2 Copyright (C) 2017-2021 Carl Hetherington <cth@carlh.net>
4 This file is part of DCP-o-matic.
6 DCP-o-matic is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
11 DCP-o-matic is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with DCP-o-matic. If not, see <http://www.gnu.org/licenses/>.
22 #include "compose.hpp"
24 #include "ffmpeg_encoder.h"
25 #include "ffmpeg_wrapper.h"
31 #include "player_video.h"
33 #include <libavutil/channel_layout.h>
41 using std::make_shared;
42 using std::shared_ptr;
45 using namespace dcpomatic;
46 #if BOOST_VERSION >= 106100
47 using namespace boost::placeholders;
51 int FFmpegFileEncoder::_video_stream_index = 0;
52 int FFmpegFileEncoder::_audio_stream_index_base = 1;
55 class ExportAudioStream
58 ExportAudioStream (string codec_name, int channels, int frame_rate, AVSampleFormat sample_format, AVFormatContext* format_context, int stream_index)
59 : _format_context (format_context)
60 , _stream_index (stream_index)
62 _codec = avcodec_find_encoder_by_name (codec_name.c_str());
64 throw EncodeError (String::compose("avcodec_find_encoder_by_name failed for %1", codec_name));
67 _codec_context = avcodec_alloc_context3 (_codec);
68 if (!_codec_context) {
69 throw std::bad_alloc ();
72 /* XXX: configurable */
73 _codec_context->bit_rate = channels * 128 * 1024;
74 _codec_context->sample_fmt = sample_format;
75 _codec_context->sample_rate = frame_rate;
76 av_channel_layout_default(&_codec_context->ch_layout, channels);
78 int r = avcodec_open2 (_codec_context, _codec, 0);
80 throw EncodeError (N_("avcodec_open2"), N_("ExportAudioStream::ExportAudioStream"), r);
83 _stream = avformat_new_stream (format_context, _codec);
85 throw EncodeError (N_("avformat_new_stream"), N_("ExportAudioStream::ExportAudioStream"));
88 _stream->id = stream_index;
89 _stream->disposition |= AV_DISPOSITION_DEFAULT;
90 r = avcodec_parameters_from_context (_stream->codecpar, _codec_context);
92 throw EncodeError (N_("avcodec_parameters_from_context"), N_("ExportAudioStream::ExportAudioStream"), r);
98 avcodec_close (_codec_context);
101 ExportAudioStream (ExportAudioStream const&) = delete;
102 ExportAudioStream& operator= (ExportAudioStream const&) = delete;
104 int frame_size () const {
105 return _codec_context->frame_size;
110 int r = avcodec_send_frame (_codec_context, nullptr);
111 if (r < 0 && r != AVERROR_EOF) {
112 /* We get EOF if we've already flushed the stream once */
113 throw EncodeError (N_("avcodec_send_frame"), N_("ExportAudioStream::flush"), r);
116 ffmpeg::Packet packet;
117 r = avcodec_receive_packet (_codec_context, packet.get());
118 if (r == AVERROR_EOF) {
121 throw EncodeError (N_("avcodec_receive_packet"), N_("ExportAudioStream::flush"), r);
124 packet->stream_index = _stream_index;
125 av_interleaved_write_frame (_format_context, packet.get());
129 void write (int size, int channel_offset, int channels, float* const* data, int64_t sample_offset)
131 DCPOMATIC_ASSERT (size);
133 auto frame = av_frame_alloc ();
134 DCPOMATIC_ASSERT (frame);
137 int const buffer_size = av_samples_get_buffer_size (&line_size, channels, size, _codec_context->sample_fmt, 0);
138 DCPOMATIC_ASSERT (buffer_size >= 0);
140 auto samples = av_malloc (buffer_size);
141 DCPOMATIC_ASSERT (samples);
143 frame->nb_samples = size;
144 frame->format = _codec_context->sample_fmt;
145 frame->ch_layout.nb_channels = channels;
146 int r = avcodec_fill_audio_frame (frame, channels, _codec_context->sample_fmt, (const uint8_t *) samples, buffer_size, 0);
147 DCPOMATIC_ASSERT (r >= 0);
149 switch (_codec_context->sample_fmt) {
150 case AV_SAMPLE_FMT_S16:
152 int16_t* q = reinterpret_cast<int16_t*> (samples);
153 for (int i = 0; i < size; ++i) {
154 for (int j = 0; j < channels; ++j) {
155 *q++ = data[j + channel_offset][i] * 32767;
160 case AV_SAMPLE_FMT_S32:
162 int32_t* q = reinterpret_cast<int32_t*> (samples);
163 for (int i = 0; i < size; ++i) {
164 for (int j = 0; j < channels; ++j) {
165 *q++ = data[j + channel_offset][i] * 2147483647;
170 case AV_SAMPLE_FMT_FLTP:
172 for (int i = 0; i < channels; ++i) {
173 memcpy (reinterpret_cast<float*>(static_cast<uint8_t*>(samples) + i * line_size), data[i + channel_offset], sizeof(float) * size);
178 DCPOMATIC_ASSERT (false);
