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"
30 #include "compose.hpp"
36 using std::runtime_error;
39 using boost::shared_ptr;
41 using boost::weak_ptr;
43 int FFmpegFileEncoder::_video_stream_index = 0;
44 int FFmpegFileEncoder::_audio_stream_index = 1;
47 force_pixel_format (AVPixelFormat, AVPixelFormat out)
52 FFmpegFileEncoder::FFmpegFileEncoder (
53 dcp::Size video_frame_size,
60 boost::filesystem::path output
63 , _audio_channels (channels)
65 , _video_frame_size (video_frame_size)
66 , _video_frame_rate (video_frame_rate)
67 , _audio_frame_rate (audio_frame_rate)
71 case EXPORT_FORMAT_PRORES:
72 _pixel_format = AV_PIX_FMT_YUV422P10;
73 _sample_format = AV_SAMPLE_FMT_S16;
74 _video_codec_name = "prores_ks";
75 _audio_codec_name = "pcm_s16le";
76 av_dict_set (&_video_options, "profile", "3", 0);
77 av_dict_set (&_video_options, "threads", "auto", 0);
79 case EXPORT_FORMAT_H264:
80 _pixel_format = AV_PIX_FMT_YUV420P;
81 _sample_format = AV_SAMPLE_FMT_FLTP;
82 _video_codec_name = "libx264";
83 _audio_codec_name = "aac";
84 av_dict_set_int (&_video_options, "crf", x264_crf, 0);
91 avformat_alloc_output_context2 (&_format_context, 0, 0, _output.string().c_str());
92 if (!_format_context) {
93 throw runtime_error ("could not allocate FFmpeg format context");
96 _video_stream = avformat_new_stream (_format_context, _video_codec);
98 throw runtime_error ("could not create FFmpeg output video stream");
101 _audio_stream = avformat_new_stream (_format_context, _audio_codec);
102 if (!_audio_stream) {
103 throw runtime_error ("could not create FFmpeg output audio stream");
106 _video_stream->id = _video_stream_index;
107 _video_stream->codec = _video_codec_context;
109 _audio_stream->id = _audio_stream_index;
110 _audio_stream->codec = _audio_codec_context;
112 if (avcodec_open2 (_video_codec_context, _video_codec, &_video_options) < 0) {
113 throw runtime_error ("could not open FFmpeg video codec");
116 int r = avcodec_open2 (_audio_codec_context, _audio_codec, 0);
119 av_strerror (r, buffer, sizeof(buffer));
120 throw runtime_error (String::compose ("could not open FFmpeg audio codec (%1)", buffer));
123 if (avio_open_boost (&_format_context->pb, _output, AVIO_FLAG_WRITE) < 0) {
124 throw runtime_error ("could not open FFmpeg output file");
127 if (avformat_write_header (_format_context, 0) < 0) {
128 throw runtime_error ("could not write header to FFmpeg output file");
131 _pending_audio.reset (new AudioBuffers(channels, 0));
135 FFmpegFileEncoder::setup_video ()
137 _video_codec = avcodec_find_encoder_by_name (_video_codec_name.c_str());
139 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _video_codec_name));
142 _video_codec_context = avcodec_alloc_context3 (_video_codec);
143 if (!_video_codec_context) {
144 throw runtime_error ("could not allocate FFmpeg video context");
147 avcodec_get_context_defaults3 (_video_codec_context, _video_codec);
149 /* Variable quantisation */
150 _video_codec_context->global_quality = 0;
151 _video_codec_context->width = _video_frame_size.width;
152 _video_codec_context->height = _video_frame_size.height;
153 _video_codec_context->time_base = (AVRational) { 1, _video_frame_rate };
154 _video_codec_context->pix_fmt = _pixel_format;
155 _video_codec_context->flags |= AV_CODEC_FLAG_QSCALE | AV_CODEC_FLAG_GLOBAL_HEADER;
159 FFmpegFileEncoder::setup_audio ()
161 _audio_codec = avcodec_find_encoder_by_name (_audio_codec_name.c_str());
163 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _audio_codec_name));
166 _audio_codec_context = avcodec_alloc_context3 (_audio_codec);
167 if (!_audio_codec_context) {
168 throw runtime_error ("could not allocate FFmpeg audio context");
171 avcodec_get_context_defaults3 (_audio_codec_context, _audio_codec);
173 /* XXX: configurable */
174 _audio_codec_context->bit_rate = 256 * 1024;
175 _audio_codec_context->sample_fmt = _sample_format;
176 _audio_codec_context->sample_rate = _audio_frame_rate;
177 _audio_codec_context->channel_layout = av_get_default_channel_layout (_audio_channels);
178 _audio_codec_context->channels = _audio_channels;
182 FFmpegFileEncoder::flush ()
184 if (_pending_audio->frames() > 0) {
185 audio_frame (_pending_audio->frames ());
188 bool flushed_video = false;
