2 Copyright (C) 2017-2018 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/>.
21 #include "ffmpeg_encoder.h"
25 #include "player_video.h"
29 #include "compose.hpp"
35 using std::runtime_error;
38 using boost::shared_ptr;
40 using boost::weak_ptr;
41 using boost::optional;
42 using namespace dcpomatic;
44 int FFmpegFileEncoder::_video_stream_index = 0;
45 int FFmpegFileEncoder::_audio_stream_index = 1;
47 FFmpegFileEncoder::FFmpegFileEncoder (
48 dcp::Size video_frame_size,
54 boost::filesystem::path output
55 #ifdef DCPOMATIC_VARIANT_SWAROOP
56 , optional<dcp::Key> key
61 , _audio_channels (channels)
63 , _video_frame_size (video_frame_size)
64 , _video_frame_rate (video_frame_rate)
65 , _audio_frame_rate (audio_frame_rate)
67 _pixel_format = pixel_format (format);
70 case EXPORT_FORMAT_PRORES:
71 _sample_format = AV_SAMPLE_FMT_S16;
72 _video_codec_name = "prores_ks";
73 _audio_codec_name = "pcm_s16le";
74 av_dict_set (&_video_options, "profile", "3", 0);
75 av_dict_set (&_video_options, "threads", "auto", 0);
77 case EXPORT_FORMAT_H264_AAC:
78 _sample_format = AV_SAMPLE_FMT_FLTP;
79 _video_codec_name = "libx264";
80 _audio_codec_name = "aac";
81 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
83 case EXPORT_FORMAT_H264_PCM:
84 _sample_format = AV_SAMPLE_FMT_S32;
85 _video_codec_name = "libx264";
86 _audio_codec_name = "pcm_s24le";
87 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
90 DCPOMATIC_ASSERT (false);
96 #ifdef DCPOMATIC_VARIANT_SWAROOP
97 int r = avformat_alloc_output_context2 (&_format_context, av_guess_format("mov", 0, 0), 0, 0);
99 int r = avformat_alloc_output_context2 (&_format_context, 0, 0, _output.string().c_str());
101 if (!_format_context) {
102 throw runtime_error (String::compose("could not allocate FFmpeg format context (%1)", r));
105 _video_stream = avformat_new_stream (_format_context, _video_codec);
106 if (!_video_stream) {
107 throw runtime_error ("could not create FFmpeg output video stream");
110 _audio_stream = avformat_new_stream (_format_context, _audio_codec);
111 if (!_audio_stream) {
112 throw runtime_error ("could not create FFmpeg output audio stream");
115 _video_stream->id = _video_stream_index;
116 _video_stream->codec = _video_codec_context;
118 _audio_stream->id = _audio_stream_index;
119 _audio_stream->codec = _audio_codec_context;
121 if (avcodec_open2 (_video_codec_context, _video_codec, &_video_options) < 0) {
122 throw runtime_error ("could not open FFmpeg video codec");
125 r = avcodec_open2 (_audio_codec_context, _audio_codec, 0);
128 av_strerror (r, buffer, sizeof(buffer));
129 throw runtime_error (String::compose ("could not open FFmpeg audio codec (%1)", buffer));
132 if (avio_open_boost (&_format_context->pb, _output, AVIO_FLAG_WRITE) < 0) {
133 throw runtime_error ("could not open FFmpeg output file");
136 AVDictionary* options = 0;
138 #ifdef DCPOMATIC_VARIANT_SWAROOP
140 av_dict_set (&options, "encryption_key", key->hex().c_str(), 0);
141 /* XXX: is this OK? */
142 av_dict_set (&options, "encryption_kid", "00000000000000000000000000000000", 0);
143 av_dict_set (&options, "encryption_scheme", "cenc-aes-ctr", 0);
147 if (av_dict_set(&_format_context->metadata, SWAROOP_ID_TAG, id->c_str(), 0) < 0) {
148 throw runtime_error ("Could not write ID to output");
153 if (avformat_write_header (_format_context, &options) < 0) {