181 DCPOMATIC_ASSERT (_codec_context->time_base.num == 1);
182 frame->pts = sample_offset * _codec_context->time_base.den / _codec_context->sample_rate;
184 r = avcodec_send_frame (_codec_context, frame);
186 av_frame_free (&frame);
188 throw EncodeError (N_("avcodec_send_frame"), N_("ExportAudioStream::write"), r);
191 ffmpeg::Packet packet;
192 r = avcodec_receive_packet (_codec_context, packet.get());
193 if (r < 0 && r != AVERROR(EAGAIN)) {
194 throw EncodeError (N_("avcodec_receive_packet"), N_("ExportAudioStream::write"), r);
196 packet->stream_index = _stream_index;
197 av_interleaved_write_frame (_format_context, packet.get());
202 AVFormatContext* _format_context;
203 AVCodec const * _codec;
204 AVCodecContext* _codec_context;
210 FFmpegFileEncoder::FFmpegFileEncoder (
211 dcp::Size video_frame_size,
212 int video_frame_rate,
213 int audio_frame_rate,
216 bool audio_stream_per_channel,
218 boost::filesystem::path output
220 : _audio_stream_per_channel (audio_stream_per_channel)
221 , _audio_channels (channels)
223 , _video_frame_size (video_frame_size)
224 , _video_frame_rate (video_frame_rate)
225 , _audio_frame_rate (audio_frame_rate)
227 _pixel_format = pixel_format (format);
230 case ExportFormat::PRORES_4444:
231 _sample_format = AV_SAMPLE_FMT_S32;
232 _video_codec_name = "prores_ks";
233 _audio_codec_name = "pcm_s24le";
234 av_dict_set(&_video_options, "profile", "4", 0);
235 av_dict_set(&_video_options, "threads", "auto", 0);
237 case ExportFormat::PRORES_HQ:
238 _sample_format = AV_SAMPLE_FMT_S32;
239 _video_codec_name = "prores_ks";
240 _audio_codec_name = "pcm_s24le";
241 av_dict_set (&_video_options, "profile", "3", 0);
242 av_dict_set (&_video_options, "threads", "auto", 0);
244 case ExportFormat::H264_AAC:
245 _sample_format = AV_SAMPLE_FMT_FLTP;
246 _video_codec_name = "libx264";
247 _audio_codec_name = "aac";
248 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
251 DCPOMATIC_ASSERT (false);
254 int r = avformat_alloc_output_context2 (&_format_context, 0, 0, _output.string().c_str());
255 if (!_format_context) {
256 throw EncodeError (N_("avformat_alloc_output_context2"), "FFmpegFileEncoder::FFmpegFileEncoder", r);
262 r = avio_open_boost (&_format_context->pb, _output, AVIO_FLAG_WRITE);
264 throw EncodeError (String::compose(_("Could not open output file %1 (%2)"), _output.string(), r));
267 AVDictionary* options = nullptr;
269 r = avformat_write_header (_format_context, &options);
271 throw EncodeError (N_("avformat_write_header"), N_("FFmpegFileEncoder::FFmpegFileEncoder"), r);
274 _pending_audio = make_shared<AudioBuffers>(channels, 0);
278 FFmpegFileEncoder::~FFmpegFileEncoder ()
280 _audio_streams.clear ();
281 avcodec_close (_video_codec_context);
282 avio_close (_format_context->pb);
283 _format_context->pb = nullptr;
284 avformat_free_context (_format_context);
289 FFmpegFileEncoder::pixel_format (ExportFormat format)
292 case ExportFormat::PRORES_4444:
293 return AV_PIX_FMT_YUV444P10;
294 case ExportFormat::PRORES_HQ:
295 return AV_PIX_FMT_YUV422P10;
296 case ExportFormat::H264_AAC:
297 return AV_PIX_FMT_YUV420P;
299 DCPOMATIC_ASSERT (false);
302 return AV_PIX_FMT_YUV422P10;
307 FFmpegFileEncoder::setup_video ()
309 _video_codec = avcodec_find_encoder_by_name (_video_codec_name.c_str());
311 throw EncodeError (String::compose("avcodec_find_encoder_by_name failed for %1", _video_codec_name));
314 _video_codec_context = avcodec_alloc_context3 (_video_codec);
315 if (!_video_codec_context) {
316 throw std::bad_alloc ();
319 /* Variable quantisation */
320 _video_codec_context->global_quality = 0;
321 _video_codec_context->width = _video_frame_size.width;
322 _video_codec_context->height = _video_frame_size.height;
323 _video_codec_context->time_base = (AVRational) { 1, _video_frame_rate };
324 _video_codec_context->pix_fmt = _pixel_format;
325 _video_codec_context->flags |= AV_CODEC_FLAG_QSCALE | AV_CODEC_FLAG_GLOBAL_HEADER;
327 if (avcodec_open2 (_video_codec_context, _video_codec, &_video_options) < 0) {
328 throw EncodeError(N_("avcodec_open2"), N_("FFmpegFileEncoder::setup_video"));
331 _video_stream = avformat_new_stream (_format_context, _video_codec);
332 if (!_video_stream) {
