/** * @file * @brief Source file for FFmpegWriter class * @author Jonathan Thomas , Fabrice Bellard * * @ref License */ /* LICENSE * * Copyright (c) 2008-2019 OpenShot Studios, LLC, Fabrice Bellard * (http://www.openshotstudios.com). This file is part of * OpenShot Library (http://www.openshot.org), an open-source project * dedicated to delivering high quality video editing and animation solutions * to the world. * * This file is originally based on the Libavformat API example, and then modified * by the libopenshot project. * * OpenShot Library (libopenshot) is free software: you can redistribute it * and/or modify it under the terms of the GNU Lesser General Public License * as published by the Free Software Foundation, either version 3 of the * License, or (at your option) any later version. * * OpenShot Library (libopenshot) is distributed in the hope that it will be * useful, but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public License * along with OpenShot Library. If not, see . */ #include "../include/FFmpegWriter.h" using namespace openshot; #if HAVE_HW_ACCEL #pragma message "You are compiling with experimental hardware encode" #else #pragma message "You are compiling only with software encode" #endif // Multiplexer parameters temporary storage AVDictionary *mux_dict = NULL; #if HAVE_HW_ACCEL int hw_en_on = 1; // Is set in UI int hw_en_supported = 0; // Is set by FFmpegWriter AVPixelFormat hw_en_av_pix_fmt = AV_PIX_FMT_NONE; AVHWDeviceType hw_en_av_device_type = AV_HWDEVICE_TYPE_VAAPI; static AVBufferRef *hw_device_ctx = NULL; AVFrame *hw_frame = NULL; static int set_hwframe_ctx(AVCodecContext *ctx, AVBufferRef *hw_device_ctx, int64_t width, int64_t height) { AVBufferRef *hw_frames_ref; AVHWFramesContext *frames_ctx = NULL; int err = 0; if (!(hw_frames_ref = av_hwframe_ctx_alloc(hw_device_ctx))) { fprintf(stderr, "Failed to create HW frame context.\n"); return -1; } frames_ctx = (AVHWFramesContext *)(hw_frames_ref->data); frames_ctx->format = hw_en_av_pix_fmt; frames_ctx->sw_format = AV_PIX_FMT_NV12; frames_ctx->width = width; frames_ctx->height = height; frames_ctx->initial_pool_size = 20; if ((err = av_hwframe_ctx_init(hw_frames_ref)) < 0) { fprintf(stderr, "Failed to initialize HW frame context." "Error code: %s\n",av_err2str(err)); av_buffer_unref(&hw_frames_ref); return err; } ctx->hw_frames_ctx = av_buffer_ref(hw_frames_ref); if (!ctx->hw_frames_ctx) err = AVERROR(ENOMEM); av_buffer_unref(&hw_frames_ref); return err; } #endif // HAVE_HW_ACCEL FFmpegWriter::FFmpegWriter(std::string path) : path(path), fmt(NULL), oc(NULL), audio_st(NULL), video_st(NULL), samples(NULL), audio_outbuf(NULL), audio_outbuf_size(0), audio_input_frame_size(0), audio_input_position(0), initial_audio_input_frame_size(0), img_convert_ctx(NULL), cache_size(8), num_of_rescalers(32), rescaler_position(0), video_codec(NULL), audio_codec(NULL), is_writing(false), write_video_count(0), write_audio_count(0), original_sample_rate(0), original_channels(0), avr(NULL), avr_planar(NULL), is_open(false), prepare_streams(false), write_header(false), write_trailer(false), audio_encoder_buffer_size(0), audio_encoder_buffer(NULL) { // Disable audio & video (so they can be independently enabled) info.has_audio = false; info.has_video = false; // Initialize FFMpeg, and register all formats and codecs AV_REGISTER_ALL // auto detect format auto_detect_format(); } // Open the writer void FFmpegWriter::Open() { if (!is_open) { // Open the writer is_open = true; // Prepare streams (if needed) if (!prepare_streams) PrepareStreams(); // Now that all the parameters are set, we can open the audio and video codecs and allocate the necessary encode buffers if (info.has_video && video_st) open_video(oc, video_st); if (info.has_audio && audio_st) open_audio(oc, audio_st); // Write header (if needed) if (!write_header) WriteHeader(); } } // auto detect format (from path) void FFmpegWriter::auto_detect_format() { // Auto detect the output format from the name. default is mpeg. fmt = av_guess_format(NULL, path.c_str(), NULL); if (!fmt) throw InvalidFormat("Could not deduce output format from file extension.", path); // Allocate the output media context AV_OUTPUT_CONTEXT(&oc, path.c_str()); if (!oc) throw OutOfMemory("Could not allocate memory for AVFormatContext.", path); // Set the AVOutputFormat for the current AVFormatContext oc->oformat = fmt; // Update codec names if (fmt->video_codec != AV_CODEC_ID_NONE && info.has_video) // Update video codec name info.vcodec = avcodec_find_encoder(fmt->video_codec)->name; if (fmt->audio_codec != AV_CODEC_ID_NONE && info.has_audio) // Update audio codec name info.acodec = avcodec_find_encoder(fmt->audio_codec)->name; } // initialize streams void FFmpegWriter::initialize_streams() { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::initialize_streams", "fmt->video_codec", fmt->video_codec, "fmt->audio_codec", fmt->audio_codec, "AV_CODEC_ID_NONE", AV_CODEC_ID_NONE); // Add the audio and video streams using the default format codecs and initialize the codecs video_st = NULL; audio_st = NULL; if (fmt->video_codec != AV_CODEC_ID_NONE && info.has_video) // Add video stream video_st = add_video_stream(); if (fmt->audio_codec != AV_CODEC_ID_NONE && info.has_audio) // Add audio stream audio_st = add_audio_stream(); } // Set video export options void FFmpegWriter::SetVideoOptions(bool has_video, std::string codec, Fraction fps, int width, int height, Fraction pixel_ratio, bool interlaced, bool top_field_first, int bit_rate) { // Set the video options if (codec.length() > 0) { AVCodec *new_codec; // Check if the codec selected is a hardware accelerated codec #if HAVE_HW_ACCEL #if defined(__linux__) if (strstr(codec.c_str(), "_vaapi") != NULL) { new_codec = avcodec_find_encoder_by_name(codec.c_str()); hw_en_on = 1; hw_en_supported = 1; hw_en_av_pix_fmt = AV_PIX_FMT_VAAPI; hw_en_av_device_type = AV_HWDEVICE_TYPE_VAAPI; } else if (strstr(codec.c_str(), "_nvenc") != NULL) { new_codec = avcodec_find_encoder_by_name(codec.c_str()); hw_en_on = 1; hw_en_supported = 1; hw_en_av_pix_fmt = AV_PIX_FMT_CUDA; hw_en_av_device_type = AV_HWDEVICE_TYPE_CUDA; } else { new_codec = avcodec_find_encoder_by_name(codec.c_str()); hw_en_on = 0; hw_en_supported = 0; } #elif defined(_WIN32) if (strstr(codec.c_str(), "_dxva2") != NULL) { new_codec = avcodec_find_encoder_by_name(codec.c_str()); hw_en_on = 1; hw_en_supported = 1; hw_en_av_pix_fmt = AV_PIX_FMT_DXVA2_VLD; hw_en_av_device_type = AV_HWDEVICE_TYPE_DXVA2; } else if (strstr(codec.c_str(), "_nvenc") != NULL) { new_codec = avcodec_find_encoder_by_name(codec.c_str()); hw_en_on = 1; hw_en_supported = 1; hw_en_av_pix_fmt = AV_PIX_FMT_CUDA; hw_en_av_device_type = AV_HWDEVICE_TYPE_CUDA; } else { new_codec = avcodec_find_encoder_by_name(codec.c_str()); hw_en_on = 0; hw_en_supported = 0; } #elif defined(__APPLE__) if (strstr(codec.c_str(), "_videotoolbox") != NULL) { new_codec = avcodec_find_encoder_by_name(codec.c_str()); hw_en_on = 1; hw_en_supported = 1; hw_en_av_pix_fmt = AV_PIX_FMT_VIDEOTOOLBOX; hw_en_av_device_type = AV_HWDEVICE_TYPE_VIDEOTOOLBOX; } else { new_codec = avcodec_find_encoder_by_name(codec.c_str()); hw_en_on = 0; hw_en_supported = 0; } #else // is FFmpeg 3 but not linux new_codec = avcodec_find_encoder_by_name(codec.c_str()); #endif //__linux__ #else // not ffmpeg 3 new_codec = avcodec_find_encoder_by_name(codec.c_str()); #endif // HAVE_HW_ACCEL if (new_codec == NULL) throw InvalidCodec("A valid video codec could not be found for this file.", path); else { // Set video codec info.vcodec = new_codec->name; // Update video codec in fmt fmt->video_codec = new_codec->id; } } if (fps.num > 0) { // Set frames per second (if provided) info.fps.num = fps.num; info.fps.den = fps.den; // Set the timebase (inverse of fps) info.video_timebase.num = info.fps.den; info.video_timebase.den = info.fps.num; } if (width >= 1) info.width = width; if (height >= 1) info.height = height; if (pixel_ratio.num > 0) { info.pixel_ratio.num = pixel_ratio.num; info.pixel_ratio.den = pixel_ratio.den; } if (bit_rate >= 1000) // bit_rate is the bitrate in b/s info.video_bit_rate = bit_rate; if ((bit_rate >= 0) && (bit_rate < 256)) // bit_rate is the bitrate in crf info.video_bit_rate = bit_rate; info.interlaced_frame = interlaced; info.top_field_first = top_field_first; // Calculate the DAR (display aspect ratio) Fraction size(info.width * info.pixel_ratio.num, info.height * info.pixel_ratio.den); // Reduce size fraction size.Reduce(); // Set the ratio based on the reduced fraction info.display_ratio.num = size.num; info.display_ratio.den = size.den; ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::SetVideoOptions (" + codec + ")", "width", width, "height", height, "size.num", size.num, "size.den", size.den, "fps.num", fps.num, "fps.den", fps.den); // Enable / Disable video info.has_video = has_video; } // Set video export options (overloaded function) void FFmpegWriter::SetVideoOptions(std::string codec, int width, int height, Fraction fps, int bit_rate) { // Call full signature with some default parameters FFmpegWriter::SetVideoOptions(true, codec, fps, width, height, openshot::Fraction(1, 1), false, true, bit_rate); } // Set audio