154 throw runtime_error ("could not write header to FFmpeg output file");
157 _pending_audio.reset (new AudioBuffers(channels, 0));
161 FFmpegFileEncoder::pixel_format (ExportFormat format)
164 case EXPORT_FORMAT_PRORES:
165 return AV_PIX_FMT_YUV422P10;
166 case EXPORT_FORMAT_H264_AAC:
167 case EXPORT_FORMAT_H264_PCM:
168 return AV_PIX_FMT_YUV420P;
170 DCPOMATIC_ASSERT (false);
173 return AV_PIX_FMT_YUV422P10;
177 FFmpegFileEncoder::setup_video ()
179 _video_codec = avcodec_find_encoder_by_name (_video_codec_name.c_str());
181 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _video_codec_name));
184 _video_codec_context = avcodec_alloc_context3 (_video_codec);
185 if (!_video_codec_context) {
186 throw runtime_error ("could not allocate FFmpeg video context");
189 avcodec_get_context_defaults3 (_video_codec_context, _video_codec);
191 /* Variable quantisation */
192 _video_codec_context->global_quality = 0;
193 _video_codec_context->width = _video_frame_size.width;
194 _video_codec_context->height = _video_frame_size.height;
195 _video_codec_context->time_base = (AVRational) { 1, _video_frame_rate };
196 _video_codec_context->pix_fmt = _pixel_format;
197 _video_codec_context->flags |= AV_CODEC_FLAG_QSCALE | AV_CODEC_FLAG_GLOBAL_HEADER;
201 FFmpegFileEncoder::setup_audio ()
203 _audio_codec = avcodec_find_encoder_by_name (_audio_codec_name.c_str());
205 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _audio_codec_name));
208 _audio_codec_context = avcodec_alloc_context3 (_audio_codec);
209 if (!_audio_codec_context) {
210 throw runtime_error ("could not allocate FFmpeg audio context");
213 avcodec_get_context_defaults3 (_audio_codec_context, _audio_codec);
215 /* XXX: configurable */
216 _audio_codec_context->bit_rate = _audio_channels * 128 * 1024;
217 _audio_codec_context->sample_fmt = _sample_format;
218 _audio_codec_context->sample_rate = _audio_frame_rate;
219 _audio_codec_context->channel_layout = av_get_default_channel_layout (_audio_channels);
220 _audio_codec_context->channels = _audio_channels;
224 FFmpegFileEncoder::flush ()
226 if (_pending_audio->frames() > 0) {
227 audio_frame (_pending_audio->frames ());
230 bool flushed_video = false;
231 bool flushed_audio = false;
233 while (!flushed_video || !flushed_audio) {
235 av_init_packet (&packet);
240 avcodec_encode_video2 (_video_codec_context, &packet, 0, &got_packet);
242 packet.stream_index = 0;
243 av_interleaved_write_frame (_format_context, &packet);
245 flushed_video = true;
247 av_packet_unref (&packet);
249 av_init_packet (&packet);
253 avcodec_encode_audio2 (_audio_codec_context, &packet, 0, &got_packet);
255 packet.stream_index = 0;
256 av_interleaved_write_frame (_format_context, &packet);
258 flushed_audio = true;
260 av_packet_unref (&packet);
263 av_write_trailer (_format_context);
265 avcodec_close (_video_codec_context);
266 avcodec_close (_audio_codec_context);
267 avio_close (_format_context->pb);
268 avformat_free_context (_format_context);
272 FFmpegFileEncoder::video (shared_ptr<PlayerVideo> video, DCPTime time)
274 shared_ptr<Image> image = video->image (
275 bind (&PlayerVideo::force, _1, _pixel_format),
280 AVFrame* frame = av_frame_alloc ();
281 DCPOMATIC_ASSERT (frame);
284 boost::mutex::scoped_lock lm (_pending_images_mutex);
285 _pending_images[image->data()[0]] = image;
288 for (int i = 0; i < 3; ++i) {
