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,
53 boost::filesystem::path output
56 , _audio_channels (channels)
58 , _video_frame_size (video_frame_size)
59 , _video_frame_rate (video_frame_rate)
60 , _audio_frame_rate (audio_frame_rate)
63 _pixel_format = pixel_format (format);
66 case EXPORT_FORMAT_PRORES:
67 _sample_format = AV_SAMPLE_FMT_S16;
68 _video_codec_name = "prores_ks";
69 _audio_codec_name = "pcm_s16le";
70 av_dict_set (&_video_options, "profile", "3", 0);
71 av_dict_set (&_video_options, "threads", "auto", 0);
73 case EXPORT_FORMAT_H264:
74 _sample_format = AV_SAMPLE_FMT_FLTP;
75 _video_codec_name = "libx264";
76 _audio_codec_name = "aac";
77 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
84 avformat_alloc_output_context2 (&_format_context, 0, 0, _output.string().c_str());
85 if (!_format_context) {
86 throw runtime_error ("could not allocate FFmpeg format context");
89 _video_stream = avformat_new_stream (_format_context, _video_codec);
91 throw runtime_error ("could not create FFmpeg output video stream");
94 _audio_stream = avformat_new_stream (_format_context, _audio_codec);
96 throw runtime_error ("could not create FFmpeg output audio stream");
99 _video_stream->id = _video_stream_index;
100 _video_stream->codec = _video_codec_context;
102 _audio_stream->id = _audio_stream_index;
103 _audio_stream->codec = _audio_codec_context;
105 if (avcodec_open2 (_video_codec_context, _video_codec, &_video_options) < 0) {
106 throw runtime_error ("could not open FFmpeg video codec");
109 int r = avcodec_open2 (_audio_codec_context, _audio_codec, 0);
112 av_strerror (r, buffer, sizeof(buffer));
113 throw runtime_error (String::compose ("could not open FFmpeg audio codec (%1)", buffer));
116 if (avio_open_boost (&_format_context->pb, _output, AVIO_FLAG_WRITE) < 0) {
117 throw runtime_error ("could not open FFmpeg output file");
120 if (avformat_write_header (_format_context, 0) < 0) {
121 throw runtime_error ("could not write header to FFmpeg output file");
124 _pending_audio.reset (new AudioBuffers(channels, 0));
128 FFmpegFileEncoder::pixel_format (ExportFormat format)
131 case EXPORT_FORMAT_PRORES:
132 return AV_PIX_FMT_YUV422P10;
133 case EXPORT_FORMAT_H264:
134 return AV_PIX_FMT_YUV420P;
136 DCPOMATIC_ASSERT (false);
139 return AV_PIX_FMT_YUV422P10;
143 FFmpegFileEncoder::setup_video ()
145 _video_codec = avcodec_find_encoder_by_name (_video_codec_name.c_str());
147 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _video_codec_name));
150 _video_codec_context = avcodec_alloc_context3 (_video_codec);
151 if (!_video_codec_context) {
152 throw runtime_error ("could not allocate FFmpeg video context");
155 avcodec_get_context_defaults3 (_video_codec_context, _video_codec);
157 /* Variable quantisation */
158 _video_codec_context->global_quality = 0;
159 _video_codec_context->width = _video_frame_size.width;
160 _video_codec_context->height = _video_frame_size.height;
161 _video_codec_context->time_base = (AVRational) { 1, _video_frame_rate };
162 _video_codec_context->pix_fmt = _pixel_format;
163 _video_codec_context->flags |= AV_CODEC_FLAG_QSCALE | AV_CODEC_FLAG_GLOBAL_HEADER;
167 FFmpegFileEncoder::setup_audio ()
169 _audio_codec = avcodec_find_encoder_by_name (_audio_codec_name.c_str());
171 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _audio_codec_name));
174 _audio_codec_context = avcodec_alloc_context3 (_audio_codec);
175 if (!_audio_codec_context) {
176 throw runtime_error ("could not allocate FFmpeg audio context");
179 avcodec_get_context_defaults3 (_audio_codec_context, _audio_codec);
181 /* XXX: configurable */
182 _audio_codec_context->bit_rate = 256 * 1024;
183 _audio_codec_context->sample_fmt = _sample_format;
184 _audio_codec_context->sample_rate = _audio_frame_rate;
185 _audio_codec_context->channel_layout = av_get_default_channel_layout (_audio_channels);
186 _audio_codec_context->channels = _audio_channels;
190 FFmpegFileEncoder::flush ()
192 if (_pending_audio->frames() > 0) {
193 audio_frame (_pending_audio->frames ());
196 bool flushed_video = false;
197 bool flushed_audio = false;
199 while (!flushed_video || !flushed_audio) {
