src/feature/ffmpeg/ffmpeg-encoder.c (view raw)
1/* Copyright (c) 2013-2015 Jeffrey Pfau
2 *
3 * This Source Code Form is subject to the terms of the Mozilla Public
4 * License, v. 2.0. If a copy of the MPL was not distributed with this
5 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
6#include "ffmpeg-encoder.h"
7
8#include <mgba/core/core.h>
9#include <mgba/gba/interface.h>
10
11#include <libavcodec/version.h>
12#include <libavcodec/avcodec.h>
13
14#include <libavfilter/buffersink.h>
15#include <libavfilter/buffersrc.h>
16
17#include <libavutil/version.h>
18#if LIBAVUTIL_VERSION_MAJOR >= 53
19#include <libavutil/buffer.h>
20#endif
21#include <libavutil/imgutils.h>
22#include <libavutil/mathematics.h>
23#include <libavutil/opt.h>
24
25#ifdef USE_LIBAVRESAMPLE
26#include <libavresample/avresample.h>
27#else
28#include <libswresample/swresample.h>
29#endif
30#include <libswscale/swscale.h>
31
32static void _ffmpegPostVideoFrame(struct mAVStream*, const color_t* pixels, size_t stride);
33static void _ffmpegPostAudioFrame(struct mAVStream*, int16_t left, int16_t right);
34static void _ffmpegSetVideoDimensions(struct mAVStream*, unsigned width, unsigned height);
35
36static bool _ffmpegWriteAudioFrame(struct FFmpegEncoder* encoder, struct AVFrame* audioFrame);
37static bool _ffmpegWriteVideoFrame(struct FFmpegEncoder* encoder, struct AVFrame* videoFrame);
38
39enum {
40 PREFERRED_SAMPLE_RATE = 0x8000
41};
42
43void FFmpegEncoderInit(struct FFmpegEncoder* encoder) {
44#if LIBAVFORMAT_VERSION_INT < AV_VERSION_INT(58, 9, 100)
45 av_register_all();
46#endif
47
48 encoder->d.videoDimensionsChanged = _ffmpegSetVideoDimensions;
49 encoder->d.postVideoFrame = _ffmpegPostVideoFrame;
50 encoder->d.postAudioFrame = _ffmpegPostAudioFrame;
51 encoder->d.postAudioBuffer = 0;
52
53 encoder->audioCodec = NULL;
54 encoder->videoCodec = NULL;
55 encoder->containerFormat = NULL;
56 FFmpegEncoderSetAudio(encoder, "flac", 0);
57 FFmpegEncoderSetVideo(encoder, "libx264", 0, 0);
58 FFmpegEncoderSetContainer(encoder, "matroska");
59 FFmpegEncoderSetDimensions(encoder, GBA_VIDEO_HORIZONTAL_PIXELS, GBA_VIDEO_VERTICAL_PIXELS);
60 encoder->iwidth = GBA_VIDEO_HORIZONTAL_PIXELS;
61 encoder->iheight = GBA_VIDEO_VERTICAL_PIXELS;
62 encoder->frameskip = 1;
63 encoder->skipResidue = 0;
64 encoder->ipixFormat =
65#ifdef COLOR_16_BIT
66#ifdef COLOR_5_6_5
67 AV_PIX_FMT_RGB565;
68#else
69 AV_PIX_FMT_BGR555;
70#endif
71#else
72#ifndef USE_LIBAV
73 AV_PIX_FMT_0BGR32;
74#else
75 AV_PIX_FMT_BGR32;
76#endif
77#endif
78 encoder->resampleContext = NULL;
79 encoder->absf = NULL;
80 encoder->context = NULL;
81 encoder->scaleContext = NULL;
82 encoder->audio = NULL;
83 encoder->audioStream = NULL;
84 encoder->audioFrame = NULL;
85 encoder->audioBuffer = NULL;
86 encoder->postaudioBuffer = NULL;
87 encoder->video = NULL;
88 encoder->videoStream = NULL;
89 encoder->videoFrame = NULL;
90 encoder->graph = NULL;
91 encoder->source = NULL;
92 encoder->sink = NULL;
93 encoder->sinkFrame = NULL;
94
95 int i;
96 for (i = 0; i < FFMPEG_FILTERS_MAX; ++i) {
97 encoder->filters[i] = NULL;
98 }
99}
100
101bool FFmpegEncoderSetAudio(struct FFmpegEncoder* encoder, const char* acodec, unsigned abr) {
102 static const struct {
103 int format;
104 int priority;
105 } priorities[] = {
106 { AV_SAMPLE_FMT_S16, 0 },
107 { AV_SAMPLE_FMT_S16P, 1 },
108 { AV_SAMPLE_FMT_S32, 2 },
109 { AV_SAMPLE_FMT_S32P, 2 },
110 { AV_SAMPLE_FMT_FLT, 3 },
111 { AV_SAMPLE_FMT_FLTP, 3 },
112 { AV_SAMPLE_FMT_DBL, 4 },
113 { AV_SAMPLE_FMT_DBLP, 4 }
114 };
115
116 if (!acodec) {
117 encoder->audioCodec = 0;