289 AVBufferRef* buffer = av_buffer_create(image->data()[i], image->stride()[i] * image->size().height, &buffer_free, this, 0);
290 frame->buf[i] = av_buffer_ref (buffer);
291 frame->data[i] = buffer->data;
292 frame->linesize[i] = image->stride()[i];
293 av_buffer_unref (&buffer);
296 frame->width = image->size().width;
297 frame->height = image->size().height;
298 frame->format = _pixel_format;
299 frame->pts = time.seconds() / av_q2d (_video_stream->time_base);
302 av_init_packet (&packet);
307 if (avcodec_encode_video2 (_video_codec_context, &packet, frame, &got_packet) < 0) {
308 throw EncodeError ("FFmpeg video encode failed");
311 if (got_packet && packet.size) {
312 packet.stream_index = _video_stream_index;
313 av_interleaved_write_frame (_format_context, &packet);
314 av_packet_unref (&packet);
317 av_frame_free (&frame);
321 /** Called when the player gives us some audio */
323 FFmpegFileEncoder::audio (shared_ptr<AudioBuffers> audio)
325 _pending_audio->append (audio);
327 int frame_size = _audio_codec_context->frame_size;
328 if (frame_size == 0) {
329 /* codec has AV_CODEC_CAP_VARIABLE_FRAME_SIZE */
330 frame_size = _audio_frame_rate / _video_frame_rate;
333 while (_pending_audio->frames() >= frame_size) {
334 audio_frame (frame_size);
339 FFmpegFileEncoder::audio_frame (int size)
341 DCPOMATIC_ASSERT (size);
343 AVFrame* frame = av_frame_alloc ();
344 DCPOMATIC_ASSERT (frame);
346 int const channels = _pending_audio->channels();
347 DCPOMATIC_ASSERT (channels);
349 int const buffer_size = av_samples_get_buffer_size (0, channels, size, _audio_codec_context->sample_fmt, 0);
350 DCPOMATIC_ASSERT (buffer_size >= 0);
352 void* samples = av_malloc (buffer_size);
353 DCPOMATIC_ASSERT (samples);
355 frame->nb_samples = size;
356 int r = avcodec_fill_audio_frame (frame, channels, _audio_codec_context->sample_fmt, (const uint8_t *) samples, buffer_size, 0);
357 DCPOMATIC_ASSERT (r >= 0);
359 float** p = _pending_audio->data ();
360 switch (_audio_codec_context->sample_fmt) {
361 case AV_SAMPLE_FMT_S16:
363 int16_t* q = reinterpret_cast<int16_t*> (samples);
364 for (int i = 0; i < size; ++i) {
365 for (int j = 0; j < channels; ++j) {
366 *q++ = p[j][i] * 32767;
371 case AV_SAMPLE_FMT_S32:
373 int32_t* q = reinterpret_cast<int32_t*> (samples);
374 for (int i = 0; i < size; ++i) {
375 for (int j = 0; j < channels; ++j) {
376 *q++ = p[j][i] * 2147483647;
381 case AV_SAMPLE_FMT_FLTP:
383 float* q = reinterpret_cast<float*> (samples);
384 for (int i = 0; i < channels; ++i) {
385 memcpy (q, p[i], sizeof(float) * size);
391 DCPOMATIC_ASSERT (false);
395 av_init_packet (&packet);
400 if (avcodec_encode_audio2 (_audio_codec_context, &packet, frame, &got_packet) < 0) {
401 throw EncodeError ("FFmpeg audio encode failed");
404 if (got_packet && packet.size) {
405 packet.stream_index = _audio_stream_index;
406 av_interleaved_write_frame (_format_context, &packet);
407 av_packet_unref (&packet);
411 av_frame_free (&frame);
413 _pending_audio->trim_start (size);
417 FFmpegFileEncoder::subtitle (PlayerText, DCPTimePeriod)
423 FFmpegFileEncoder::buffer_free (void* opaque, uint8_t* data)
425 reinterpret_cast<FFmpegFileEncoder*>(opaque)->buffer_free2(data);
429 FFmpegFileEncoder::buffer_free2 (uint8_t* data)
431 boost::mutex::scoped_lock lm (_pending_images_mutex);
432 if (_pending_images.find(data) != _pending_images.end()) {
433 _pending_images.erase (data);