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);
94 #ifdef DCPOMATIC_VARIANT_SWAROOP
95 int r = avformat_alloc_output_context2 (&_format_context, av_guess_format("mov", 0, 0), 0, 0);
97 int r = avformat_alloc_output_context2 (&_format_context, 0, 0, _output.string().c_str());
99 if (!_format_context) {
100 throw runtime_error (String::compose("could not allocate FFmpeg format context (%1)", r));
103 _video_stream = avformat_new_stream (_format_context, _video_codec);
104 if (!_video_stream) {
105 throw runtime_error ("could not create FFmpeg output video stream");
108 _audio_stream = avformat_new_stream (_format_context, _audio_codec);
109 if (!_audio_stream) {
110 throw runtime_error ("could not create FFmpeg output audio stream");
113 _video_stream->id = _video_stream_index;
114 _video_stream->codec = _video_codec_context;
116 _audio_stream->id = _audio_stream_index;
117 _audio_stream->codec = _audio_codec_context;
119 if (avcodec_open2 (_video_codec_context, _video_codec, &_video_options) < 0) {
120 throw runtime_error ("could not open FFmpeg video codec");
123 r = avcodec_open2 (_audio_codec_context, _audio_codec, 0);
126 av_strerror (r, buffer, sizeof(buffer));
127 throw runtime_error (String::compose ("could not open FFmpeg audio codec (%1)", buffer));
130 if (avio_open_boost (&_format_context->pb, _output, AVIO_FLAG_WRITE) < 0) {
131 throw runtime_error ("could not open FFmpeg output file");
134 AVDictionary* options = 0;
136 #ifdef DCPOMATIC_VARIANT_SWAROOP
138 av_dict_set (&options, "encryption_key", key->hex().c_str(), 0);
139 /* XXX: is this OK? */
140 av_dict_set (&options, "encryption_kid", "00000000000000000000000000000000", 0);
141 av_dict_set (&options, "encryption_scheme", "cenc-aes-ctr", 0);
145 if (av_dict_set(&_format_context->metadata, SWAROOP_ID_TAG, id->c_str(), 0) < 0) {
146 throw runtime_error ("Could not write ID to output");
151 if (avformat_write_header (_format_context, &options) < 0) {
152 throw runtime_error ("could not write header to FFmpeg output file");
155 _pending_audio.reset (new AudioBuffers(channels, 0));
159 FFmpegFileEncoder::pixel_format (ExportFormat format)
162 case EXPORT_FORMAT_PRORES:
163 return AV_PIX_FMT_YUV422P10;
164 case EXPORT_FORMAT_H264_AAC:
165 case EXPORT_FORMAT_H264_PCM:
166 return AV_PIX_FMT_YUV420P;
168 DCPOMATIC_ASSERT (false);
171 return AV_PIX_FMT_YUV422P10;
175 FFmpegFileEncoder::setup_video ()
177 _video_codec = avcodec_find_encoder_by_name (_video_codec_name.c_str());
179 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _video_codec_name));
182 _video_codec_context = avcodec_alloc_context3 (_video_codec);
183 if (!_video_codec_context) {
184 throw runtime_error ("could not allocate FFmpeg video context");
187 avcodec_get_context_defaults3 (_video_codec_context, _video_codec);
189 /* Variable quantisation */
190 _video_codec_context->global_quality = 0;
191 _video_codec_context->width = _video_frame_size.width;
192 _video_codec_context->height = _video_frame_size.height;
193 _video_codec_context->time_base = (AVRational) { 1, _video_frame_rate };
194 _video_codec_context->pix_fmt = _pixel_format;
195 _video_codec_context->flags |= AV_CODEC_FLAG_QSCALE | AV_CODEC_FLAG_GLOBAL_HEADER;
199 FFmpegFileEncoder::setup_audio ()
201 _audio_codec = avcodec_find_encoder_by_name (_audio_codec_name.c_str());
203 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _audio_codec_name));
206 _audio_codec_context = avcodec_alloc_context3 (_audio_codec);
207 if (!_audio_codec_context) {
208 throw runtime_error ("could not allocate FFmpeg audio context");
211 avcodec_get_context_defaults3 (_audio_codec_context, _audio_codec);
213 /* XXX: configurable */
214 _audio_codec_context->bit_rate = _audio_channels * 128 * 1024;
215 _audio_codec_context->sample_fmt = _sample_format;
216 _audio_codec_context->sample_rate = _audio_frame_rate;
217 _audio_codec_context->channel_layout = av_get_default_channel_layout (_audio_channels);
218 _audio_codec_context->channels = _audio_channels;
222 FFmpegFileEncoder::flush ()
224 if (_pending_audio->frames() > 0) {
225 audio_frame (_pending_audio->frames ());
228 bool flushed_video = false;
229 bool flushed_audio = false;
231 while (!flushed_video || !flushed_audio) {
233 av_init_packet (&packet);
238 avcodec_encode_video2 (_video_codec_context, &packet, 0, &got_packet);
240 packet.stream_index = 0;
