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;
42 int FFmpegFileEncoder::_video_stream_index = 0;
43 int FFmpegFileEncoder::_audio_stream_index = 1;
45 FFmpegFileEncoder::FFmpegFileEncoder (
46 dcp::Size video_frame_size,
52 boost::filesystem::path output
55 , _audio_channels (channels)
57 , _video_frame_size (video_frame_size)
58 , _video_frame_rate (video_frame_rate)
59 , _audio_frame_rate (audio_frame_rate)
61 _pixel_format = pixel_format (format);
64 case EXPORT_FORMAT_PRORES:
65 _sample_format = AV_SAMPLE_FMT_S16;
66 _video_codec_name = "prores_ks";
67 _audio_codec_name = "pcm_s16le";
68 av_dict_set (&_video_options, "profile", "3", 0);
69 av_dict_set (&_video_options, "threads", "auto", 0);
71 case EXPORT_FORMAT_H264:
72 _sample_format = AV_SAMPLE_FMT_FLTP;
73 _video_codec_name = "libx264";
74 _audio_codec_name = "aac";
75 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
82 int r = avformat_alloc_output_context2 (&_format_context, 0, 0, _output.string().c_str());
83 if (!_format_context) {
84 throw runtime_error (String::compose("could not allocate FFmpeg format context (%1)", r));
87 _video_stream = avformat_new_stream (_format_context, _video_codec);
89 throw runtime_error ("could not create FFmpeg output video stream");
92 _audio_stream = avformat_new_stream (_format_context, _audio_codec);
94 throw runtime_error ("could not create FFmpeg output audio stream");
97 _video_stream->id = _video_stream_index;
98 _video_stream->codec = _video_codec_context;
100 _audio_stream->id = _audio_stream_index;
101 _audio_stream->codec = _audio_codec_context;
103 if (avcodec_open2 (_video_codec_context, _video_codec, &_video_options) < 0) {
104 throw runtime_error ("could not open FFmpeg video codec");
107 r = avcodec_open2 (_audio_codec_context, _audio_codec, 0);
110 av_strerror (r, buffer, sizeof(buffer));
111 throw runtime_error (String::compose ("could not open FFmpeg audio codec (%1)", buffer));
114 if (avio_open_boost (&_format_context->pb, _output, AVIO_FLAG_WRITE) < 0) {
115 throw runtime_error ("could not open FFmpeg output file");
118 if (avformat_write_header (_format_context, 0) < 0) {
119 throw runtime_error ("could not write header to FFmpeg output file");
122 _pending_audio.reset (new AudioBuffers(channels, 0));
126 FFmpegFileEncoder::~FFmpegFileEncoder ()
128 avcodec_close (_video_codec_context);
129 avcodec_close (_audio_codec_context);
130 avformat_close_input (&_format_context);
135 FFmpegFileEncoder::pixel_format (ExportFormat format)
138 case EXPORT_FORMAT_PRORES:
139 return AV_PIX_FMT_YUV422P10;
140 case EXPORT_FORMAT_H264:
141 return AV_PIX_FMT_YUV420P;
143 DCPOMATIC_ASSERT (false);
146 return AV_PIX_FMT_YUV422P10;
150 FFmpegFileEncoder::setup_video ()
152 _video_codec = avcodec_find_encoder_by_name (_video_codec_name.c_str());
154 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _video_codec_name));
157 _video_codec_context = avcodec_alloc_context3 (_video_codec);
158 if (!_video_codec_context) {
159 throw runtime_error ("could not allocate FFmpeg video context");
162 avcodec_get_context_defaults3 (_video_codec_context, _video_codec);
164 /* Variable quantisation */
165 _video_codec_context->global_quality = 0;
166 _video_codec_context->width = _video_frame_size.width;
167 _video_codec_context->height = _video_frame_size.height;
168 _video_codec_context->time_base = (AVRational) { 1, _video_frame_rate };
169 _video_codec_context->pix_fmt = _pixel_format;
170 _video_codec_context->flags |= AV_CODEC_FLAG_QSCALE | AV_CODEC_FLAG_GLOBAL_HEADER;
174 FFmpegFileEncoder::setup_audio ()
176 _audio_codec = avcodec_find_encoder_by_name (_audio_codec_name.c_str());
178 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _audio_codec_name));
181 _audio_codec_context = avcodec_alloc_context3 (_audio_codec);
182 if (!_audio_codec_context) {
183 throw runtime_error ("could not allocate FFmpeg audio context");
186 avcodec_get_context_defaults3 (_audio_codec_context, _audio_codec);
188 /* XXX: configurable */
189 _audio_codec_context->bit_rate = 256 * 1024;
190 _audio_codec_context->sample_fmt = _sample_format;
191 _audio_codec_context->sample_rate = _audio_frame_rate;
192 _audio_codec_context->channel_layout = av_get_default_channel_layout (_audio_channels);
193 _audio_codec_context->channels = _audio_channels;
197 FFmpegFileEncoder::flush ()
199 if (_pending_audio->frames() > 0) {