118 return true;
119 }
120
121 AVCodec* codec = avcodec_find_encoder_by_name(acodec);
122 if (!codec) {
123 return false;
124 }
125
126 if (!codec->sample_fmts) {
127 return false;
128 }
129 size_t i;
130 size_t j;
131 int priority = INT_MAX;
132 encoder->sampleFormat = AV_SAMPLE_FMT_NONE;
133 for (i = 0; codec->sample_fmts[i] != AV_SAMPLE_FMT_NONE; ++i) {
134 for (j = 0; j < sizeof(priorities) / sizeof(*priorities); ++j) {
135 if (codec->sample_fmts[i] == priorities[j].format && priority > priorities[j].priority) {
136 priority = priorities[j].priority;
137 encoder->sampleFormat = codec->sample_fmts[i];
138 }
139 }
140 }
141 if (encoder->sampleFormat == AV_SAMPLE_FMT_NONE) {
142 return false;
143 }
144 encoder->sampleRate = PREFERRED_SAMPLE_RATE;
145 if (codec->supported_samplerates) {
146 for (i = 0; codec->supported_samplerates[i]; ++i) {
147 if (codec->supported_samplerates[i] < PREFERRED_SAMPLE_RATE) {
148 continue;
149 }
150 if (encoder->sampleRate == PREFERRED_SAMPLE_RATE || encoder->sampleRate > codec->supported_samplerates[i]) {
151 encoder->sampleRate = codec->supported_samplerates[i];
152 }
153 }
154 } else if (codec->id == AV_CODEC_ID_AAC) {
155 // HACK: AAC doesn't support 32768Hz (it rounds to 32000), but libfaac doesn't tell us that
156 encoder->sampleRate = 44100;
157 }
158 encoder->audioCodec = acodec;
159 encoder->audioBitrate = abr;
160 return true;
161}
162
163bool FFmpegEncoderSetVideo(struct FFmpegEncoder* encoder, const char* vcodec, unsigned vbr, int frameskip) {
164 static const struct {
165 enum AVPixelFormat format;
166 int priority;
167 } priorities[] = {
168 { AV_PIX_FMT_RGB555, 0 },
169 { AV_PIX_FMT_BGR555, 0 },
170 { AV_PIX_FMT_RGB565, 1 },
171 { AV_PIX_FMT_BGR565, 1 },
172 { AV_PIX_FMT_RGB24, 2 },
173 { AV_PIX_FMT_BGR24, 2 },
174#ifndef USE_LIBAV
175 { AV_PIX_FMT_BGR0, 3 },
176 { AV_PIX_FMT_RGB0, 3 },
177 { AV_PIX_FMT_0BGR, 3 },
178 { AV_PIX_FMT_0RGB, 3 },
179#endif
180 { AV_PIX_FMT_YUV422P, 4 },
181 { AV_PIX_FMT_YUV444P, 5 },
182 { AV_PIX_FMT_YUV420P, 6 },
183 { AV_PIX_FMT_PAL8, 7 },
184 };
185
186 if (!vcodec) {
187 encoder->videoCodec = 0;
188 return true;
189 }
190
191 AVCodec* codec = avcodec_find_encoder_by_name(vcodec);
192 if (!codec) {
193 return false;
194 }
195
196 size_t i;
197 size_t j;
198 int priority = INT_MAX;
199 encoder->pixFormat = AV_PIX_FMT_NONE;
200 for (i = 0; codec->pix_fmts[i] != AV_PIX_FMT_NONE; ++i) {
201 for (j = 0; j < sizeof(priorities) / sizeof(*priorities); ++j) {
202 if (codec->pix_fmts[i] == priorities[j].format && priority > priorities[j].priority) {
203 priority = priorities[j].priority;
204 encoder->pixFormat = codec->pix_fmts[i];
205 }
206 }
207 }
208 if (encoder->pixFormat == AV_PIX_FMT_NONE) {
209 return false;
210 }
211 encoder->videoCodec = vcodec;
212 encoder->videoBitrate = vbr;
213 encoder->frameskip = frameskip + 1;
214 return true;
215}
216
217bool FFmpegEncoderSetContainer(struct FFmpegEncoder* encoder, const char* container) {
218 AVOutputFormat* oformat = av_guess_format(container, 0, 0);
219 if (!oformat) {
220 return false;
221 }
222 encoder->containerFormat = container;
223 return true;
224}
225
226void FFmpegEncoderSetDimensions(struct FFmpegEncoder* encoder, int width, int height) {
227 encoder->width = width > 0 ? width : GBA_VIDEO_HORIZONTAL_PIXELS;
228 encoder->height = height > 0 ? height : GBA_VIDEO_VERTICAL_PIXELS;
229}
230
231bool FFmpegEncoderVerifyContainer(struct FFmpegEncoder* encoder) {
232 AVOutputFormat* oformat = av_guess_format(encoder->containerFormat, 0, 0);
233 AVCodec* acodec = avcodec_find_encoder_by_name(encoder->audioCodec);