241 av_interleaved_write_frame (_format_context, &packet);
243 flushed_video = true;
245 av_packet_unref (&packet);
247 av_init_packet (&packet);
251 avcodec_encode_audio2 (_audio_codec_context, &packet, 0, &got_packet);
253 packet.stream_index = 0;
254 av_interleaved_write_frame (_format_context, &packet);
256 flushed_audio = true;
258 av_packet_unref (&packet);
261 av_write_trailer (_format_context);
263 avcodec_close (_video_codec_context);
264 avcodec_close (_audio_codec_context);
265 avio_close (_format_context->pb);
266 avformat_free_context (_format_context);
270 FFmpegFileEncoder::video (shared_ptr<PlayerVideo> video, DCPTime time)
272 shared_ptr<Image> image = video->image (
273 bind (&PlayerVideo::force, _1, _pixel_format),
278 AVFrame* frame = av_frame_alloc ();
279 DCPOMATIC_ASSERT (frame);
282 boost::mutex::scoped_lock lm (_pending_images_mutex);
283 _pending_images[image->data()[0]] = image;
286 for (int i = 0; i < 3; ++i) {
287 AVBufferRef* buffer = av_buffer_create(image->data()[i], image->stride()[i] * image->size().height, &buffer_free, this, 0);
288 frame->buf[i] = av_buffer_ref (buffer);
289 frame->data[i] = buffer->data;
290 frame->linesize[i] = image->stride()[i];
291 av_buffer_unref (&buffer);
294 frame->width = image->size().width;
295 frame->height = image->size().height;
296 frame->format = _pixel_format;
297 frame->pts = time.seconds() / av_q2d (_video_stream->time_base);
300 av_init_packet (&packet);
305 if (avcodec_encode_video2 (_video_codec_context, &packet, frame, &got_packet) < 0) {
306 throw EncodeError ("FFmpeg video encode failed");
309 if (got_packet && packet.size) {
310 packet.stream_index = _video_stream_index;
311 av_interleaved_write_frame (_format_context, &packet);
312 av_packet_unref (&packet);
315 av_frame_free (&frame);
319 /** Called when the player gives us some audio */
321 FFmpegFileEncoder::audio (shared_ptr<AudioBuffers> audio)
323 _pending_audio->append (audio);
325 int frame_size = _audio_codec_context->frame_size;
326 if (frame_size == 0) {
327 /* codec has AV_CODEC_CAP_VARIABLE_FRAME_SIZE */
328 frame_size = _audio_frame_rate / _video_frame_rate;
331 while (_pending_audio->frames() >= frame_size) {
332 audio_frame (frame_size);
337 FFmpegFileEncoder::audio_frame (int size)
339 DCPOMATIC_ASSERT (size);
341 AVFrame* frame = av_frame_alloc ();
342 DCPOMATIC_ASSERT (frame);
344 int const channels = _pending_audio->channels();
345 DCPOMATIC_ASSERT (channels);
347 int const buffer_size = av_samples_get_buffer_size (0, channels, size, _audio_codec_context->sample_fmt, 0);
348 DCPOMATIC_ASSERT (buffer_size >= 0);
350 void* samples = av_malloc (buffer_size);
351 DCPOMATIC_ASSERT (samples);
353 frame->nb_samples = size;
354 int r = avcodec_fill_audio_frame (frame, channels, _audio_codec_context->sample_fmt, (const uint8_t *) samples, buffer_size, 0);
355 DCPOMATIC_ASSERT (r >= 0);
357 float** p = _pending_audio->data ();
358 switch (_audio_codec_context->sample_fmt) {
359 case AV_SAMPLE_FMT_S16:
361 int16_t* q = reinterpret_cast<int16_t*> (samples);
362 for (int i = 0; i < size; ++i) {
363 for (int j = 0; j < channels; ++j) {
364 *q++ = p[j][i] * 32767;
369 case AV_SAMPLE_FMT_S32:
371 int32_t* q = reinterpret_cast<int32_t*> (samples);
372 for (int i = 0; i < size; ++i) {
373 for (int j = 0; j < channels; ++j) {
374 *q++ = p[j][i] * 2147483647;
379 case AV_SAMPLE_FMT_FLTP:
381 float* q = reinterpret_cast<float*> (samples);
382 for (int i = 0; i < channels; ++i) {
383 memcpy (q, p[i], sizeof(float) * size);
389 DCPOMATIC_ASSERT (false);
393 av_init_packet (&packet);
398 if (avcodec_encode_audio2 (_audio_codec_context, &packet, frame, &got_packet) < 0) {
399 throw EncodeError ("FFmpeg audio encode failed");
402 if (got_packet && packet.size) {
403 packet.stream_index = _audio_stream_index;
404 av_interleaved_write_frame (_format_context, &packet);
405 av_packet_unref (&packet);
409 av_frame_free (&frame);
411 _pending_audio->trim_start (size);
415 FFmpegFileEncoder::subtitle (PlayerText, DCPTimePeriod)
421 FFmpegFileEncoder::buffer_free (void* opaque, uint8_t* data)
423 reinterpret_cast<FFmpegFileEncoder*>(opaque)->buffer_free2(data);
427 FFmpegFileEncoder::buffer_free2 (uint8_t* data)
429 boost::mutex::scoped_lock lm (_pending_images_mutex);
430 if (_pending_images.find(data) != _pending_images.end()) {
431 _pending_images.erase (data);