.size() >= 3)\r
+ {\r
+ for(auto opt_it = params.begin()+2; opt_it != params.end();)\r
+ {\r
+ auto name = u8(boost::trim_copy(boost::to_lower_copy(*opt_it++))).substr(1);\r
+ auto value = u8(boost::trim_copy(boost::to_lower_copy(*opt_it++)));\r
+ \r
+ if(value == "h264")\r
+ value = "libx264";\r
+ else if(value == "dvcpro")\r
+ value = "dvvideo";\r
+\r
+ options.push_back(option(name, value));\r
+ }\r
+ }\r
+ \r
+ return spl::make_shared<ffmpeg_consumer_proxy>(env::media_folder() + filename, options);\r
}\r
\r
+spl::shared_ptr<core::frame_consumer> create_consumer(const boost::property_tree::wptree& ptree)\r
+{\r
+ auto filename = ptree.get<std::wstring>(L"path");\r
+ auto codec = ptree.get(L"vcodec", L"libx264");\r
+\r
+ std::vector<option> options;\r
+ options.push_back(option("vcodec", u8(codec)));\r
+ \r
+ return spl::make_shared<ffmpeg_consumer_proxy>(env::media_folder() + filename, options);\r
}\r
+\r
+}}\r