234 AVCodec* vcodec = avcodec_find_encoder_by_name(encoder->videoCodec);
235 if ((encoder->audioCodec && !acodec) || (encoder->videoCodec && !vcodec) || !oformat || (!acodec && !vcodec)) {
236 return false;
237 }
238 if (encoder->audioCodec && !avformat_query_codec(oformat, acodec->id, FF_COMPLIANCE_EXPERIMENTAL)) {
239 return false;
240 }
241 if (encoder->videoCodec && !avformat_query_codec(oformat, vcodec->id, FF_COMPLIANCE_EXPERIMENTAL)) {
242 return false;
243 }
244 return true;
245}
246
247bool FFmpegEncoderOpen(struct FFmpegEncoder* encoder, const char* outfile) {
248 AVCodec* acodec = avcodec_find_encoder_by_name(encoder->audioCodec);
249 AVCodec* vcodec = avcodec_find_encoder_by_name(encoder->videoCodec);
250 if ((encoder->audioCodec && !acodec) || (encoder->videoCodec && !vcodec) || !FFmpegEncoderVerifyContainer(encoder)) {
251 return false;
252 }
253
254 if (encoder->context) {
255 return false;
256 }
257
258 encoder->currentAudioSample = 0;
259 encoder->currentAudioFrame = 0;
260 encoder->currentVideoFrame = 0;
261 encoder->skipResidue = 0;
262
263 AVOutputFormat* oformat = av_guess_format(encoder->containerFormat, 0, 0);
264#ifndef USE_LIBAV
265 avformat_alloc_output_context2(&encoder->context, oformat, 0, outfile);
266#else
267 encoder->context = avformat_alloc_context();
268 strncpy(encoder->context->filename, outfile, sizeof(encoder->context->filename) - 1);
269 encoder->context->filename[sizeof(encoder->context->filename) - 1] = '\0';
270 encoder->context->oformat = oformat;
271#endif
272
273 if (acodec) {
274#ifdef FFMPEG_USE_CODECPAR
275 encoder->audioStream = avformat_new_stream(encoder->context, NULL);
276 encoder->audio = avcodec_alloc_context3(acodec);
277#else
278 encoder->audioStream = avformat_new_stream(encoder->context, acodec);
279 encoder->audio = encoder->audioStream->codec;
280#endif
281 encoder->audio->bit_rate = encoder->audioBitrate;
282 encoder->audio->channels = 2;
283 encoder->audio->channel_layout = AV_CH_LAYOUT_STEREO;
284 encoder->audio->sample_rate = encoder->sampleRate;
285 encoder->audio->sample_fmt = encoder->sampleFormat;
286 AVDictionary* opts = 0;
287 av_dict_set(&opts, "strict", "-2", 0);
288 if (encoder->context->oformat->flags & AVFMT_GLOBALHEADER) {
289#ifdef AV_CODEC_FLAG_GLOBAL_HEADER
290 encoder->audio->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
291#else
292 encoder->audio->flags |= CODEC_FLAG_GLOBAL_HEADER;
293#endif
294 }
295 int res = avcodec_open2(encoder->audio, acodec, &opts);
296 av_dict_free(&opts);
297 if (res < 0) {
298 FFmpegEncoderClose(encoder);
299 return false;
300 }
301#if LIBAVCODEC_VERSION_MAJOR >= 55
302 encoder->audioFrame = av_frame_alloc();
303#else
304 encoder->audioFrame = avcodec_alloc_frame();
305#endif
306 if (!encoder->audio->frame_size) {
307 encoder->audio->frame_size = 1;
308 }
309 encoder->audioFrame->nb_samples = encoder->audio->frame_size;
310 encoder->audioFrame->format = encoder->audio->sample_fmt;
311 encoder->audioFrame->pts = 0;
312#ifdef USE_LIBAVRESAMPLE
313 encoder->resampleContext = avresample_alloc_context();
314 av_opt_set_int(encoder->resampleContext, "in_channel_layout", AV_CH_LAYOUT_STEREO, 0);
315 av_opt_set_int(encoder->resampleContext, "out_channel_layout", AV_CH_LAYOUT_STEREO, 0);
316 av_opt_set_int(encoder->resampleContext, "in_sample_rate", PREFERRED_SAMPLE_RATE, 0);
317 av_opt_set_int(encoder->resampleContext, "out_sample_rate", encoder->sampleRate, 0);
318 av_opt_set_int(encoder->resampleContext, "in_sample_fmt", AV_SAMPLE_FMT_S16, 0);