export options void FFmpegWriter::SetAudioOptions(bool has_audio, std::string codec, int sample_rate, int channels, ChannelLayout channel_layout, int bit_rate) { // Set audio options if (codec.length() > 0) { AVCodec *new_codec = avcodec_find_encoder_by_name(codec.c_str()); if (new_codec == NULL) throw InvalidCodec("A valid audio codec could not be found for this file.", path); else { // Set audio codec info.acodec = new_codec->name; // Update audio codec in fmt fmt->audio_codec = new_codec->id; } } if (sample_rate > 7999) info.sample_rate = sample_rate; if (channels > 0) info.channels = channels; if (bit_rate > 999) info.audio_bit_rate = bit_rate; info.channel_layout = channel_layout; // init resample options (if zero) if (original_sample_rate == 0) original_sample_rate = info.sample_rate; if (original_channels == 0) original_channels = info.channels; ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::SetAudioOptions (" + codec + ")", "sample_rate", sample_rate, "channels", channels, "bit_rate", bit_rate); // Enable / Disable audio info.has_audio = has_audio; } // Set audio export options (overloaded function) void FFmpegWriter::SetAudioOptions(std::string codec, int sample_rate, int bit_rate) { // Call full signature with some default parameters FFmpegWriter::SetAudioOptions(true, codec, sample_rate, 2, openshot::LAYOUT_STEREO, bit_rate); } // Set custom options (some codecs accept additional params) void FFmpegWriter::SetOption(StreamType stream, std::string name, std::string value) { // Declare codec context AVCodecContext *c = NULL; AVStream *st = NULL; std::stringstream convert(value); if (info.has_video && stream == VIDEO_STREAM && video_st) { st = video_st; // Get codec context c = AV_GET_CODEC_PAR_CONTEXT(st, video_codec); } else if (info.has_audio && stream == AUDIO_STREAM && audio_st) { st = audio_st; // Get codec context c = AV_GET_CODEC_PAR_CONTEXT(st, audio_codec); } else throw NoStreamsFound("The stream was not found. Be sure to call PrepareStreams() first.", path); // Init AVOption const AVOption *option = NULL; // Was a codec / stream found? if (c) // Find AVOption (if it exists) option = AV_OPTION_FIND(c->priv_data, name.c_str()); // Was option found? if (option || (name == "g" || name == "qmin" || name == "qmax" || name == "max_b_frames" || name == "mb_decision" || name == "level" || name == "profile" || name == "slices" || name == "rc_min_rate" || name == "rc_max_rate" || name == "rc_buffer_size" || name == "crf" || name == "cqp" || name == "qp")) { // Check for specific named options if (name == "g") // Set gop_size convert >> c->gop_size; else if (name == "qmin") // Minimum quantizer convert >> c->qmin; else if (name == "qmax") // Maximum quantizer convert >> c->qmax; else if (name == "max_b_frames") // Maximum number of B-frames between non-B-frames convert >> c->max_b_frames; else if (name == "mb_decision") // Macroblock decision mode convert >> c->mb_decision; else if (name == "level") // Set codec level convert >> c->level; else if (name == "profile") // Set codec profile convert >> c->profile; else if (name == "slices") // Indicates number of picture subdivisions convert >> c->slices; else if (name == "rc_min_rate") // Minimum bitrate convert >> c->rc_min_rate; else if (name == "rc_max_rate") // Maximum bitrate convert >> c->rc_max_rate; else if (name == "rc_buffer_size") // Buffer size convert >> c->rc_buffer_size; else if (name == "cqp") { // encode quality and special settings like lossless // This might be better in an extra methods as more options // and way to set quality are possible #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(55, 39, 101) #if HAVE_HW_ACCEL if (hw_en_on) { av_opt_set_int(c->priv_data, "qp", std::min(std::stoi(value),63), 0); // 0-63 } else #endif // HAVE_HW_ACCEL { switch (c->codec_id) { #if (LIBAVCODEC_VERSION_MAJOR >= 58) case AV_CODEC_ID_AV1 : c->bit_rate = 0; av_opt_set_int(c->priv_data, "qp", std::min(std::stoi(value),63), 0); // 0-63 break; #endif case AV_CODEC_ID_VP8 : c->bit_rate = 10000000; av_opt_set_int(c->priv_data, "qp", std::max(std::min(std::stoi(value), 63), 4), 0); // 4-63 break; case AV_CODEC_ID_VP9 : c->bit_rate = 0; // Must be zero! av_opt_set_int(c->priv_data, "qp", std::min(std::stoi(value), 63), 0); // 0-63 if (std::stoi(value) == 0) { av_opt_set(c->priv_data, "preset", "veryslow", 0); av_opt_set_int(c->priv_data, "lossless", 1, 0); } break; case AV_CODEC_ID_H264 : av_opt_set_int(c->priv_data, "qp", std::min(std::stoi(value), 51), 0); // 0-51 if (std::stoi(value) == 0) { av_opt_set(c->priv_data, "preset", "veryslow", 0); } break; case AV_CODEC_ID_HEVC : av_opt_set_int(c->priv_data, "qp", std::min(std::stoi(value), 51), 0); // 0-51 if (std::stoi(value) == 0) { av_opt_set(c->priv_data, "preset", "veryslow", 0); av_opt_set_int(c->priv_data, "lossless", 1, 0); } break; default: // For all other codecs assume a range of 0-63 av_opt_set_int(c->priv_data, "qp", std::min(std::stoi(value), 63), 0); // 0-63 c->bit_rate = 0; } } #endif } else if (name == "crf") { // encode quality and special settings like lossless // This might be better in an extra methods as more options // and way to set quality are possible #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(55, 39, 101) #if HAVE_HW_ACCEL if (hw_en_on) { double mbs = 15000000.0; if (info.video_bit_rate > 0) { if (info.video_bit_rate > 42) { mbs = 380000.0; } else { mbs *= std::pow(0.912,info.video_bit_rate); } } c->bit_rate = (int)(mbs); } else #endif // HAVE_HW_ACCEL { switch (c->codec_id) { #if (LIBAVCODEC_VERSION_MAJOR >= 58) case AV_CODEC_ID_AV1 : c->bit_rate = 0; // AV1 only supports "crf" quality values av_opt_set_int(c->priv_data, "crf", std::min(std::stoi(value),63), 0); break; #endif case AV_CODEC_ID_VP8 : c->bit_rate = 10000000; av_opt_set_int(c->priv_data, "crf", std::max(std::min(std::stoi(value), 63), 4), 0); // 4-63 break; case AV_CODEC_ID_VP9 : c->bit_rate = 0; // Must be zero! av_opt_set_int(c->priv_data, "crf", std::min(std::stoi(value), 63), 0); // 0-63 if (std::stoi(value) == 0) { av_opt_set(c->priv_data, "preset", "veryslow", 0); av_opt_set_int(c->priv_data, "lossless", 1, 0); } break; case AV_CODEC_ID_H264 : av_opt_set_int(c->priv_data, "crf", std::min(std::stoi(value), 51), 0); // 0-51 if (std::stoi(value) == 0) { av_opt_set(c->priv_data, "preset", "veryslow", 0); } break; case AV_CODEC_ID_HEVC : if (strstr(info.vcodec.c_str(), "svt_hevc") != NULL) { av_opt_set_int(c->priv_data, "preset", 7, 0); av_opt_set_int(c->priv_data, "forced-idr",1,0); av_opt_set_int(c->priv_data, "qp",std::min(std::stoi(value), 51),0); } else { av_opt_set_int(c->priv_data, "crf", std::min(std::stoi(value), 51), 0); // 0-51 } if (std::stoi(value) == 0) { av_opt_set(c->priv_data, "preset", "veryslow", 0); av_opt_set_int(c->priv_data, "lossless", 1, 0); } break; default: // If this codec doesn't support crf calculate a bitrate // TODO: find better formula double mbs = 15000000.0; if (info.video_bit_rate > 0) { if (info.video_bit_rate > 42) { mbs = 380000.0; } else { mbs *= std::pow(0.912, info.video_bit_rate); } } c->bit_rate = (int) (mbs); } } #endif } else if (name == "qp") { // encode quality and special settings like lossless // This might be better in an extra methods as more options // and way to set quality are possible #if (LIBAVCODEC_VERSION_MAJOR >= 58) switch (c->codec_id) { case AV_CODEC_ID_AV1 : c->bit_rate = 0; if (strstr(info.vcodec.c_str(), "svt_av1") != NULL) { av_opt_set_int(c->priv_data, "qp", std::min(std::stoi(value),63), 0); } else if (strstr(info.vcodec.c_str(), "rav1e") != NULL) { // Set number of tiles to a fixed value // TODO Let user choose number of tiles av_opt_set_int(c->priv_data, "qp", std::min(std::stoi(value),255), 0); } else if (strstr(info.vcodec.c_str(), "aom") != NULL) { // Set number of tiles to a fixed value // TODO Let user choose number of tiles // libaom doesn't have qp only crf av_opt_set_int(c->priv_data, "crf", std::min(std::stoi(value),63), 0); } else { av_opt_set_int(c->priv_data, "crf", std::min(std::stoi(value),63), 0); } case AV_CODEC_ID_HEVC : c->bit_rate = 0; if (strstr(info.vcodec.c_str(), "svt_hevc") != NULL) { av_opt_set_int(c->priv_data, "qp", std::min(std::stoi(value),51), 0); av_opt_set_int(c->priv_data, "preset", 7, 0); av_opt_set_int(c->priv_data, "forced-idr",1,0); } break; } #endif } else { // Set AVOption AV_OPTION_SET(st, c->priv_data, name.c_str(), value.c_str(), c); } ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::SetOption (" + (std::string)name + ")", "stream == VIDEO_STREAM", stream == VIDEO_STREAM); // Muxing dictionary is not part of the codec context. // Just reusing SetOption function to set popular multiplexing presets. } else if (name == "muxing_preset") { if (value == "mp4_faststart") { // 'moov' box to the beginning; only for MOV, MP4 av_dict_set(&mux_dict, "movflags", "faststart", 0); } else if (value == "mp4_fragmented") { // write selfcontained fragmented file, minimum length of the fragment 8 sec; only for MOV, MP4 av_dict_set(&mux_dict, "movflags", "frag_keyframe", 0); av_dict_set(&mux_dict, "min_frag_duration", "8000000", 0); } } else { throw InvalidOptions("The option is not valid for this codec.", path); } } /// Determine if codec name is valid bool FFmpegWriter::IsValidCodec(std::string codec_name) { // Initialize FFMpeg, and register all formats and codecs AV_REGISTER_ALL // Find the codec (if any) if (avcodec_find_encoder_by_name(codec_name.c_str()) == NULL) return false; else return true; } // Prepare & initialize streams and open codecs void FFmpegWriter::PrepareStreams() { if (!info.has_audio && !info.has_video) throw InvalidOptions("No video or audio options have been set. You must set has_video or has_audio (or both).", path); ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::PrepareStreams [" + path + "]", "info.has_audio", info.has_audio, "info.has_video", info.has_video); // Initialize the streams (i.e. add the streams) initialize_streams(); // Mark as 'prepared' prepare_streams = true; } // Write the file header (after the options are set) void FFmpegWriter::WriteHeader() { if (!info.has_audio && !info.has_video) throw InvalidOptions("No video or audio options have been set. You must set has_video or has_audio (or both).", path); // Open the output file, if needed if (!