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:
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);
88 #ifdef DCPOMATIC_VARIANT_SWAROOP
89 int r = avformat_alloc_output_context2 (&_format_context, av_guess_format("mp4", 0, 0), 0, 0);
91 int r = avformat_alloc_output_context2 (&_format_context, 0, 0, _output.string().c_str());
93 if (!_format_context) {
94 throw runtime_error (String::compose("could not allocate FFmpeg format context (%1)", r));
97 _video_stream = avformat_new_stream (_format_context, _video_codec);
99 throw runtime_error ("could not create FFmpeg output video stream");
102 _audio_stream = avformat_new_stream (_format_context, _audio_codec);
103 if (!_audio_stream) {
104 throw runtime_error ("could not create FFmpeg output audio stream");
107 _video_stream->id = _video_stream_index;
108 _video_stream->codec = _video_codec_context;
110 _audio_stream->id = _audio_stream_index;
111 _audio_stream->codec = _audio_codec_context;
113 if (avcodec_open2 (_video_codec_context, _video_codec, &_video_options) < 0) {
114 throw runtime_error ("could not open FFmpeg video codec");
117 r = avcodec_open2 (_audio_codec_context, _audio_codec, 0);
120 av_strerror (r, buffer, sizeof(buffer));
121 throw runtime_error (String::compose ("could not open FFmpeg audio codec (%1)", buffer));
124 if (avio_open_boost (&_format_context->pb, _output, AVIO_FLAG_WRITE) < 0) {
125 throw runtime_error ("could not open FFmpeg output file");
128 AVDictionary* options = 0;
130 #ifdef DCPOMATIC_VARIANT_SWAROOP
132 av_dict_set (&options, "encryption_key", key->hex().c_str(), 0);
133 /* XXX: is this OK? */
134 av_dict_set (&options, "encryption_kid", "00000000000000000000000000000000", 0);
135 av_dict_set (&options, "encryption_scheme", "cenc-aes-ctr", 0);
139 if (av_dict_set(&_format_context->metadata, SWAROOP_ID_TAG, id->c_str(), 0) < 0) {
140 throw runtime_error ("Could not write ID to output");
145 if (avformat_write_header (_format_context, &options) < 0) {
146 throw runtime_error ("could not write header to FFmpeg output file");
149 _pending_audio.reset (new AudioBuffers(channels, 0));
153 FFmpegFileEncoder::pixel_format (ExportFormat format)
156 case EXPORT_FORMAT_PRORES:
157 return AV_PIX_FMT_YUV422P10;
158 case EXPORT_FORMAT_H264:
159 return AV_PIX_FMT_YUV420P;
161 DCPOMATIC_ASSERT (false);
164 return AV_PIX_FMT_YUV422P10;
168 FFmpegFileEncoder::setup_video ()
170 _video_codec = avcodec_find_encoder_by_name (_video_codec_name.c_str());
172 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _video_codec_name));
175 _video_codec_context = avcodec_alloc_context3 (_video_codec);
176 if (!_video_codec_context) {
177 throw runtime_error ("could not allocate FFmpeg video context");
180 avcodec_get_context_defaults3 (_video_codec_context, _video_codec);
182 /* Variable quantisation */
183 _video_codec_context->global_quality = 0;
184 _video_codec_context->width = _video_frame_size.width;
185 _video_codec_context->height = _video_frame_size.height;
186 _video_codec_context->time_base = (AVRational) { 1, _video_frame_rate };
187 _video_codec_context->pix_fmt = _pixel_format;
188 _video_codec_context->flags |= AV_CODEC_FLAG_QSCALE | AV_CODEC_FLAG_GLOBAL_HEADER;
192 FFmpegFileEncoder::setup_audio ()
194 _audio_codec = avcodec_find_encoder_by_name (_audio_codec_name.c_str());
196 throw runtime_error (String::compose ("could not find FFmpeg encoder %1", _audio_codec_name));
199 _audio_codec_context = avcodec_alloc_context3 (_audio_codec);
200 if (!_audio_codec_context) {
201 throw runtime_error ("could not allocate FFmpeg audio context");
204 avcodec_get_context_defaults3 (_audio_codec_context, _audio_codec);
206 /* XXX: configurable */
207 _audio_codec_context->bit_rate = _audio_channels * 128 * 1024;
208 _audio_codec_context->sample_fmt = _sample_format;
209 _audio_codec_context->sample_rate = _audio_frame_rate;
210 _audio_codec_context->channel_layout = av_get_default_channel_layout (_audio_channels);
211 _audio_codec_context->channels = _audio_channels;
215 FFmpegFileEncoder::flush ()
217 if (_pending_audio->frames() > 0) {
218 audio_frame (_pending_audio->frames ());
221 bool flushed_video = false;
222 bool flushed_audio = false;