319 av_opt_set_int(encoder->resampleContext, "out_sample_fmt", encoder->sampleFormat, 0);
320 avresample_open(encoder->resampleContext);
321#else
322 encoder->resampleContext = swr_alloc_set_opts(NULL, AV_CH_LAYOUT_STEREO, encoder->sampleFormat, encoder->sampleRate,
323 AV_CH_LAYOUT_STEREO, AV_SAMPLE_FMT_S16, PREFERRED_SAMPLE_RATE, 0, NULL);
324 swr_init(encoder->resampleContext);
325#endif
326 encoder->audioBufferSize = (encoder->audioFrame->nb_samples * PREFERRED_SAMPLE_RATE / encoder->sampleRate) * 4;
327 encoder->audioBuffer = av_malloc(encoder->audioBufferSize);
328 encoder->postaudioBufferSize = av_samples_get_buffer_size(0, encoder->audio->channels, encoder->audio->frame_size, encoder->audio->sample_fmt, 0);
329 encoder->postaudioBuffer = av_malloc(encoder->postaudioBufferSize);
330 avcodec_fill_audio_frame(encoder->audioFrame, encoder->audio->channels, encoder->audio->sample_fmt, (const uint8_t*) encoder->postaudioBuffer, encoder->postaudioBufferSize, 0);
331
332 if (encoder->audio->codec->id == AV_CODEC_ID_AAC &&
333 (strcasecmp(encoder->containerFormat, "mp4") ||
334 strcasecmp(encoder->containerFormat, "m4v") ||
335 strcasecmp(encoder->containerFormat, "mov"))) {
336 // MP4 container doesn't support the raw ADTS AAC format that the encoder spits out
337#ifdef FFMPEG_USE_NEW_BSF
338 av_bsf_alloc(av_bsf_get_by_name("aac_adtstoasc"), &encoder->absf);
339 avcodec_parameters_from_context(encoder->absf->par_in, encoder->audio);
340 av_bsf_init(encoder->absf);
341#else
342 encoder->absf = av_bitstream_filter_init("aac_adtstoasc");
343#endif
344 }
345#ifdef FFMPEG_USE_CODECPAR
346 avcodec_parameters_from_context(encoder->audioStream->codecpar, encoder->audio);
347#endif
348 }
349
350 if (vcodec) {
351#ifdef FFMPEG_USE_CODECPAR
352 encoder->videoStream = avformat_new_stream(encoder->context, NULL);
353 encoder->video = avcodec_alloc_context3(vcodec);
354#else
355 encoder->videoStream = avformat_new_stream(encoder->context, vcodec);
356 encoder->video = encoder->videoStream->codec;
357#endif
358 encoder->video->bit_rate = encoder->videoBitrate;
359 encoder->video->width = encoder->width;
360 encoder->video->height = encoder->height;
361 encoder->video->time_base = (AVRational) { VIDEO_TOTAL_LENGTH * encoder->frameskip, GBA_ARM7TDMI_FREQUENCY };
362 encoder->video->framerate = (AVRational) { GBA_ARM7TDMI_FREQUENCY, VIDEO_TOTAL_LENGTH * encoder->frameskip };
363 encoder->video->pix_fmt = encoder->pixFormat;
364 encoder->video->gop_size = 60;
365 encoder->video->max_b_frames = 3;
366 if (encoder->context->oformat->flags & AVFMT_GLOBALHEADER) {
367#ifdef AV_CODEC_FLAG_GLOBAL_HEADER
368 encoder->video->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
369#else
370 encoder->video->flags |= CODEC_FLAG_GLOBAL_HEADER;
371#endif
372 }
373
374 if (encoder->video->codec->id == AV_CODEC_ID_H264 &&
375 (strcasecmp(encoder->containerFormat, "mp4") ||
376 strcasecmp(encoder->containerFormat, "m4v") ||
377 strcasecmp(encoder->containerFormat, "mov"))) {
378 // QuickTime and a few other things require YUV420
379 encoder->video->pix_fmt = AV_PIX_FMT_YUV420P;
380 }
381#if LIBAVCODEC_VERSION_MAJOR >= 57
382 if (encoder->video->codec->id == AV_CODEC_ID_FFV1) {
383 av_opt_set(encoder->video->priv_data, "coder", "range_tab", 0);
384 }
385#endif
386
387 if (strcmp(vcodec->name, "libx264") == 0) {
388 // Try to adaptively figure out when you can use a slower encoder
389 if (encoder->width * encoder->height > 1000000) {
390 av_opt_set(encoder->video->priv_data, "preset", "superfast", 0);
391 } else if (encoder->width * encoder->height > 500000) {