(fmt->flags & AVFMT_NOFILE)) { if (avio_open(&oc->pb, path.c_str(), AVIO_FLAG_WRITE) < 0) throw InvalidFile("Could not open or write file.", path); } // Force the output filename (which doesn't always happen for some reason) AV_SET_FILENAME(oc, path.c_str()); // Add general metadata (if any) for (std::map::iterator iter = info.metadata.begin(); iter != info.metadata.end(); ++iter) { av_dict_set(&oc->metadata, iter->first.c_str(), iter->second.c_str(), 0); } // Set multiplexing parameters AVDictionary *dict = NULL; bool is_mp4 = strcmp(oc->oformat->name, "mp4"); bool is_mov = strcmp(oc->oformat->name, "mov"); // Set dictionary preset only for MP4 and MOV files if (is_mp4 || is_mov) av_dict_copy(&dict, mux_dict, 0); // Write the stream header if (avformat_write_header(oc, &dict) != 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::WriteHeader (avformat_write_header)"); throw InvalidFile("Could not write header to file.", path); }; // Free multiplexing dictionaries sets if (dict) av_dict_free(&dict); if (mux_dict) av_dict_free(&mux_dict); // Mark as 'written' write_header = true; ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::WriteHeader"); } // Add a frame to the queue waiting to be encoded. void FFmpegWriter::WriteFrame(std::shared_ptr frame) { // Check for open reader (or throw exception) if (!is_open) throw WriterClosed("The FFmpegWriter is closed. Call Open() before calling this method.", path); // Add frame pointer to "queue", waiting to be processed the next // time the WriteFrames() method is called. if (info.has_video && video_st) spooled_video_frames.push_back(frame); if (info.has_audio && audio_st) spooled_audio_frames.push_back(frame); ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::WriteFrame", "frame->number", frame->number, "spooled_video_frames.size()", spooled_video_frames.size(), "spooled_audio_frames.size()", spooled_audio_frames.size(), "cache_size", cache_size, "is_writing", is_writing); // Write the frames once it reaches the correct cache size if ((int)spooled_video_frames.size() == cache_size || (int)spooled_audio_frames.size() == cache_size) { // Is writer currently writing? if (!is_writing) // Write frames to video file write_queued_frames(); else { // Write frames to video file write_queued_frames(); } } // Keep track of the last frame added last_frame = frame; } // Write all frames in the queue to the video file. void FFmpegWriter::write_queued_frames() { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_queued_frames", "spooled_video_frames.size()", spooled_video_frames.size(), "spooled_audio_frames.size()", spooled_audio_frames.size()); // Flip writing flag is_writing = true; // Transfer spool to queue queued_video_frames = spooled_video_frames; queued_audio_frames = spooled_audio_frames; // Empty spool spooled_video_frames.clear(); spooled_audio_frames.clear(); // Set the number of threads in OpenMP omp_set_num_threads(OPEN_MP_NUM_PROCESSORS); // Allow nested OpenMP sections omp_set_nested(true); // Create blank exception bool has_error_encoding_video = false; #pragma omp parallel { #pragma omp single { // Process all audio frames (in a separate thread) if (info.has_audio && audio_st && !queued_audio_frames.empty()) write_audio_packets(false); // Loop through each queued image frame while (!queued_video_frames.empty()) { // Get front frame (from the queue) std::shared_ptr frame = queued_video_frames.front(); // Add to processed queue processed_frames.push_back(frame); // Encode and add the frame to the output file if (info.has_video && video_st) process_video_packet(frame); // Remove front item queued_video_frames.pop_front(); } // end while } // end omp single #pragma omp single { // Loop back through the frames (in order), and write them to the video file while (!processed_frames.empty()) { // Get front frame (from the queue) std::shared_ptr frame = processed_frames.front(); if (info.has_video && video_st) { // Add to deallocate queue (so we can remove the AVFrames when we are done) deallocate_frames.push_back(frame); // Does this frame's AVFrame still exist if (av_frames.count(frame)) { // Get AVFrame AVFrame *frame_final = av_frames[frame]; // Write frame to video file bool success = write_video_packet(frame, frame_final); if (!success) has_error_encoding_video = true; } } // Remove front item processed_frames.pop_front(); } // Loop through, and deallocate AVFrames while (!deallocate_frames.empty()) { // Get front frame (from the queue) std::shared_ptr frame = deallocate_frames.front(); // Does this frame's AVFrame still exist if (av_frames.count(frame)) { // Get AVFrame AVFrame *av_frame = av_frames[frame]; // Deallocate AVPicture and AVFrame av_freep(&(av_frame->data[0])); AV_FREE_FRAME(&av_frame); av_frames.erase(frame); } // Remove front item deallocate_frames.pop_front(); } // Done writing is_writing = false; } // end omp single } // end omp parallel // Raise exception from main thread if (has_error_encoding_video) throw ErrorEncodingVideo("Error while writing raw video frame", -1); } // Write a block of frames from a reader void FFmpegWriter::WriteFrame(ReaderBase *reader, int64_t start, int64_t length) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::WriteFrame (from Reader)", "start", start, "length", length); // Loop through each frame (and encoded it) for (int64_t number = start; number <= length; number++) { // Get the frame std::shared_ptr f = reader->GetFrame(number); // Encode frame WriteFrame(f); } } // Write the file trailer (after all frames are written) void FFmpegWriter::WriteTrailer() { // Write any remaining queued frames to video file write_queued_frames(); // Process final audio frame (if any) if (info.has_audio && audio_st) write_audio_packets(true); // Flush encoders (who sometimes hold on to frames) flush_encoders(); /* write the trailer, if any. The trailer must be written * before you close the CodecContexts open when you wrote the * header; otherwise write_trailer may try to use memory that * was freed on av_codec_close() */ av_write_trailer(oc); // Mark as 'written' write_trailer = true; ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::WriteTrailer"); } // Flush encoders void FFmpegWriter::flush_encoders() { if (info.has_audio && audio_codec && AV_GET_CODEC_TYPE(audio_st) == AVMEDIA_TYPE_AUDIO && AV_GET_CODEC_ATTRIBUTES(audio_st, audio_codec)->frame_size <= 1) return; #if (LIBAVFORMAT_VERSION_MAJOR < 58) if (info.has_video && video_codec && AV_GET_CODEC_TYPE(video_st) == AVMEDIA_TYPE_VIDEO && (oc->oformat->flags & AVFMT_RAWPICTURE) && AV_FIND_DECODER_CODEC_ID(video_st) == AV_CODEC_ID_RAWVIDEO) return; #else if (info.has_video && video_codec && AV_GET_CODEC_TYPE(video_st) == AVMEDIA_TYPE_VIDEO && AV_FIND_DECODER_CODEC_ID(video_st) == AV_CODEC_ID_RAWVIDEO) return; #endif // FLUSH VIDEO ENCODER if (info.has_video) for (;;) { // Increment PTS (in frames and scaled to the codec's timebase) write_video_count += av_rescale_q(1, (AVRational) {info.fps.den, info.fps.num}, video_codec->time_base); AVPacket pkt; av_init_packet(&pkt); pkt.data = NULL; pkt.size = 0; // Pointer for video buffer (if using old FFmpeg version) uint8_t *video_outbuf = NULL; /* encode the image */ int got_packet = 0; int error_code = 0; #if IS_FFMPEG_3_2 #pragma omp critical (write_video_packet) { // Encode video packet (latest version of FFmpeg) error_code = avcodec_send_frame(video_codec, NULL); got_packet = 0; while (error_code >= 0) { error_code = avcodec_receive_packet(video_codec, &pkt); if (error_code == AVERROR(EAGAIN)|| error_code == AVERROR_EOF) { got_packet = 0; // Write packet avcodec_flush_buffers(video_codec); break; } if (pkt.pts != AV_NOPTS_VALUE) pkt.pts = av_rescale_q(pkt.pts, video_codec->time_base, video_st->time_base); if (pkt.dts != AV_NOPTS_VALUE) pkt.dts = av_rescale_q(pkt.dts, video_codec->time_base, video_st->time_base); if (pkt.duration > 0) pkt.duration = av_rescale_q(pkt.duration, video_codec->time_base, video_st->time_base); pkt.stream_index = video_st->index; error_code = av_interleaved_write_frame(oc, &pkt); } } #else // IS_FFMPEG_3_2 #if LIBAVFORMAT_VERSION_MAJOR >= 54 // Encode video packet (older than FFmpeg 3.2) error_code = avcodec_encode_video2(video_codec, &pkt, NULL, &got_packet); #else // Encode video packet (even older version of FFmpeg) int video_outbuf_size = 0; /* encode the image */ int out_size = avcodec_encode_video(video_codec, NULL, video_outbuf_size, NULL); /* if zero size, it means the image was buffered */ if (out_size > 0) { if(video_codec->coded_frame->key_frame) pkt.flags |= AV_PKT_FLAG_KEY; pkt.data= video_outbuf; pkt.size= out_size; // got data back (so encode this frame) got_packet = 1; } #endif // LIBAVFORMAT_VERSION_MAJOR >= 54 #endif // IS_FFMPEG_3_2 if (error_code < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::flush_encoders ERROR [" + (std::string) av_err2str(error_code) + "]", "error_code", error_code); } if (!got_packet) { break; } // set the timestamp if (pkt.pts != AV_NOPTS_VALUE) pkt.pts = av_rescale_q(pkt.pts, video_codec->time_base, video_st->time_base); if (pkt.dts != AV_NOPTS_VALUE) pkt.dts = av_rescale_q(pkt.dts, video_codec->time_base, video_st->time_base); if (pkt.duration > 0) pkt.duration = av_rescale_q(pkt.duration, video_codec->time_base, video_st->time_base); pkt.stream_index = video_st->index; // Write packet error_code = av_interleaved_write_frame(oc, &pkt); if (error_code < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::flush_encoders ERROR [" + (std::string)av_err2str(error_code) + "]", "error_code", error_code); } } // FLUSH AUDIO ENCODER if (info.has_audio) for (;;) { // Increment PTS (in samples and scaled to the codec's timebase) #if LIBAVFORMAT_VERSION_MAJOR >= 54 // for some reason, it requires me to multiply channels X 2 write_audio_count += av_rescale_q(audio_input_position / (audio_codec->channels * av_get_bytes_per_sample(AV_SAMPLE_FMT_S16)), (AVRational){1, info.sample_rate}, audio_codec->time_base); #else write_audio_count += av_rescale_q(audio_input_position / audio_codec->channels, (AVRational){1, info.sample_rate}, audio_codec->time_base); #endif AVPacket pkt; av_init_packet(&pkt); pkt.data = NULL; pkt.size = 0; pkt.pts = pkt.dts = write_audio_count; /* encode the image */ int error_code = 0; int got_packet = 0; #if IS_FFMPEG_3_2 error_code = avcodec_send_frame(audio_codec, NULL); #else error_code = avcodec_encode_audio2(audio_codec, &pkt, NULL, &got_packet); #endif if (error_code < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::flush_encoders ERROR [" + (std::string)av_err2str(error_code) + "]", "error_code", error_code); } if (!got_packet) { break; } // Since the PTS can change during encoding, set the value again. This seems like a huge hack, // but it fixes lots of PTS related issues when I do this. pkt.pts = pkt.dts = write_audio_count; // Scale the PTS to the audio stream timebase (which is sometimes different than the codec's timebase) if (pkt.pts != AV_NOPTS_VALUE) pkt.pts = av_rescale_q(pkt.pts, audio_codec->time_base, audio_st->time_base); if (pkt.dts != AV_NOPTS_VALUE) pkt.dts = av_rescale_q(pkt.dts, audio_codec->time_base, audio_st->time_base); if (pkt.duration > 0) pkt.duration = av_rescale_q(pkt.duration, audio_codec->time_base, audio_st->time_base); // set stream pkt.stream_index = audio_st->index; pkt.flags |= AV_PKT_FLAG_KEY; // Write packet error_code = av_interleaved_write_frame(oc, &pkt); if (error_code < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::flush_encoders ERROR [" + (std::string)av_err2str(error_code) + "]", "error_code", error_code); } // deallocate memory for packet AV_FREE_PACKET(&pkt); } } // Close the video codec void FFmpegWriter::close_video(AVFormatContext *oc, AVStream *st) { #if HAVE_HW_ACCEL if (hw_en_on && hw_en_supported) { if (hw_device_ctx) { av_buffer_unref(&hw_device_ctx); hw_device_ctx = NULL; } } #endif // HAVE_HW_ACCEL } // Close the audio codec void FFmpegWriter::close_audio(AVFormatContext *oc, AVStream *st) { // Clear buffers delete[] samples; delete[] audio_outbuf; delete[] audio_encoder_buffer; samples = NULL; audio_outbuf = NULL; audio_encoder_buffer = NULL; // Deallocate resample buffer if (avr) { SWR_CLOSE(avr); SWR_FREE(&avr); avr = NULL; } if (avr_planar) { SWR_CLOSE(avr_planar); SWR_FREE(&avr_planar); avr_planar = NULL; } } // Close the writer void FFmpegWriter::Close() { // Write trailer (if needed) if (!write_trailer) WriteTrailer(); // Close each codec if (video_st) close_video(oc, video_st); if (audio_st) close_audio(oc, audio_st); // Deallocate image scalers if (image_rescalers.size() > 0) RemoveScalers(); if (!(fmt->flags & AVFMT_NOFILE)) { /* close the output file */ avio_close(oc->pb); } // Reset frame counters write_video_count = 0; write_audio_count = 0; // Free the context which frees the streams too avformat_free_context(oc); oc = NULL; // Close writer is_open = false; prepare_streams = false; write_header = false; write_trailer = false; ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::Close"); } // Add an AVFrame to the cache void FFmpegWriter::add_avframe(std::shared_ptr frame, AVFrame *av_frame) { // Add AVFrame to map (if it does not already exist) if (!av_frames.count(frame)) { // Add av_frame av_frames[frame] = av_frame; } else { // Do not add, and deallocate this AVFrame AV_FREE_FRAME(&av_frame); } } // Add an audio output stream AVStream *FFmpegWriter::add_audio_stream() { AVCodecContext *c; AVStream *st; // Find the audio codec AVCodec *codec = avcodec_find_encoder_by_name(info.acodec.c_str()); if (codec == NULL) throw InvalidCodec("A valid audio codec could not be found for this file.", path); // Create a new audio stream AV_FORMAT_NEW_STREAM(oc, audio_codec, codec, st) c->codec_id = codec->id; #if LIBAVFORMAT_VERSION_MAJOR >= 53 c->codec_type = AVMEDIA_TYPE_AUDIO; #else c->codec_type = CODEC_TYPE_AUDIO; #endif // Set the sample parameters c->bit_rate = info.audio_bit_rate; c->channels = info.channels; // Set valid sample rate (or throw error) if (codec->supported_samplerates) { int i; for (i = 0; codec->supported_samplerates[i] != 0; i++) if (info.sample_rate == codec->supported_samplerates[i]) { // Set the valid sample rate c->sample_rate = info.sample_rate; break; } if (codec->supported_samplerates[i] == 0) throw InvalidSampleRate("An invalid sample rate was detected for this codec.", path); } else // Set sample rate c->sample_rate = info.sample_rate; // Set a valid number of channels (or throw error) const uint64_t channel_layout = info.channel_layout; if (codec->channel_layouts) { int i; for (i = 0; codec->channel_layouts[i] != 0; i++) if (channel_layout == codec->channel_layouts[i]) { // Set valid channel layout c->channel_layout = channel_layout; break; } if (codec->channel_layouts[i] == 0) throw InvalidChannels("An invalid channel layout was detected (i.e. MONO / STEREO).", path); } else // Set valid channel layout c->channel_layout = channel_layout; // Choose a valid sample_fmt if (codec->sample_fmts) { for (int i = 0; codec->sample_fmts[i] != AV_SAMPLE_FMT_NONE; i++) { // Set sample format to 1st valid format (and then exit loop) c->sample_fmt = codec->sample_fmts[i]; break; } } if (c->sample_fmt == AV_SAMPLE_FMT_NONE) { // Default if no sample formats found c->sample_fmt = AV_SAMPLE_FMT_S16; } // some formats want stream headers to be separate if (oc->oformat->flags & AVFMT_GLOBALHEADER) #if (LIBAVCODEC_VERSION_MAJOR >= 57) c->flags |= AV_CODEC_FLAG_GLOBAL_HEADER; #else c->flags |= CODEC_FLAG_GLOBAL_HEADER; #endif AV_COPY_PARAMS_FROM_CONTEXT(st, c); ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::add_audio_stream", "c->codec_id", c->codec_id, "c->bit_rate", c->bit_rate, "c->channels", c->channels, "c->sample_fmt", c->sample_fmt, "c->channel_layout", c->channel_layout, "c->sample_rate", c->sample_rate); return st; } // Add a video output stream AVStream *FFmpegWriter::add_video_stream() { AVCodecContext *c; AVStream *st; // Find the video codec AVCodec *codec = avcodec_find_encoder_by_name(info.vcodec.c_str()); if (codec == NULL) throw InvalidCodec("A valid video codec could not be found for this file.", path); // Create a new video stream AV_FORMAT_NEW_STREAM(oc, video_codec, codec, st) c->codec_id = codec->id; #if LIBAVFORMAT_VERSION_MAJOR >= 53 c->codec_type = AVMEDIA_TYPE_VIDEO; #else c->codec_type = CODEC_TYPE_VIDEO; #endif /* Init video encoder options */ if (info.video_bit_rate >= 1000 #if (LIBAVCODEC_VERSION_MAJOR >= 58) && c->codec_id != AV_CODEC_ID_AV1 #endif ) { c->bit_rate = info.video_bit_rate; if (info.video_bit_rate >= 1500000) { c->qmin = 2; c->qmax = 30; } // Here should be the setting for low fixed bitrate // Defaults are used because mpeg2 otherwise had problems } else { // Check if codec supports crf or qp switch (c->codec_id) { #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(55, 39, 101) #if (LIBAVCODEC_VERSION_MAJOR >= 58) case AV_CODEC_ID_AV1 : // TODO: Set `crf` or `qp` according to bitrate, as bitrate is not supported by these encoders yet. if (info.video_bit_rate >= 1000) { c->bit_rate = 0; if (strstr(info.vcodec.c_str(), "aom") != NULL) { int calculated_quality = 35; if (info.video_bit_rate < 500000) calculated_quality = 50; if (info.video_bit_rate > 5000000) calculated_quality = 10; av_opt_set_int(c->priv_data, "crf", calculated_quality, 0); info.video_bit_rate = calculated_quality; } else { int