201 av_init_packet (&packet);
206 avcodec_encode_video2 (_video_codec_context, &packet, 0, &got_packet);
208 packet.stream_index = 0;
209 av_interleaved_write_frame (_format_context, &packet);
211 flushed_video = true;
213 av_packet_unref (&packet);
215 av_init_packet (&packet);
219 avcodec_encode_audio2 (_audio_codec_context, &packet, 0, &got_packet);
221 packet.stream_index = 0;
222 av_interleaved_write_frame (_format_context, &packet);
224 flushed_audio = true;
226 av_packet_unref (&packet);
229 av_write_trailer (_format_context);
231 avcodec_close (_video_codec_context);
232 avcodec_close (_audio_codec_context);
233 avio_close (_format_context->pb);
234 avformat_free_context (_format_context);
238 FFmpegFileEncoder::video (shared_ptr<PlayerVideo> video, DCPTime time)
240 shared_ptr<Image> image = video->image (
241 bind (&PlayerVideo::force, _1, _pixel_format),
246 AVFrame* frame = av_frame_alloc ();
247 DCPOMATIC_ASSERT (frame);
249 _pending_images[image->data()[0]] = image;
250 for (int i = 0; i < 3; ++i) {
251 AVBufferRef* buffer = av_buffer_create(image->data()[i], image->stride()[i] * image->size().height, &buffer_free, this, 0);
252 frame->buf[i] = av_buffer_ref (buffer);
253 frame->data[i] = buffer->data;
254 frame->linesize[i] = image->stride()[i];
255 av_buffer_unref (&buffer);
258 frame->width = image->size().width;
259 frame->height = image->size().height;
260 frame->format = _pixel_format;
261 frame->pts = time.seconds() / av_q2d (_video_stream->time_base);
264 av_init_packet (&packet);
269 if (avcodec_encode_video2 (_video_codec_context, &packet, frame, &got_packet) < 0) {
270 throw EncodeError ("FFmpeg video encode failed");
273 if (got_packet && packet.size) {
274 packet.stream_index = _video_stream_index;
275 av_interleaved_write_frame (_format_context, &packet);
276 av_packet_unref (&packet);
279 av_frame_free (&frame);
283 /** Called when the player gives us some audio */
285 FFmpegFileEncoder::audio (shared_ptr<AudioBuffers> audio)
287 _pending_audio->append (audio);
289 int frame_size = _audio_codec_context->frame_size;
290 if (frame_size == 0) {
291 /* codec has AV_CODEC_CAP_VARIABLE_FRAME_SIZE */
292 frame_size = _audio_frame_rate / _video_frame_rate;
295 while (_pending_audio->frames() >= frame_size) {
296 audio_frame (frame_size);
301 FFmpegFileEncoder::audio_frame (int size)
303 DCPOMATIC_ASSERT (size);
305 AVFrame* frame = av_frame_alloc ();
306 DCPOMATIC_ASSERT (frame);
308 int const channels = _pending_audio->channels();
309 DCPOMATIC_ASSERT (channels);
311 int const buffer_size = av_samples_get_buffer_size (0, channels, size, _audio_codec_context->sample_fmt, 0);
312 DCPOMATIC_ASSERT (buffer_size >= 0);
314 void* samples = av_malloc (buffer_size);
315 DCPOMATIC_ASSERT (samples);
317 frame->nb_samples = size;
318 int r = avcodec_fill_audio_frame (frame, channels, _audio_codec_context->sample_fmt, (const uint8_t *) samples, buffer_size, 0);
319 DCPOMATIC_ASSERT (r >= 0);
321 float** p = _pending_audio->data ();
322 switch (_audio_codec_context->sample_fmt) {
323 case AV_SAMPLE_FMT_S16:
325 int16_t* q = reinterpret_cast<int16_t*> (samples);
326 for (int i = 0; i < size; ++i) {
327 for (int j = 0; j < channels; ++j) {
328 *q++ = p[j][i] * 32767;
333 case AV_SAMPLE_FMT_FLTP:
335 float* q = reinterpret_cast<float*> (samples);
336 for (int i = 0; i < channels; ++i) {
337 memcpy (q, p[i], sizeof(float) * size);
343 DCPOMATIC_ASSERT (false);
347 av_init_packet (&packet);
352 if (avcodec_encode_audio2 (_audio_codec_context, &packet, frame, &got_packet) < 0) {
353 throw EncodeError ("FFmpeg audio encode failed");
356 if (got_packet && packet.size) {
357 packet.stream_index = _audio_stream_index;
358 av_interleaved_write_frame (_format_context, &packet);
359 av_packet_unref (&packet);
363 av_frame_free (&frame);
365 _pending_audio->trim_start (size);
369 FFmpegFileEncoder::subtitle (PlayerText, DCPTimePeriod)
375 FFmpegFileEncoder::buffer_free (void* opaque, uint8_t* data)
377 reinterpret_cast<FFmpegFileEncoder*>(opaque)->buffer_free2(data);
381 FFmpegFileEncoder::buffer_free2 (uint8_t* data)
383 _pending_images.erase (data);