333 throw EncodeError (N_("avformat_new_stream"), N_("FFmpegFileEncoder::setup_video"));
336 _video_stream->id = _video_stream_index;
337 int r = avcodec_parameters_from_context (_video_stream->codecpar, _video_codec_context);
339 throw EncodeError (N_("avcodec_parameters_from_context"), N_("FFmpegFileEncoder::setup_video"), r);
345 FFmpegFileEncoder::setup_audio ()
347 int const streams = _audio_stream_per_channel ? _audio_channels : 1;
348 int const channels_per_stream = _audio_stream_per_channel ? 1 : _audio_channels;
350 for (int i = 0; i < streams; ++i) {
351 _audio_streams.push_back(
352 make_shared<ExportAudioStream>(
353 _audio_codec_name, channels_per_stream, _audio_frame_rate, _sample_format, _format_context, _audio_stream_index_base + i
361 FFmpegFileEncoder::flush ()
363 if (_pending_audio->frames() > 0) {
364 audio_frame (_pending_audio->frames ());
367 bool flushed_video = false;
368 bool flushed_audio = false;
370 while (!flushed_video || !flushed_audio) {
371 int r = avcodec_send_frame (_video_codec_context, nullptr);
372 if (r < 0 && r != AVERROR_EOF) {
373 /* We get EOF if we've already flushed the stream once */
374 throw EncodeError (N_("avcodec_send_frame"), N_("FFmpegFileEncoder::flush"), r);
377 ffmpeg::Packet packet;
378 r = avcodec_receive_packet (_video_codec_context, packet.get());
379 if (r == AVERROR_EOF) {
380 flushed_video = true;
382 throw EncodeError (N_("avcodec_receive_packet"), N_("FFmpegFileEncoder::flush"), r);
384 packet->stream_index = _video_stream_index;
385 packet->duration = _video_stream->time_base.den / _video_frame_rate;
386 av_interleaved_write_frame (_format_context, packet.get());
389 flushed_audio = true;
390 for (auto const& i: _audio_streams) {
392 flushed_audio = false;
397 auto const r = av_write_trailer(_format_context);
399 throw EncodeError(N_("av_write_trailer"), N_("FFmpegFileEncoder::flush"), r);
405 FFmpegFileEncoder::video (shared_ptr<PlayerVideo> video, DCPTime time)
407 /* All our output formats are video range at the moment */
408 auto image = video->image (
409 bind (&PlayerVideo::force, _pixel_format),
414 auto frame = av_frame_alloc ();
415 DCPOMATIC_ASSERT (frame);
417 for (int i = 0; i < 3; ++i) {
418 auto buffer = _pending_images.create_buffer(image, i);
419 frame->buf[i] = av_buffer_ref (buffer);
420 frame->data[i] = buffer->data;
421 frame->linesize[i] = image->stride()[i];
422 av_buffer_unref (&buffer);
425 frame->width = image->size().width;
426 frame->height = image->size().height;
427 frame->format = _pixel_format;
428 DCPOMATIC_ASSERT (_video_stream->time_base.num == 1);
429 frame->pts = time.get() * _video_stream->time_base.den / DCPTime::HZ;
431 int r = avcodec_send_frame (_video_codec_context, frame);
432 av_frame_free (&frame);
434 throw EncodeError (N_("avcodec_send_frame"), N_("FFmpegFileEncoder::video"), r);
437 ffmpeg::Packet packet;
438 r = avcodec_receive_packet (_video_codec_context, packet.get());
439 if (r < 0 && r != AVERROR(EAGAIN)) {
440 throw EncodeError (N_("avcodec_receive_packet"), N_("FFmpegFileEncoder::video"), r);
442 packet->stream_index = _video_stream_index;
443 packet->duration = _video_stream->time_base.den / _video_frame_rate;
444 av_interleaved_write_frame (_format_context, packet.get());
449 /** Called when the player gives us some audio */
451 FFmpegFileEncoder::audio (shared_ptr<AudioBuffers> audio)
453 _pending_audio->append (audio);
455 DCPOMATIC_ASSERT (!_audio_streams.empty());
456 int frame_size = _audio_streams[0]->frame_size();
457 if (frame_size == 0) {
458 /* codec has AV_CODEC_CAP_VARIABLE_FRAME_SIZE */
459 frame_size = _audio_frame_rate / _video_frame_rate;
462 while (_pending_audio->frames() >= frame_size) {
463 audio_frame (frame_size);
469 FFmpegFileEncoder::audio_frame (int size)
471 if (_audio_stream_per_channel) {
473 for (auto i: _audio_streams) {
474 i->write (size, offset, 1, _pending_audio->data(), _audio_frames);
478 DCPOMATIC_ASSERT (!_audio_streams.empty());
479 DCPOMATIC_ASSERT (_pending_audio->channels());
480 _audio_streams[0]->write (size, 0, _pending_audio->channels(), _pending_audio->data(), _audio_frames);
483 _pending_audio->trim_start (size);
484 _audio_frames += size;
489 FFmpegFileEncoder::subtitle (PlayerText, DCPTimePeriod)