calculated_quality = 50; if (info.video_bit_rate < 500000) calculated_quality = 60; if (info.video_bit_rate > 5000000) calculated_quality = 15; av_opt_set_int(c->priv_data, "qp", calculated_quality, 0); info.video_bit_rate = calculated_quality; } // medium } if (strstr(info.vcodec.c_str(), "svt_av1") != NULL) { av_opt_set_int(c->priv_data, "preset", 6, 0); av_opt_set_int(c->priv_data, "forced-idr",1,0); } else if (strstr(info.vcodec.c_str(), "rav1e") != NULL) { av_opt_set_int(c->priv_data, "speed", 7, 0); av_opt_set_int(c->priv_data, "tile-rows", 2, 0); av_opt_set_int(c->priv_data, "tile-columns", 4, 0); } else if (strstr(info.vcodec.c_str(), "aom") != NULL) { // Set number of tiles to a fixed value // TODO: Allow user to chose their own number of tiles av_opt_set_int(c->priv_data, "tile-rows", 1, 0); // log2 of number of rows av_opt_set_int(c->priv_data, "tile-columns", 2, 0); // log2 of number of columns av_opt_set_int(c->priv_data, "row-mt", 1, 0); // use multiple cores av_opt_set_int(c->priv_data, "cpu-used", 3, 0); // default is 1, usable is 4 } //break; #endif case AV_CODEC_ID_VP9 : case AV_CODEC_ID_HEVC : #endif case AV_CODEC_ID_VP8 : case AV_CODEC_ID_H264 : if (info.video_bit_rate < 40) { c->qmin = 0; c->qmax = 63; } else { c->qmin = info.video_bit_rate - 5; c->qmax = 63; } break; default: // Here should be the setting for codecs that don't support crf // For now defaults are used break; } } //TODO: Implement variable bitrate feature (which actually works). This implementation throws //invalid bitrate errors and rc buffer underflow errors, etc... //c->rc_min_rate = info.video_bit_rate; //c->rc_max_rate = info.video_bit_rate; //c->rc_buffer_size = FFMAX(c->rc_max_rate, 15000000) * 112L / 15000000 * 16384; //if ( !c->rc_initial_buffer_occupancy ) // c->rc_initial_buffer_occupancy = c->rc_buffer_size * 3/4; /* resolution must be a multiple of two */ // TODO: require /2 height and width c->width = info.width; c->height = info.height; /* time base: this is the fundamental unit of time (in seconds) in terms of which frame timestamps are represented. for fixed-fps content, timebase should be 1/framerate and timestamp increments should be identically 1. */ c->time_base.num = info.video_timebase.num; c->time_base.den = info.video_timebase.den; // AVCodecContext->framerate was added in FFmpeg 2.2 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(56, 26, 0) c->framerate = av_inv_q(c->time_base); #endif st->avg_frame_rate = av_inv_q(c->time_base); st->time_base.num = info.video_timebase.num; st->time_base.den = info.video_timebase.den; #if (LIBAVFORMAT_VERSION_MAJOR >= 58) #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-Wdeprecated-declarations" st->codec->time_base.num = info.video_timebase.num; st->codec->time_base.den = info.video_timebase.den; #pragma GCC diagnostic pop #endif c->gop_size = 12; /* TODO: add this to "info"... emit one intra frame every twelve frames at most */ c->max_b_frames = 10; if (c->codec_id == AV_CODEC_ID_MPEG2VIDEO) /* just for testing, we also add B frames */ c->max_b_frames = 2; if (c->codec_id == AV_CODEC_ID_MPEG1VIDEO) /* Needed to avoid using macroblocks in which some coeffs overflow. This does not happen with normal video, it just happens here as the motion of the chroma plane does not match the luma plane. */ c->mb_decision = 2; // some formats want stream headers to be separate if (oc->oformat->flags & AVFMT_GLOBALHEADER) #if (LIBAVCODEC_VERSION_MAJOR >= 57) c->flags |= AV_CODEC_FLAG_GLOBAL_HEADER; #else c->flags |= CODEC_FLAG_GLOBAL_HEADER; #endif // Find all supported pixel formats for this codec const PixelFormat *supported_pixel_formats = codec->pix_fmts; while (supported_pixel_formats != NULL && *supported_pixel_formats != PIX_FMT_NONE) { // Assign the 1st valid pixel format (if one is missing) if (c->pix_fmt == PIX_FMT_NONE) c->pix_fmt = *supported_pixel_formats; ++supported_pixel_formats; } // Codec doesn't have any pix formats? if (c->pix_fmt == PIX_FMT_NONE) { if (fmt->video_codec == AV_CODEC_ID_RAWVIDEO) { // Raw video should use RGB24 c->pix_fmt = PIX_FMT_RGB24; #if (LIBAVFORMAT_VERSION_MAJOR < 58) if (strcmp(fmt->name, "gif") != 0) // If not GIF format, skip the encoding process // Set raw picture flag (so we don't encode this video) oc->oformat->flags |= AVFMT_RAWPICTURE; #endif } else { // Set the default codec c->pix_fmt = PIX_FMT_YUV420P; } } AV_COPY_PARAMS_FROM_CONTEXT(st, c); #if (LIBAVFORMAT_VERSION_MAJOR < 58) ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::add_video_stream (" + (std::string)fmt->name + " : " + (std::string)av_get_pix_fmt_name(c->pix_fmt) + ")", "c->codec_id", c->codec_id, "c->bit_rate", c->bit_rate, "c->pix_fmt", c->pix_fmt, "oc->oformat->flags", oc->oformat->flags, "AVFMT_RAWPICTURE", AVFMT_RAWPICTURE); #else ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::add_video_stream (" + (std::string)fmt->name + " : " + (std::string)av_get_pix_fmt_name(c->pix_fmt) + ")", "c->codec_id", c->codec_id, "c->bit_rate", c->bit_rate, "c->pix_fmt", c->pix_fmt, "oc->oformat->flags", oc->oformat->flags); #endif return st; } // open audio codec void FFmpegWriter::open_audio(AVFormatContext *oc, AVStream *st) { AVCodec *codec; AV_GET_CODEC_FROM_STREAM(st, audio_codec) // Set number of threads equal to number of processors (not to exceed 16) audio_codec->thread_count = std::min(FF_NUM_PROCESSORS, 16); // Find the audio encoder codec = avcodec_find_encoder_by_name(info.acodec.c_str()); if (!codec) codec = avcodec_find_encoder(audio_codec->codec_id); if (!codec) throw InvalidCodec("Could not find codec", path); // Init options AVDictionary *opts = NULL; av_dict_set(&opts, "strict", "experimental", 0); // Open the codec if (avcodec_open2(audio_codec, codec, &opts) < 0) throw InvalidCodec("Could not open audio codec", path); AV_COPY_PARAMS_FROM_CONTEXT(st, audio_codec); // Free options av_dict_free(&opts); // Calculate the size of the input frame (i..e how many samples per packet), and the output buffer // TODO: Ugly hack for PCM codecs (will be removed ASAP with new PCM support to compute the input frame size in samples if (audio_codec->frame_size <= 1) { // No frame size found... so calculate audio_input_frame_size = 50000 / info.channels; int s = AV_FIND_DECODER_CODEC_ID(st); switch (s) { case AV_CODEC_ID_PCM_S16LE: case AV_CODEC_ID_PCM_S16BE: case AV_CODEC_ID_PCM_U16LE: case AV_CODEC_ID_PCM_U16BE: audio_input_frame_size >>= 1; break; default: break; } } else { // Set frame size based on the codec audio_input_frame_size = audio_codec->frame_size; } // Set the initial frame size (since it might change during resampling) initial_audio_input_frame_size = audio_input_frame_size; // Allocate array for samples samples = new int16_t[AVCODEC_MAX_AUDIO_FRAME_SIZE]; // Set audio output buffer (used to store the encoded audio) audio_outbuf_size = AVCODEC_MAX_AUDIO_FRAME_SIZE; audio_outbuf = new uint8_t[audio_outbuf_size]; // Set audio packet encoding buffer audio_encoder_buffer_size = AUDIO_PACKET_ENCODING_SIZE; audio_encoder_buffer = new uint8_t[audio_encoder_buffer_size]; // Add audio metadata (if any) for (std::map::iterator iter = info.metadata.begin(); iter != info.metadata.end(); ++iter) { av_dict_set(&st->metadata, iter->first.c_str(), iter->second.c_str(), 0); } ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::open_audio", "audio_codec->thread_count", audio_codec->thread_count, "audio_input_frame_size", audio_input_frame_size, "buffer_size", AVCODEC_MAX_AUDIO_FRAME_SIZE + MY_INPUT_BUFFER_PADDING_SIZE); } // open video codec void FFmpegWriter::open_video(AVFormatContext *oc, AVStream *st) { AVCodec *codec; AV_GET_CODEC_FROM_STREAM(st, video_codec) // Set number of threads equal to number of processors (not to exceed 16) video_codec->thread_count = std::min(FF_NUM_PROCESSORS, 16); #if HAVE_HW_ACCEL if (hw_en_on && hw_en_supported) { //char *dev_hw = NULL; char adapter[256]; char *adapter_ptr = NULL; int adapter_num; // Use the hw device given in the environment variable HW_EN_DEVICE_SET or the default if not set adapter_num = openshot::Settings::Instance()->HW_EN_DEVICE_SET; fprintf(stderr, "\n\nEncodiing Device Nr: %d\n", adapter_num); if (adapter_num < 3 && adapter_num >=0) { #if defined(__linux__) snprintf(adapter,sizeof(adapter),"/dev/dri/renderD%d", adapter_num+128); // Maybe 127 is better because the first card would be 1?! adapter_ptr = adapter; #elif defined(_WIN32) adapter_ptr = NULL; #elif defined(__APPLE__) adapter_ptr = NULL; #endif } else { adapter_ptr = NULL; // Just to be sure } // Check if it is there and writable #if defined(__linux__) if( adapter_ptr != NULL && access( adapter_ptr, W_OK ) == 0 ) { #elif defined(_WIN32) if( adapter_ptr != NULL ) { #elif defined(__APPLE__) if( adapter_ptr != NULL ) { #endif ZmqLogger::Instance()->AppendDebugMethod("Encode Device present using device", "adapter", adapter_num); } else { adapter_ptr = NULL; // use default ZmqLogger::Instance()->AppendDebugMethod("Encode Device not present using default"); } if (av_hwdevice_ctx_create(&hw_device_ctx, hw_en_av_device_type, adapter_ptr, NULL, 0) < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::open_video : Codec name: ", info.vcodec.c_str(), -1, " ERROR creating\n", -1); throw