392 av_opt_set(encoder->video->priv_data, "preset", "veryfast", 0);
393 } else {
394 av_opt_set(encoder->video->priv_data, "preset", "faster", 0);
395 }
396 if (encoder->videoBitrate == 0) {
397 av_opt_set(encoder->video->priv_data, "crf", "0", 0);
398 encoder->video->pix_fmt = AV_PIX_FMT_YUV444P;
399 }
400 }
401 if (strcmp(vcodec->name, "libvpx-vp9") == 0 && encoder->videoBitrate == 0) {
402 av_opt_set(encoder->video->priv_data, "lossless", "1", 0);
403 encoder->video->pix_fmt = AV_PIX_FMT_YUV444P;
404 }
405
406 if (encoder->pixFormat == AV_PIX_FMT_PAL8) {
407 encoder->graph = avfilter_graph_alloc();
408
409 const struct AVFilter* source = avfilter_get_by_name("buffer");
410 const struct AVFilter* sink = avfilter_get_by_name("buffersink");
411 const struct AVFilter* split = avfilter_get_by_name("split");
412 const struct AVFilter* palettegen = avfilter_get_by_name("palettegen");
413 const struct AVFilter* paletteuse = avfilter_get_by_name("paletteuse");
414
415 if (!source || !sink || !split || !palettegen || !paletteuse || !encoder->graph) {
416 FFmpegEncoderClose(encoder);
417 return false;
418 }
419
420 char args[256];
421 snprintf(args, sizeof(args), "video_size=%dx%d:pix_fmt=%d:time_base=%d/%d",
422 encoder->video->width, encoder->video->height, encoder->ipixFormat,
423 encoder->video->time_base.num, encoder->video->time_base.den);
424
425 int res = 0;
426 res |= avfilter_graph_create_filter(&encoder->source, source, NULL, args, NULL, encoder->graph);
427 res |= avfilter_graph_create_filter(&encoder->sink, sink, NULL, NULL, NULL, encoder->graph);
428 res |= avfilter_graph_create_filter(&encoder->filters[0], split, NULL, NULL, NULL, encoder->graph);
429 res |= avfilter_graph_create_filter(&encoder->filters[1], palettegen, NULL, "reserve_transparent=off", NULL, encoder->graph);
430 res |= avfilter_graph_create_filter(&encoder->filters[2], paletteuse, NULL, "dither=none", NULL, encoder->graph);
431 if (res < 0) {
432 FFmpegEncoderClose(encoder);
433 return false;
434 }
435
436 res = 0;
437 res |= avfilter_link(encoder->source, 0, encoder->filters[0], 0);
438 res |= avfilter_link(encoder->filters[0], 0, encoder->filters[1], 0);
439 res |= avfilter_link(encoder->filters[0], 1, encoder->filters[2], 0);
440 res |= avfilter_link(encoder->filters[1], 0, encoder->filters[2], 1);
441 res |= avfilter_link(encoder->filters[2], 0, encoder->sink, 0);
442 if (res < 0 || avfilter_graph_config(encoder->graph, NULL) < 0) {
443 FFmpegEncoderClose(encoder);
444 return false;
445 }
446
447#if LIBAVCODEC_VERSION_MAJOR >= 55
448 encoder->sinkFrame = av_frame_alloc();
449#else
450 encoder->sinkFrame = avcodec_alloc_frame();
451#endif
452 }
453 AVDictionary* opts = 0;
454 av_dict_set(&opts, "strict", "-2", 0);
455 int res = avcodec_open2(encoder->video, vcodec, &opts);
456 av_dict_free(&opts);
457 if (res < 0) {
458 FFmpegEncoderClose(encoder);
459 return false;
460 }
461#if LIBAVCODEC_VERSION_MAJOR >= 55
462 encoder->videoFrame = av_frame_alloc();
463#else
464 encoder->videoFrame = avcodec_alloc_frame();
465#endif
466 encoder->videoFrame->format = encoder->video->pix_fmt != AV_PIX_FMT_PAL8 ? encoder->video->pix_fmt : encoder->ipixFormat;
467 encoder->videoFrame->width = encoder->video->width;
468 encoder->videoFrame->height = encoder->video->height;
469 encoder->videoFrame->pts = 0;
470 _ffmpegSetVideoDimensions(&encoder->d, encoder->iwidth, encoder->iheight);
471 av_image_alloc(encoder->videoFrame->data, encoder->videoFrame->linesize, encoder->videoFrame->width, encoder->videoFrame->height, encoder->videoFrame->format, 32);