189 bool flushed_audio = false;
191 while (!flushed_video || !flushed_audio) {
193 av_init_packet (&packet);
198 avcodec_encode_video2 (_video_codec_context, &packet, 0, &got_packet);
200 packet.stream_index = 0;
201 av_interleaved_write_frame (_format_context, &packet);
203 flushed_video = true;
205 av_packet_unref (&packet);
207 av_init_packet (&packet);
211 avcodec_encode_audio2 (_audio_codec_context, &packet, 0, &got_packet);
213 packet.stream_index = 0;
214 av_interleaved_write_frame (_format_context, &packet);
216 flushed_audio = true;
218 av_packet_unref (&packet);
221 av_write_trailer (_format_context);
223 avcodec_close (_video_codec_context);
224 avcodec_close (_audio_codec_context);
225 avio_close (_format_context->pb);
226 avformat_free_context (_format_context);
230 FFmpegFileEncoder::video (shared_ptr<PlayerVideo> video, DCPTime time)
232 shared_ptr<Image> image = video->image (
233 bind (&Log::dcp_log, _log.get(), _1, _2),
234 bind (&force_pixel_format, _1, _pixel_format),
239 AVFrame* frame = av_frame_alloc ();
240 DCPOMATIC_ASSERT (frame);
242 _pending_images[image->data()[0]] = image;
243 for (int i = 0; i < 3; ++i) {
244 AVBufferRef* buffer = av_buffer_create(image->data()[i], image->stride()[i] * image->size().height, &buffer_free, this, 0);
245 frame->buf[i] = av_buffer_ref (buffer);
246 frame->data[i] = buffer->data;
247 frame->linesize[i] = image->stride()[i];
248 av_buffer_unref (&buffer);
251 frame->width = image->size().width;
252 frame->height = image->size().height;
253 frame->format = _pixel_format;
254 frame->pts = time.seconds() / av_q2d (_video_stream->time_base);
257 av_init_packet (&packet);
262 if (avcodec_encode_video2 (_video_codec_context, &packet, frame, &got_packet) < 0) {
263 throw EncodeError ("FFmpeg video encode failed");
266 if (got_packet && packet.size) {
267 packet.stream_index = _video_stream_index;
268 av_interleaved_write_frame (_format_context, &packet);
269 av_packet_unref (&packet);
272 av_frame_free (&frame);
276 /** Called when the player gives us some audio */
278 FFmpegFileEncoder::audio (shared_ptr<AudioBuffers> audio)
280 _pending_audio->append (audio);
282 int frame_size = _audio_codec_context->frame_size;
283 if (frame_size == 0) {
284 /* codec has AV_CODEC_CAP_VARIABLE_FRAME_SIZE */
285 frame_size = _audio_frame_rate / _video_frame_rate;
288 while (_pending_audio->frames() >= frame_size) {
289 audio_frame (frame_size);
294 FFmpegFileEncoder::audio_frame (int size)
296 DCPOMATIC_ASSERT (size);
298 AVFrame* frame = av_frame_alloc ();
299 DCPOMATIC_ASSERT (frame);
301 int const channels = _pending_audio->channels();
302 DCPOMATIC_ASSERT (channels);
304 int const buffer_size = av_samples_get_buffer_size (0, channels, size, _audio_codec_context->sample_fmt, 0);
305 DCPOMATIC_ASSERT (buffer_size >= 0);
307 void* samples = av_malloc (buffer_size);
308 DCPOMATIC_ASSERT (samples);
310 frame->nb_samples = size;
311 int r = avcodec_fill_audio_frame (frame, channels, _audio_codec_context->sample_fmt, (const uint8_t *) samples, buffer_size, 0);
312 DCPOMATIC_ASSERT (r >= 0);
314 float** p = _pending_audio->data ();
315 switch (_audio_codec_context->sample_fmt) {
316 case AV_SAMPLE_FMT_S16:
318 int16_t* q = reinterpret_cast<int16_t*> (samples);
319 for (int i = 0; i < size; ++i) {
320 for (int j = 0; j < channels; ++j) {
321 *q++ = p[j][i] * 32767;
326 case AV_SAMPLE_FMT_FLTP:
328 float* q = reinterpret_cast<float*> (samples);
329 for (int i = 0; i < channels; ++i) {
330 memcpy (q, p[i], sizeof(float) * size);
336 DCPOMATIC_ASSERT (false);
340 av_init_packet (&packet);
345 if (avcodec_encode_audio2 (_audio_codec_context, &packet, frame, &got_packet) < 0) {
346 throw EncodeError ("FFmpeg audio encode failed");
349 if (got_packet && packet.size) {
350 packet.stream_index = _audio_stream_index;
351 av_interleaved_write_frame (_format_context, &packet);
352 av_packet_unref (&packet);
356 av_frame_free (&frame);
358 _pending_audio->trim_start (size);
362 FFmpegFileEncoder::subtitle (PlayerText, DCPTimePeriod)
368 FFmpegFileEncoder::buffer_free (void* opaque, uint8_t* data)
370 reinterpret_cast<FFmpegFileEncoder*>(opaque)->buffer_free2(data);
374 FFmpegFileEncoder::buffer_free2 (uint8_t* data)
376 _pending_images.erase (data);