224 while (!flushed_video || !flushed_audio) {
226 av_init_packet (&packet);
231 avcodec_encode_video2 (_video_codec_context, &packet, 0, &got_packet);
233 packet.stream_index = 0;
234 av_interleaved_write_frame (_format_context, &packet);
236 flushed_video = true;
238 av_packet_unref (&packet);
240 av_init_packet (&packet);
244 avcodec_encode_audio2 (_audio_codec_context, &packet, 0, &got_packet);
246 packet.stream_index = 0;
247 av_interleaved_write_frame (_format_context, &packet);
249 flushed_audio = true;
251 av_packet_unref (&packet);
254 av_write_trailer (_format_context);
256 avcodec_close (_video_codec_context);
257 avcodec_close (_audio_codec_context);
258 avio_close (_format_context->pb);
259 avformat_free_context (_format_context);
263 FFmpegFileEncoder::video (shared_ptr<PlayerVideo> video, DCPTime time)
265 shared_ptr<Image> image = video->image (
266 bind (&PlayerVideo::force, _1, _pixel_format),
271 AVFrame* frame = av_frame_alloc ();
272 DCPOMATIC_ASSERT (frame);
274 _pending_images[image->data()[0]] = image;
275 for (int i = 0; i < 3; ++i) {
276 AVBufferRef* buffer = av_buffer_create(image->data()[i], image->stride()[i] * image->size().height, &buffer_free, this, 0);
277 frame->buf[i] = av_buffer_ref (buffer);
278 frame->data[i] = buffer->data;
279 frame->linesize[i] = image->stride()[i];
280 av_buffer_unref (&buffer);
283 frame->width = image->size().width;
284 frame->height = image->size().height;
285 frame->format = _pixel_format;
286 frame->pts = time.seconds() / av_q2d (_video_stream->time_base);
289 av_init_packet (&packet);
294 if (avcodec_encode_video2 (_video_codec_context, &packet, frame, &got_packet) < 0) {
295 throw EncodeError ("FFmpeg video encode failed");
298 if (got_packet && packet.size) {
299 packet.stream_index = _video_stream_index;
300 av_interleaved_write_frame (_format_context, &packet);
301 av_packet_unref (&packet);
304 av_frame_free (&frame);
308 /** Called when the player gives us some audio */
310 FFmpegFileEncoder::audio (shared_ptr<AudioBuffers> audio)
312 _pending_audio->append (audio);
314 int frame_size = _audio_codec_context->frame_size;
315 if (frame_size == 0) {
316 /* codec has AV_CODEC_CAP_VARIABLE_FRAME_SIZE */
317 frame_size = _audio_frame_rate / _video_frame_rate;
320 while (_pending_audio->frames() >= frame_size) {
321 audio_frame (frame_size);
326 FFmpegFileEncoder::audio_frame (int size)
328 DCPOMATIC_ASSERT (size);
330 AVFrame* frame = av_frame_alloc ();
331 DCPOMATIC_ASSERT (frame);
333 int const channels = _pending_audio->channels();
334 DCPOMATIC_ASSERT (channels);
336 int const buffer_size = av_samples_get_buffer_size (0, channels, size, _audio_codec_context->sample_fmt, 0);
337 DCPOMATIC_ASSERT (buffer_size >= 0);
339 void* samples = av_malloc (buffer_size);
340 DCPOMATIC_ASSERT (samples);
342 frame->nb_samples = size;
343 int r = avcodec_fill_audio_frame (frame, channels, _audio_codec_context->sample_fmt, (const uint8_t *) samples, buffer_size, 0);
344 DCPOMATIC_ASSERT (r >= 0);
346 float** p = _pending_audio->data ();
347 switch (_audio_codec_context->sample_fmt) {
348 case AV_SAMPLE_FMT_S16:
350 int16_t* q = reinterpret_cast<int16_t*> (samples);
351 for (int i = 0; i < size; ++i) {
352 for (int j = 0; j < channels; ++j) {
353 *q++ = p[j][i] * 32767;
358 case AV_SAMPLE_FMT_FLTP:
360 float* q = reinterpret_cast<float*> (samples);
361 for (int i = 0; i < channels; ++i) {
362 memcpy (q, p[i], sizeof(float) * size);
368 DCPOMATIC_ASSERT (false);
372 av_init_packet (&packet);
377 if (avcodec_encode_audio2 (_audio_codec_context, &packet, frame, &got_packet) < 0) {
378 throw EncodeError ("FFmpeg audio encode failed");
381 if (got_packet && packet.size) {
382 packet.stream_index = _audio_stream_index;
383 av_interleaved_write_frame (_format_context, &packet);
384 av_packet_unref (&packet);
388 av_frame_free (&frame);
390 _pending_audio->trim_start (size);
394 FFmpegFileEncoder::subtitle (PlayerText, DCPTimePeriod)
400 FFmpegFileEncoder::buffer_free (void* opaque, uint8_t* data)
402 reinterpret_cast<FFmpegFileEncoder*>(opaque)->buffer_free2(data);
406 FFmpegFileEncoder::buffer_free2 (uint8_t* data)
408 _pending_images.erase (data);