472#ifdef FFMPEG_USE_CODECPAR
473 avcodec_parameters_from_context(encoder->videoStream->codecpar, encoder->video);
474#endif
475 }
476
477 AVDictionary* opts = 0;
478 av_dict_set(&opts, "strict", "-2", 0);
479 bool res = avio_open(&encoder->context->pb, outfile, AVIO_FLAG_WRITE) < 0 || avformat_write_header(encoder->context, &opts) < 0;
480 av_dict_free(&opts);
481 if (res) {
482 FFmpegEncoderClose(encoder);
483 return false;
484 }
485 return true;
486}
487
488void FFmpegEncoderClose(struct FFmpegEncoder* encoder) {
489 if (encoder->audio) {
490 while (true) {
491 if (!_ffmpegWriteAudioFrame(encoder, NULL)) {
492 break;
493 }
494 }
495 }
496 if (encoder->video) {
497 if (encoder->graph) {
498 if (av_buffersrc_add_frame(encoder->source, NULL) >= 0) {
499 while (true) {
500 int res = av_buffersink_get_frame(encoder->sink, encoder->sinkFrame);
501 if (res < 0) {
502 break;
503 }
504 _ffmpegWriteVideoFrame(encoder, encoder->sinkFrame);
505 av_frame_unref(encoder->sinkFrame);
506 }
507 }
508 }
509 while (true) {
510 if (!_ffmpegWriteVideoFrame(encoder, NULL)) {
511 break;
512 }
513 }
514 }
515
516 if (encoder->context && encoder->context->pb) {
517 av_write_trailer(encoder->context);
518 avio_close(encoder->context->pb);
519 }
520
521 if (encoder->postaudioBuffer) {
522 av_free(encoder->postaudioBuffer);
523 encoder->postaudioBuffer = NULL;
524 }
525 if (encoder->audioBuffer) {
526 av_free(encoder->audioBuffer);
527 encoder->audioBuffer = NULL;
528 }
529
530 if (encoder->audioFrame) {
531#if LIBAVCODEC_VERSION_MAJOR >= 55
532 av_frame_free(&encoder->audioFrame);
533#else
534 avcodec_free_frame(&encoder->audioFrame);
535#endif
536 }
537 if (encoder->audio) {
538 avcodec_close(encoder->audio);
539 encoder->audio = NULL;
540 }
541
542 if (encoder->resampleContext) {
543#ifdef USE_LIBAVRESAMPLE
544 avresample_close(encoder->resampleContext);
545 encoder->resampleContext = NULL;
546#else
547 swr_free(&encoder->resampleContext);
548#endif
549 }
550
551 if (encoder->absf) {
552#ifdef FFMPEG_USE_NEW_BSF
553 av_bsf_free(&encoder->absf);
554#else
555 av_bitstream_filter_close(encoder->absf);
556 encoder->absf = NULL;
557#endif
558 }
559
560 if (encoder->videoFrame) {
561#if LIBAVCODEC_VERSION_MAJOR >= 55
562 av_frame_free(&encoder->videoFrame);
563#else
564 avcodec_free_frame(&encoder->videoFrame);
565#endif
566 }
567
568 if (encoder->sinkFrame) {
569#if LIBAVCODEC_VERSION_MAJOR >= 55
570 av_frame_free(&encoder->sinkFrame);
571#else
572 avcodec_free_frame(&encoder->sinkFrame);
573#endif
574 encoder->sinkFrame = NULL;
575 }
576
577 if (encoder->video) {
578 avcodec_close(encoder->video);
579 encoder->video = NULL;
580 }
581
582 if (encoder->scaleContext) {
583 sws_freeContext(encoder->scaleContext);
584 encoder->scaleContext = NULL;
585 }
586
587 if (encoder->graph) {
588 avfilter_graph_free(&encoder->graph);
589 encoder->graph = NULL;
590 encoder->source = NULL;
591 encoder->sink = NULL;
592
593 int i;
594 for (i = 0; i < FFMPEG_FILTERS_MAX; ++i) {
595 encoder->filters[i] = NULL;
596 }
597 }
598
599 if (encoder->context) {
600 avformat_free_context(encoder->context);
601 encoder->context = NULL;
602 }
603}
604
605bool FFmpegEncoderIsOpen(struct FFmpegEncoder* encoder) {
606 return !!encoder->context;
607}
608
609void _ffmpegPostAudioFrame(struct mAVStream* stream, int16_t left, int16_t right) {
610 struct FFmpegEncoder* encoder = (struct FFmpegEncoder*) stream;
611 if (!encoder->context || !encoder->audioCodec) {
612 return;
613 }
614
615 if (encoder->absf && !left) {
616 // XXX: AVBSF doesn't like silence. Figure out why.