200 audio_frame (_pending_audio->frames ());
203 bool flushed_video = false;
204 bool flushed_audio = false;
206 while (!flushed_video || !flushed_audio) {
208 av_init_packet (&packet);
213 avcodec_encode_video2 (_video_codec_context, &packet, 0, &got_packet);
215 packet.stream_index = 0;
216 av_interleaved_write_frame (_format_context, &packet);
218 flushed_video = true;
220 av_packet_unref (&packet);
222 av_init_packet (&packet);
226 avcodec_encode_audio2 (_audio_codec_context, &packet, 0, &got_packet);
228 packet.stream_index = 0;
229 av_interleaved_write_frame (_format_context, &packet);
231 flushed_audio = true;
233 av_packet_unref (&packet);
236 av_write_trailer (_format_context);
240 FFmpegFileEncoder::video (shared_ptr<PlayerVideo> video, DCPTime time)
242 shared_ptr<Image> image = video->image (
243 bind (&PlayerVideo::force, _1, _pixel_format),
248 AVFrame* frame = av_frame_alloc ();
249 DCPOMATIC_ASSERT (frame);
252 boost::mutex::scoped_lock lm (_pending_images_mutex);
253 _pending_images[image->data()[0]] = image;
256 for (int i = 0; i < 3; ++i) {
257 AVBufferRef* buffer = av_buffer_create(image->data()[i], image->stride()[i] * image->size().height, &buffer_free, this, 0);
258 frame->buf[i] = av_buffer_ref (buffer);
259 frame->data[i] = buffer->data;
260 frame->linesize[i] = image->stride()[i];
261 av_buffer_unref (&buffer);
264 frame->width = image->size().width;
265 frame->height = image->size().height;
266 frame->format = _pixel_format;
267 DCPOMATIC_ASSERT (_video_stream->time_base.num == 1);
268 frame->pts = time.get() * _video_stream->time_base.den / DCPTime::HZ;
271 av_init_packet (&packet);
276 if (avcodec_encode_video2 (_video_codec_context, &packet, frame, &got_packet) < 0) {
277 throw EncodeError ("FFmpeg video encode failed");
280 if (got_packet && packet.size) {
281 packet.stream_index = _video_stream_index;
282 av_interleaved_write_frame (_format_context, &packet);
283 av_packet_unref (&packet);
286 av_frame_free (&frame);
290 /** Called when the player gives us some audio */
292 FFmpegFileEncoder::audio (shared_ptr<AudioBuffers> audio)
294 _pending_audio->append (audio);
296 int frame_size = _audio_codec_context->frame_size;
297 if (frame_size == 0) {
298 /* codec has AV_CODEC_CAP_VARIABLE_FRAME_SIZE */
299 frame_size = _audio_frame_rate / _video_frame_rate;
302 while (_pending_audio->frames() >= frame_size) {
303 audio_frame (frame_size);
308 FFmpegFileEncoder::audio_frame (int size)
310 DCPOMATIC_ASSERT (size);
312 AVFrame* frame = av_frame_alloc ();
313 DCPOMATIC_ASSERT (frame);
315 int const channels = _pending_audio->channels();
316 DCPOMATIC_ASSERT (channels);
318 int const buffer_size = av_samples_get_buffer_size (0, channels, size, _audio_codec_context->sample_fmt, 0);
319 DCPOMATIC_ASSERT (buffer_size >= 0);
321 void* samples = av_malloc (buffer_size);
322 DCPOMATIC_ASSERT (samples);
324 frame->nb_samples = size;
325 int r = avcodec_fill_audio_frame (frame, channels, _audio_codec_context->sample_fmt, (const uint8_t *) samples, buffer_size, 0);
326 DCPOMATIC_ASSERT (r >= 0);
328 float** p = _pending_audio->data ();
329 switch (_audio_codec_context->sample_fmt) {
330 case AV_SAMPLE_FMT_S16:
332 int16_t* q = reinterpret_cast<int16_t*> (samples);
333 for (int i = 0; i < size; ++i) {
334 for (int j = 0; j < channels; ++j) {
335 *q++ = p[j][i] * 32767;
340 case AV_SAMPLE_FMT_FLTP:
342 float* q = reinterpret_cast<float*> (samples);
343 for (int i = 0; i < channels; ++i) {
344 memcpy (q, p[i], sizeof(float) * size);
350 DCPOMATIC_ASSERT (false);
354 av_init_packet (&packet);
359 if (avcodec_encode_audio2 (_audio_codec_context, &packet, frame, &got_packet) < 0) {
360 throw EncodeError ("FFmpeg audio encode failed");
363 if (got_packet && packet.size) {
364 packet.stream_index = _audio_stream_index;
365 av_interleaved_write_frame (_format_context, &packet);
366 av_packet_unref (&packet);
370 av_frame_free (&frame);
372 _pending_audio->trim_start (size);
376 FFmpegFileEncoder::subtitle (PlayerText, DCPTimePeriod)
382 FFmpegFileEncoder::buffer_free (void* opaque, uint8_t* data)
384 reinterpret_cast<FFmpegFileEncoder*>(opaque)->buffer_free2(data);
388 FFmpegFileEncoder::buffer_free2 (uint8_t* data)
390 boost::mutex::scoped_lock lm (_pending_images_mutex);
391 if (_pending_images.find(data) != _pending_images.end()) {
392 _pending_images.erase (data);