InvalidCodec("Could not create hwdevice", path); } } #endif // HAVE_HW_ACCEL /* find the video encoder */ codec = avcodec_find_encoder_by_name(info.vcodec.c_str()); if (!codec) codec = avcodec_find_encoder(AV_FIND_DECODER_CODEC_ID(st)); if (!codec) throw InvalidCodec("Could not find codec", path); /* Force max_b_frames to 0 in some cases (i.e. for mjpeg image sequences */ if (video_codec->max_b_frames && video_codec->codec_id != AV_CODEC_ID_MPEG4 && video_codec->codec_id != AV_CODEC_ID_MPEG1VIDEO && video_codec->codec_id != AV_CODEC_ID_MPEG2VIDEO) video_codec->max_b_frames = 0; // Init options AVDictionary *opts = NULL; av_dict_set(&opts, "strict", "experimental", 0); #if HAVE_HW_ACCEL if (hw_en_on && hw_en_supported) { video_codec->pix_fmt = hw_en_av_pix_fmt; // for the list of possible options, see the list of codec-specific options: // e.g. ffmpeg -h encoder=h264_vaapi or ffmpeg -h encoder=hevc_vaapi // and "man ffmpeg-codecs" // For VAAPI, it is safer to explicitly set rc_mode instead of relying on auto-selection // which is ffmpeg version-specific. if (hw_en_av_pix_fmt == AV_PIX_FMT_VAAPI) { int64_t qp; if (av_opt_get_int(video_codec->priv_data, "qp", 0, &qp) != 0 || qp == 0) { // unless "qp" was set for CQP, switch to VBR RC mode av_opt_set(video_codec->priv_data, "rc_mode", "VBR", 0); // In the current state (ffmpeg-4.2-4 libva-mesa-driver-19.1.5-1) to use VBR, // one has to specify both bit_rate and maxrate, otherwise a small low quality file is generated on Intel iGPU). video_codec->rc_max_rate = video_codec->bit_rate; } } switch (video_codec->codec_id) { case AV_CODEC_ID_H264: video_codec->max_b_frames = 0; // At least this GPU doesn't support b-frames video_codec->profile = FF_PROFILE_H264_BASELINE | FF_PROFILE_H264_CONSTRAINED; av_opt_set(video_codec->priv_data, "preset", "slow", 0); av_opt_set(video_codec->priv_data, "tune", "zerolatency", 0); av_opt_set(video_codec->priv_data, "vprofile", "baseline", AV_OPT_SEARCH_CHILDREN); break; case AV_CODEC_ID_HEVC: // tested to work with defaults break; case AV_CODEC_ID_VP9: // tested to work with defaults break; default: ZmqLogger::Instance()->AppendDebugMethod("No codec-specific options defined for this codec. HW encoding may fail", "codec_id", video_codec->codec_id); break; } // set hw_frames_ctx for encoder's AVCodecContext int err; if ((err = set_hwframe_ctx(video_codec, hw_device_ctx, info.width, info.height)) < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::open_video (set_hwframe_ctx) ERROR faled to set hwframe context", "width", info.width, "height", info.height, av_err2str(err), -1); } } #endif // HAVE_HW_ACCEL /* open the codec */ if (avcodec_open2(video_codec, codec, &opts) < 0) throw InvalidCodec("Could not open video codec", path); AV_COPY_PARAMS_FROM_CONTEXT(st, video_codec); // Free options av_dict_free(&opts); // Add video metadata (if any) for (std::map::iterator iter = info.metadata.begin(); iter != info.metadata.end(); ++iter) { av_dict_set(&st->metadata, iter->first.c_str(), iter->second.c_str(), 0); } ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::open_video", "video_codec->thread_count", video_codec->thread_count); } // write all queued frames' audio to the video file void FFmpegWriter::write_audio_packets(bool is_final) { #pragma omp task firstprivate(is_final) { // Init audio buffers / variables int total_frame_samples = 0; int frame_position = 0; int channels_in_frame = 0; int sample_rate_in_frame = 0; int samples_in_frame = 0; ChannelLayout channel_layout_in_frame = LAYOUT_MONO; // default channel layout // Create a new array (to hold all S16 audio samples, for the current queued frames unsigned int all_queued_samples_size = sizeof(int16_t) * (queued_audio_frames.size() * AVCODEC_MAX_AUDIO_FRAME_SIZE); int16_t *all_queued_samples = (int16_t *) av_malloc(all_queued_samples_size); int16_t *all_resampled_samples = NULL; int16_t *final_samples_planar = NULL; int16_t *final_samples = NULL; // Loop through each queued audio frame while (!queued_audio_frames.empty()) { // Get front frame (from the queue) std::shared_ptr frame = queued_audio_frames.front(); // Get the audio details from this frame sample_rate_in_frame = frame->SampleRate(); samples_in_frame = frame->GetAudioSamplesCount(); channels_in_frame = frame->GetAudioChannelsCount(); channel_layout_in_frame = frame->ChannelsLayout(); // Get audio sample array float *frame_samples_float = NULL; // Get samples interleaved together (c1 c2 c1 c2 c1 c2) frame_samples_float = frame->GetInterleavedAudioSamples(sample_rate_in_frame, NULL, &samples_in_frame); // Calculate total samples total_frame_samples = samples_in_frame * channels_in_frame; // Translate audio sample values back to 16 bit integers with saturation const int16_t max16 = 32767; const int16_t min16 = -32768; for (int s = 0; s < total_frame_samples; s++, frame_position++) { float valF = frame_samples_float[s] * (1 << 15); int16_t conv; if (valF > max16) { conv = max16; } else if (valF < min16) { conv = min16; } else { conv = int(valF + 32768.5) - 32768; // +0.5 is for rounding } // Copy into buffer all_queued_samples[frame_position] = conv; } // Deallocate float array delete[] frame_samples_float; // Remove front item queued_audio_frames.pop_front(); } // end while // Update total samples (since we've combined all queued frames) total_frame_samples = frame_position; int remaining_frame_samples = total_frame_samples; int samples_position = 0; ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_audio_packets", "is_final", is_final, "total_frame_samples", total_frame_samples, "channel_layout_in_frame", channel_layout_in_frame, "channels_in_frame", channels_in_frame, "samples_in_frame", samples_in_frame, "LAYOUT_MONO", LAYOUT_MONO); // Keep track of the original sample format AVSampleFormat output_sample_fmt = audio_codec->sample_fmt; AVFrame *audio_frame = NULL; if (!is_final) { // Create input frame (and allocate arrays) audio_frame = AV_ALLOCATE_FRAME(); AV_RESET_FRAME(audio_frame); audio_frame->nb_samples = total_frame_samples / channels_in_frame; // Fill input frame with sample data int error_code = avcodec_fill_audio_frame(audio_frame, channels_in_frame, AV_SAMPLE_FMT_S16, (uint8_t *) all_queued_samples, all_queued_samples_size, 0); if (error_code < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_audio_packets ERROR [" + (std::string) av_err2str(error_code) + "]", "error_code", error_code); } // Do not convert audio to planar format (yet). We need to keep everything interleaved at this point. switch (audio_codec->sample_fmt) { case AV_SAMPLE_FMT_FLTP: { output_sample_fmt = AV_SAMPLE_FMT_FLT; break; } case AV_SAMPLE_FMT_S32P: { output_sample_fmt = AV_SAMPLE_FMT_S32; break; } case AV_SAMPLE_FMT_S16P: { output_sample_fmt = AV_SAMPLE_FMT_S16; break; } case AV_SAMPLE_FMT_U8P: { output_sample_fmt = AV_SAMPLE_FMT_U8; break; } default: { // This is only here to silence unused-enum warnings break; } } // Update total samples & input frame size (due to bigger or smaller data types) total_frame_samples *= (float(info.sample_rate) / sample_rate_in_frame); // adjust for different byte sizes total_frame_samples *= (float(info.channels) / channels_in_frame); // adjust for different # of channels // Create output frame (and allocate arrays) AVFrame *audio_converted = AV_ALLOCATE_FRAME(); AV_RESET_FRAME(audio_converted); audio_converted->nb_samples = total_frame_samples / channels_in_frame; av_samples_alloc(audio_converted->data, audio_converted->linesize, info.channels, audio_converted->nb_samples, output_sample_fmt, 0); ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_audio_packets (1st resampling)", "in_sample_fmt", AV_SAMPLE_FMT_S16, "out_sample_fmt", output_sample_fmt, "in_sample_rate", sample_rate_in_frame, "out_sample_rate", info.sample_rate, "in_channels", channels_in_frame, "out_channels", info.channels); // setup resample context if (!avr) { avr = SWR_ALLOC(); av_opt_set_int(avr, "in_channel_layout", channel_layout_in_frame, 0); av_opt_set_int(avr, "out_channel_layout", info.channel_layout, 0); av_opt_set_int(avr, "in_sample_fmt", AV_SAMPLE_FMT_S16, 0); av_opt_set_int(avr, "out_sample_fmt", output_sample_fmt, 0); // planar not allowed here av_opt_set_int(avr, "in_sample_rate", sample_rate_in_frame, 0); av_opt_set_int(avr, "out_sample_rate", info.sample_rate, 0); av_opt_set_int(avr, "in_channels", channels_in_frame, 0); av_opt_set_int(avr, "out_channels", info.channels, 0); SWR_INIT(avr); } int nb_samples = 0; // Convert audio samples nb_samples = SWR_CONVERT(avr, // audio resample context audio_converted->data, // output data pointers audio_converted->linesize[0], // output plane size, in bytes. (0 if unknown) audio_converted->nb_samples, // maximum number of samples that the output buffer can hold audio_frame->data, // input data pointers audio_frame->linesize[0], // input plane size, in bytes (0 if unknown) audio_frame->nb_samples); // number of input samples to convert // Set remaining samples remaining_frame_samples = nb_samples * av_get_bytes_per_sample(AV_SAMPLE_FMT_S16); // Create a new array (to hold all resampled S16 audio samples) all_resampled_samples = (int16_t *) av_malloc( sizeof(int16_t) * nb_samples * info.channels * (av_get_bytes_per_sample(output_sample_fmt) / av_get_bytes_per_sample(AV_SAMPLE_FMT_S16))); // Copy audio samples over original samples memcpy(all_resampled_samples, audio_converted->data[0], nb_samples * info.channels * av_get_bytes_per_sample(output_sample_fmt)); // Remove converted audio av_freep(&(audio_frame->data[0])); AV_FREE_FRAME(&audio_frame); av_freep(&audio_converted->data[0]); AV_FREE_FRAME(&audio_converted); all_queued_samples = NULL; // this array cleared with above call ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_audio_packets (Successfully completed 1st resampling)", "nb_samples", nb_samples, "remaining_frame_samples", remaining_frame_samples); } // Loop until no more samples while (remaining_frame_samples > 0 || is_final) { // Get remaining samples needed for this packet int remaining_packet_samples = (audio_input_frame_size * info.channels) - audio_input_position; // Determine how many samples we need int diff = 0; if (remaining_frame_samples >= remaining_packet_samples) { diff = remaining_packet_samples; } else { diff = remaining_frame_samples; } // Copy frame samples into the packet samples array if (!is_final) //TODO: Make this more sane memcpy(samples + (audio_input_position * (av_get_bytes_per_sample(output_sample_fmt) / av_get_bytes_per_sample(AV_SAMPLE_FMT_S16))), all_resampled_samples + samples_position, diff * av_get_bytes_per_sample(output_sample_fmt)); // Increment counters audio_input_position += diff; samples_position += diff * (av_get_bytes_per_sample(output_sample_fmt) / av_get_bytes_per_sample(AV_SAMPLE_FMT_S16)); remaining_frame_samples -= diff; // Do we have enough samples to proceed? if (audio_input_position < (audio_input_frame_size * info.channels) && !is_final) // Not enough samples to encode... so wait until the next frame break; // Convert to planar (if needed by audio codec) AVFrame *frame_final = AV_ALLOCATE_FRAME(); AV_RESET_FRAME(frame_final); if (av_sample_fmt_is_planar(audio_codec->sample_fmt)) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_audio_packets (2nd resampling for Planar formats)", "in_sample_fmt", output_sample_fmt, "out_sample_fmt", audio_codec->sample_fmt, "in_sample_rate", info.sample_rate, "out_sample_rate", info.sample_rate, "in_channels", info.channels, "out_channels", info.channels); // setup resample context if (!avr_planar) { avr_planar = SWR_ALLOC(); av_opt_set_int(avr_planar, "in_channel_layout", info.channel_layout, 0); av_opt_set_int(avr_planar, "out_channel_layout", info.channel_layout, 0); av_opt_set_int(avr_planar, "in_sample_fmt", output_sample_fmt, 0); av_opt_set_int(avr_planar, "out_sample_fmt", audio_codec->sample_fmt, 0); // planar not allowed here av_opt_set_int(avr_planar, "in_sample_rate", info.sample_rate, 0); av_opt_set_int(avr_planar, "out_sample_rate", info.sample_rate, 0); av_opt_set_int(avr_planar, "in_channels", info.channels, 0); av_opt_set_int(avr_planar, "out_channels", info.channels, 0); SWR_INIT(avr_planar); } // Create input frame (and allocate arrays) audio_frame = AV_ALLOCATE_FRAME(); AV_RESET_FRAME(audio_frame); audio_frame->nb_samples = audio_input_position / info.channels; // Create a new array final_samples_planar = (int16_t *) av_malloc( sizeof(int16_t) * audio_frame->nb_samples * info.channels * (av_get_bytes_per_sample(output_sample_fmt) / av_get_bytes_per_sample(AV_SAMPLE_FMT_S16))); // Copy audio into buffer for frame memcpy(final_samples_planar, samples, audio_frame->nb_samples * info.channels * av_get_bytes_per_sample(output_sample_fmt)); // Fill input frame with sample data avcodec_fill_audio_frame(audio_frame, info.channels, output_sample_fmt, (uint8_t *) final_samples_planar, audio_encoder_buffer_size, 0); // Create output frame (and allocate arrays) frame_final->nb_samples = audio_input_frame_size; av_samples_alloc(frame_final->data, frame_final->linesize, info.channels, frame_final->nb_samples, audio_codec->sample_fmt, 0); // Convert audio samples int nb_samples = SWR_CONVERT(avr_planar, // audio resample context frame_final->data, // output data pointers frame_final->linesize[0], // output plane size, in bytes. (0 if unknown) frame_final->nb_samples, // maximum number of samples that the output buffer can hold audio_frame->data, // input data pointers audio_frame->linesize[0], // input plane size, in bytes (0 if unknown) audio_frame->nb_samples); // number of input samples to convert // Copy audio samples over original samples if (nb_samples > 0) memcpy(samples, frame_final->data[0], nb_samples * av_get_bytes_per_sample(audio_codec->sample_fmt) * info.channels); // deallocate AVFrame av_freep(&(audio_frame->data[0])); AV_FREE_FRAME(&audio_frame); all_queued_samples = NULL; // this array cleared with above call ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_audio_packets (Successfully completed 2nd resampling for Planar formats)", "nb_samples", nb_samples); } else { // Create a new array final_samples = (int16_t *) av_malloc( sizeof(int16_t) * audio_input_position * (av_get_bytes_per_sample(audio_codec->sample_fmt) / av_get_bytes_per_sample(AV_SAMPLE_FMT_S16))); // Copy audio into buffer for frame memcpy(final_samples, samples, audio_input_position * av_get_bytes_per_sample(audio_codec->sample_fmt)); // Init the nb_samples property frame_final->nb_samples = audio_input_frame_size; // Fill the final_frame AVFrame with audio (non planar) avcodec_fill_audio_frame(frame_final, audio_codec->channels, audio_codec->sample_fmt, (uint8_t *) final_samples, audio_encoder_buffer_size, 0); } // Increment PTS (in samples) write_audio_count += FFMIN(audio_input_frame_size, audio_input_position); frame_final->pts = write_audio_count; // Set the AVFrame's PTS // Init the packet AVPacket pkt; av_init_packet(&pkt); pkt.data = audio_encoder_buffer; pkt.size = audio_encoder_buffer_size; // Set the packet's PTS prior to encoding pkt.pts = pkt.dts = write_audio_count; /* encode the audio samples */ int got_packet_ptr = 0; #if IS_FFMPEG_3_2 // Encode audio (latest version of FFmpeg) int error_code; int ret = 0; int frame_finished = 0; error_code = ret = avcodec_send_frame(audio_codec, frame_final); if (ret < 0 && ret != AVERROR(EINVAL) && ret != AVERROR_EOF) { avcodec_send_frame(audio_codec, NULL); } else { if (ret >= 0) pkt.size = 0; ret = avcodec_receive_packet(audio_codec, &pkt); if (ret >= 0) frame_finished = 1; if(ret == AVERROR(EINVAL) || ret == AVERROR_EOF) { avcodec_flush_buffers(audio_codec); ret = 0; } if (ret >= 0) { ret = frame_finished; } } if (!pkt.data && !frame_finished) { ret = -1; } got_packet_ptr = ret; #else // Encode audio (older versions of FFmpeg) int error_code = avcodec_encode_audio2(audio_codec, &pkt, frame_final, &got_packet_ptr); #endif /* if zero size, it means the image was buffered */ if (error_code == 0 && got_packet_ptr) { // Since the PTS can change during encoding, set the value again. This seems like a huge hack, // but it fixes lots of PTS related issues when I do this. pkt.pts = pkt.dts = write_audio_count; // Scale the PTS to the audio stream timebase (which is sometimes different than the codec's timebase) if (pkt.pts != AV_NOPTS_VALUE) pkt.pts = av_rescale_q(pkt.pts, audio_codec->time_base, audio_st->time_base); if (pkt.dts != AV_NOPTS_VALUE) pkt.dts = av_rescale_q(pkt.dts, audio_codec->time_base, audio_st->time_base); if (pkt.duration > 0) pkt.duration = av_rescale_q(pkt.duration, audio_codec->time_base, audio_st->time_base); // set stream pkt.stream_index = audio_st->index; pkt.flags |= AV_PKT_FLAG_KEY; /* write the compressed frame in the media file */ int error_code = av_interleaved_write_frame(oc, &pkt); if (error_code < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_audio_packets ERROR [" + (std::string) av_err2str(error_code) + "]", "error_code", error_code); } } if (error_code < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_audio_packets ERROR [" + (std::string) av_err2str(error_code) + "]", "error_code", error_code); } // deallocate AVFrame av_freep(&(frame_final->data[0])); AV_FREE_FRAME(&frame_final); // deallocate memory for packet AV_FREE_PACKET(&pkt); // Reset position audio_input_position = 0; is_final = false; } // Delete arrays (if needed) if (all_resampled_samples) { av_freep(&all_resampled_samples); all_resampled_samples = NULL; } if (all_queued_samples) { av_freep(&all_queued_samples); all_queued_samples = NULL; } } // end task } // Allocate an AVFrame object AVFrame *FFmpegWriter::allocate_avframe(PixelFormat pix_fmt, int width, int height, int *buffer_size, uint8_t *new_buffer) { // Create an RGB AVFrame AVFrame *new_av_frame = NULL; // Allocate an AVFrame structure new_av_frame = AV_ALLOCATE_FRAME(); if (new_av_frame == NULL) throw OutOfMemory("Could not allocate AVFrame", path); // Determine required buffer size and allocate buffer *buffer_size = AV_GET_IMAGE_SIZE(pix_fmt, width, height); // Create buffer (if not provided) if (!new_buffer) { // New Buffer new_buffer = (uint8_t *) av_malloc(*buffer_size * sizeof(uint8_t)); // Attach buffer to AVFrame AV_COPY_PICTURE_DATA(new_av_frame, new_buffer, pix_fmt, width, height); new_av_frame->width = width; new_av_frame->height = height; new_av_frame->format = pix_fmt; } // return AVFrame return new_av_frame; } // process