617 left = 1;
618 }
619
620 encoder->audioBuffer[encoder->currentAudioSample * 2] = left;
621 encoder->audioBuffer[encoder->currentAudioSample * 2 + 1] = right;
622
623 ++encoder->currentAudioSample;
624
625 if (encoder->currentAudioSample * 4 < encoder->audioBufferSize) {
626 return;
627 }
628
629 int channelSize = 2 * av_get_bytes_per_sample(encoder->audio->sample_fmt);
630 encoder->currentAudioSample = 0;
631#ifdef USE_LIBAVRESAMPLE
632 avresample_convert(encoder->resampleContext, 0, 0, 0,
633 (uint8_t**) &encoder->audioBuffer, 0, encoder->audioBufferSize / 4);
634
635 if (avresample_available(encoder->resampleContext) < encoder->audioFrame->nb_samples) {
636 return;
637 }
638#if LIBAVCODEC_VERSION_MAJOR >= 55
639 av_frame_make_writable(encoder->audioFrame);
640#endif
641 int samples = avresample_read(encoder->resampleContext, encoder->audioFrame->data, encoder->postaudioBufferSize / channelSize);
642#else
643#if LIBAVCODEC_VERSION_MAJOR >= 55
644 av_frame_make_writable(encoder->audioFrame);
645#endif
646 if (swr_get_out_samples(encoder->resampleContext, 1) < encoder->audioFrame->nb_samples) {
647 swr_convert(encoder->resampleContext, NULL, 0, (const uint8_t**) &encoder->audioBuffer, encoder->audioBufferSize / 4);
648 return;
649 }
650 int samples = swr_convert(encoder->resampleContext, encoder->audioFrame->data, encoder->postaudioBufferSize / channelSize,
651 (const uint8_t**) &encoder->audioBuffer, encoder->audioBufferSize / 4);
652#endif
653
654 encoder->audioFrame->pts = encoder->currentAudioFrame;
655 encoder->currentAudioFrame += samples;
656
657 _ffmpegWriteAudioFrame(encoder, encoder->audioFrame);
658}
659
660bool _ffmpegWriteAudioFrame(struct FFmpegEncoder* encoder, struct AVFrame* audioFrame) {
661 AVPacket packet;
662 av_init_packet(&packet);
663 packet.data = 0;
664 packet.size = 0;
665
666 int gotData;
667#ifdef FFMPEG_USE_PACKETS
668 avcodec_send_frame(encoder->audio, audioFrame);
669 gotData = avcodec_receive_packet(encoder->audio, &packet);
670 gotData = (gotData == 0) && packet.size;
671#else
672 avcodec_encode_audio2(encoder->audio, &packet, audioFrame, &gotData);
673#endif
674 packet.pts = av_rescale_q(packet.pts, encoder->audio->time_base, encoder->audioStream->time_base);
675 packet.dts = packet.pts;
676
677 if (gotData) {
678 if (encoder->absf) {
679 AVPacket tempPacket;
680
681#ifdef FFMPEG_USE_NEW_BSF
682 int success = av_bsf_send_packet(encoder->absf, &packet);
683 if (success >= 0) {
684 success = av_bsf_receive_packet(encoder->absf, &tempPacket);
685 }
686#else
687 int success = av_bitstream_filter_filter(encoder->absf, encoder->audio, 0,
688 &tempPacket.data, &tempPacket.size,
689 packet.data, packet.size, 0);
690#endif
691
692 if (success >= 0) {
693#if LIBAVUTIL_VERSION_MAJOR >= 53
694 tempPacket.buf = av_buffer_create(tempPacket.data, tempPacket.size, av_buffer_default_free, 0, 0);
695#endif
696
697#ifdef FFMPEG_USE_PACKET_UNREF
698 av_packet_move_ref(&packet, &tempPacket);
699#else
700 av_free_packet(&packet);
701 packet = tempPacket;
702#endif
703
704 packet.stream_index = encoder->audioStream->index;
705 av_interleaved_write_frame(encoder->context, &packet);
706 }
707 } else {
708 packet.stream_index = encoder->audioStream->index;
709 av_interleaved_write_frame(encoder->context, &packet);
710 }
711 }
712#ifdef FFMPEG_USE_PACKET_UNREF
713 av_packet_unref(&packet);
714#else
715 av_free_packet(&packet);
716#endif
717 return gotData;
718}
719
720void _ffmpegPostVideoFrame(struct mAVStream* stream, const color_t* pixels, size_t stride) {
721 struct FFmpegEncoder* encoder = (struct FFmpegEncoder*) stream;
722 if (!encoder->context || !encoder->videoCodec) {
723 return;
724 }
725 encoder->skipResidue = (encoder->skipResidue + 1) % encoder->frameskip;
726 if (encoder->skipResidue) {
727 return;
728 }
729 stride *= BYTES_PER_PIXEL;
730
731#if LIBAVCODEC_VERSION_MAJOR >= 55
732 av_frame_make_writable(encoder->videoFrame);
733#endif
734 encoder->videoFrame->pts = av_rescale_q(encoder->currentVideoFrame, encoder->video->time_base, encoder->videoStream->time_base);
735 ++encoder->currentVideoFrame;
736
737 sws_scale(encoder->scaleContext, (const uint8_t* const*) &pixels, (const int*) &stride, 0, encoder->iheight, encoder->videoFrame->data, encoder->videoFrame->linesize);
738
739 if (encoder->graph) {
740 if (av_buffersrc_add_frame(encoder->source, encoder->videoFrame) < 0) {
741 return;
742 }
743 while (true) {
744 int res = av_buffersink_get_frame(encoder->sink, encoder->sinkFrame);
745 if (res < 0) {
746 break;
747 }
748 _ffmpegWriteVideoFrame(encoder, encoder->sinkFrame);
749 av_frame_unref(encoder->sinkFrame);
750 }
751 } else {
752 _ffmpegWriteVideoFrame(encoder, encoder->videoFrame);
753 }
754}
755
756bool _ffmpegWriteVideoFrame(struct FFmpegEncoder* encoder, struct AVFrame* videoFrame) {
757 AVPacket packet;
758
759 av_init_packet(&packet);
760 packet.data = 0;
761 packet.size = 0;
762
763 int gotData;
764#ifdef FFMPEG_USE_PACKETS
765 avcodec_send_frame(encoder->video, videoFrame);
766 gotData = avcodec_receive_packet(encoder->video, &packet) == 0;
767#else
768 avcodec_encode_video2(encoder->video, &packet, videoFrame, &gotData);
769#endif
770 if (gotData) {
771#ifndef FFMPEG_USE_PACKET_UNREF
772 if (encoder->video->coded_frame->key_frame) {
773 packet.flags |= AV_PKT_FLAG_KEY;
774 }
775#endif
776 packet.stream_index = encoder->videoStream->index;
777 av_interleaved_write_frame(encoder->context, &packet);
778 }
779#ifdef FFMPEG_USE_PACKET_UNREF
780 av_packet_unref(&packet);
781#else
782 av_free_packet(&packet);
783#endif
784
785 return gotData;
786}
787
788static void _ffmpegSetVideoDimensions(struct mAVStream* stream, unsigned width, unsigned height) {
789 struct FFmpegEncoder* encoder = (struct FFmpegEncoder*) stream;
790 if (!encoder->context || !encoder->videoCodec) {
791 return;
792 }
793 encoder->iwidth = width;
794 encoder->iheight = height;
795 if (encoder->scaleContext) {
796 sws_freeContext(encoder->scaleContext);
797 }
798 encoder->scaleContext = sws_getContext(encoder->iwidth, encoder->iheight, encoder->ipixFormat,
799 encoder->videoFrame->width, encoder->videoFrame->height, encoder->videoFrame->format,
800 SWS_POINT, 0, 0, 0);
801}