video frame void FFmpegWriter::process_video_packet(std::shared_ptr frame) { // Determine the height & width of the source image int source_image_width = frame->GetWidth(); int source_image_height = frame->GetHeight(); // Do nothing if size is 1x1 (i.e. no image in this frame) if (source_image_height == 1 && source_image_width == 1) return; // Init rescalers (if not initialized yet) if (image_rescalers.size() == 0) InitScalers(source_image_width, source_image_height); // Get a unique rescaler (for this thread) SwsContext *scaler = image_rescalers[rescaler_position]; rescaler_position++; if (rescaler_position == num_of_rescalers) rescaler_position = 0; #pragma omp task firstprivate(frame, scaler, source_image_width, source_image_height) { // Allocate an RGB frame & final output frame int bytes_source = 0; int bytes_final = 0; AVFrame *frame_source = NULL; const uchar *pixels = NULL; // Get a list of pixels from source image pixels = frame->GetPixels(); // Init AVFrame for source image & final (converted image) frame_source = allocate_avframe(PIX_FMT_RGBA, source_image_width, source_image_height, &bytes_source, (uint8_t *) pixels); #if IS_FFMPEG_3_2 AVFrame *frame_final; #if HAVE_HW_ACCEL if (hw_en_on && hw_en_supported) { frame_final = allocate_avframe(AV_PIX_FMT_NV12, info.width, info.height, &bytes_final, NULL); } else #endif // HAVE_HW_ACCEL { frame_final = allocate_avframe((AVPixelFormat)(video_st->codecpar->format), info.width, info.height, &bytes_final, NULL); } #else AVFrame *frame_final = allocate_avframe(video_codec->pix_fmt, info.width, info.height, &bytes_final, NULL); #endif // IS_FFMPEG_3_2 // Fill with data AV_COPY_PICTURE_DATA(frame_source, (uint8_t *) pixels, PIX_FMT_RGBA, source_image_width, source_image_height); ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::process_video_packet", "frame->number", frame->number, "bytes_source", bytes_source, "bytes_final", bytes_final); // Resize & convert pixel format sws_scale(scaler, frame_source->data, frame_source->linesize, 0, source_image_height, frame_final->data, frame_final->linesize); // Add resized AVFrame to av_frames map #pragma omp critical (av_frames_section) add_avframe(frame, frame_final); // Deallocate memory AV_FREE_FRAME(&frame_source); } // end task } // write video frame bool FFmpegWriter::write_video_packet(std::shared_ptr frame, AVFrame *frame_final) { #if (LIBAVFORMAT_VERSION_MAJOR >= 58) ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_video_packet", "frame->number", frame->number, "oc->oformat->flags", oc->oformat->flags); if (AV_GET_CODEC_TYPE(video_st) == AVMEDIA_TYPE_VIDEO && AV_FIND_DECODER_CODEC_ID(video_st) == AV_CODEC_ID_RAWVIDEO) { #else ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_video_packet", "frame->number", frame->number, "oc->oformat->flags & AVFMT_RAWPICTURE", oc->oformat->flags & AVFMT_RAWPICTURE); if (oc->oformat->flags & AVFMT_RAWPICTURE) { #endif // Raw video case. AVPacket pkt; av_init_packet(&pkt); pkt.flags |= AV_PKT_FLAG_KEY; pkt.stream_index = video_st->index; pkt.data = (uint8_t *) frame_final->data; pkt.size = sizeof(AVPicture); // Increment PTS (in frames and scaled to the codec's timebase) write_video_count += av_rescale_q(1, (AVRational) {info.fps.den, info.fps.num}, video_codec->time_base); pkt.pts = write_video_count; /* write the compressed frame in the media file */ int error_code = av_interleaved_write_frame(oc, &pkt); if (error_code < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_video_packet ERROR [" + (std::string) av_err2str(error_code) + "]", "error_code", error_code); return false; } // Deallocate packet AV_FREE_PACKET(&pkt); } else { AVPacket pkt; av_init_packet(&pkt); pkt.data = NULL; pkt.size = 0; pkt.pts = pkt.dts = AV_NOPTS_VALUE; // Pointer for video buffer (if using old FFmpeg version) uint8_t *video_outbuf = NULL; // Increment PTS (in frames and scaled to the codec's timebase) write_video_count += av_rescale_q(1, (AVRational) {info.fps.den, info.fps.num}, video_codec->time_base); // Assign the initial AVFrame PTS from the frame counter frame_final->pts = write_video_count; #if HAVE_HW_ACCEL if (hw_en_on && hw_en_supported) { if (!(hw_frame = av_frame_alloc())) { fprintf(stderr, "Error code: av_hwframe_alloc\n"); } if (av_hwframe_get_buffer(video_codec->hw_frames_ctx, hw_frame, 0) < 0) { fprintf(stderr, "Error code: av_hwframe_get_buffer\n"); } if (!hw_frame->hw_frames_ctx) { fprintf(stderr, "Error hw_frames_ctx.\n"); } hw_frame->format = AV_PIX_FMT_NV12; if ( av_hwframe_transfer_data(hw_frame, frame_final, 0) < 0) { fprintf(stderr, "Error while transferring frame data to surface.\n"); } av_frame_copy_props(hw_frame, frame_final); } #endif // HAVE_HW_ACCEL /* encode the image */ int got_packet_ptr = 0; int error_code = 0; #if IS_FFMPEG_3_2 // Write video packet (latest version of FFmpeg) int ret; #if HAVE_HW_ACCEL if (hw_en_on && hw_en_supported) { ret = avcodec_send_frame(video_codec, hw_frame); //hw_frame!!! } else #endif // HAVE_HW_ACCEL { ret = avcodec_send_frame(video_codec, frame_final); } error_code = ret; if (ret < 0 ) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_video_packet (Frame not sent)"); if (ret == AVERROR(EAGAIN) ) { std::cerr << "Frame EAGAIN" << "\n"; } if (ret == AVERROR_EOF ) { std::cerr << "Frame AVERROR_EOF" << "\n"; } avcodec_send_frame(video_codec, NULL); } else { while (ret >= 0) { ret = avcodec_receive_packet(video_codec, &pkt); if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF) { avcodec_flush_buffers(video_codec); got_packet_ptr = 0; break; } if (ret == 0) { got_packet_ptr = 1; break; } } } #else #if LIBAVFORMAT_VERSION_MAJOR >= 54 // Write video packet (older than FFmpeg 3.2) error_code = avcodec_encode_video2(video_codec, &pkt, frame_final, &got_packet_ptr); if (error_code != 0) { std::cerr << "Frame AVERROR_EOF" << "\n"; } if (got_packet_ptr == 0) { std::cerr << "Frame gotpacket error" << "\n"; } #else // Write video packet (even older versions of FFmpeg) int video_outbuf_size = 200000; video_outbuf = (uint8_t*) av_malloc(200000); /* encode the image */ int out_size = avcodec_encode_video(video_codec, video_outbuf, video_outbuf_size, frame_final); /* if zero size, it means the image was buffered */ if (out_size > 0) { if(video_codec->coded_frame->key_frame) pkt.flags |= AV_PKT_FLAG_KEY; pkt.data= video_outbuf; pkt.size= out_size; // got data back (so encode this frame) got_packet_ptr = 1; } #endif // LIBAVFORMAT_VERSION_MAJOR >= 54 #endif // IS_FFMPEG_3_2 /* if zero size, it means the image was buffered */ if (error_code == 0 && got_packet_ptr) { // Since the PTS can change during encoding, set the value again. This seems like a huge hack, // but it fixes lots of PTS related issues when I do this. //pkt.pts = pkt.dts = write_video_count; // set the timestamp if (pkt.pts != AV_NOPTS_VALUE) pkt.pts = av_rescale_q(pkt.pts, video_codec->time_base, video_st->time_base); if (pkt.dts != AV_NOPTS_VALUE) pkt.dts = av_rescale_q(pkt.dts, video_codec->time_base, video_st->time_base); if (pkt.duration > 0) pkt.duration = av_rescale_q(pkt.duration, video_codec->time_base, video_st->time_base); pkt.stream_index = video_st->index; /* write the compressed frame in the media file */ int error_code = av_interleaved_write_frame(oc, &pkt); if (error_code < 0) { ZmqLogger::Instance()->AppendDebugMethod("FFmpegWriter::write_video_packet ERROR [" + (std::string) av_err2str(error_code) + "]", "error_code", error_code); return false; } } // Deallocate memory (if needed) if (video_outbuf) delete[] video_outbuf; // Deallocate packet AV_FREE_PACKET(&pkt); #if HAVE_HW_ACCEL if (hw_en_on && hw_en_supported) { if (hw_frame) { av_frame_free(&hw_frame); hw_frame = NULL; } } #endif // HAVE_HW_ACCEL } // Success return true; } // Output the ffmpeg info about this format, streams, and codecs (i.e. dump format) void FFmpegWriter::OutputStreamInfo() { // output debug info av_dump_format(oc, 0, path.c_str(), 1); } // Init a collection of software rescalers (thread safe) void FFmpegWriter::InitScalers(int source_width, int source_height) { int scale_mode = SWS_FAST_BILINEAR; if (openshot::Settings::Instance()->HIGH_QUALITY_SCALING) { scale_mode = SWS_BICUBIC; } // Init software rescalers vector (many of them, one for each thread) for (int x = 0; x < num_of_rescalers; x++) { // Init the software scaler from FFMpeg #if HAVE_HW_ACCEL if (hw_en_on && hw_en_supported) { img_convert_ctx = sws_getContext(source_width, source_height, PIX_FMT_RGBA, info.width, info.height, AV_PIX_FMT_NV12, scale_mode, NULL, NULL, NULL); } else #endif // HAVE_HW_ACCEL { img_convert_ctx = sws_getContext(source_width, source_height, PIX_FMT_RGBA, info.width, info.height, AV_GET_CODEC_PIXEL_FORMAT(video_st, video_st->codec), scale_mode, NULL, NULL, NULL); } // Add rescaler to vector image_rescalers.push_back(img_convert_ctx); } } // Set audio resample options void FFmpegWriter::ResampleAudio(int sample_rate, int channels) { original_sample_rate = sample_rate; original_channels = channels; } // Remove & deallocate all software scalers void FFmpegWriter::RemoveScalers() { // Close all rescalers for (int x = 0; x < num_of_rescalers; x++) sws_freeContext(image_rescalers[x]); // Clear vector image_rescalers.clear(); }