src/platform/ffmpeg/ffmpeg-encoder.c (view raw)
1#include "ffmpeg-encoder.h"
2
3#include "gba-video.h"
4
5#include <libavcodec/avcodec.h>
6
7#include <libavutil/imgutils.h>
8#include <libavutil/opt.h>
9
10#include <libavresample/avresample.h>
11#include <libswscale/swscale.h>
12
13static void _ffmpegPostVideoFrame(struct GBAAVStream*, struct GBAVideoRenderer* renderer);
14static void _ffmpegPostAudioFrame(struct GBAAVStream*, int32_t left, int32_t right);
15
16enum {
17 PREFERRED_SAMPLE_RATE = 0x8000
18};
19
20void FFmpegEncoderInit(struct FFmpegEncoder* encoder) {
21 av_register_all();
22
23 encoder->d.postVideoFrame = _ffmpegPostVideoFrame;
24 encoder->d.postAudioFrame = _ffmpegPostAudioFrame;
25
26 encoder->audioCodec = 0;
27 encoder->videoCodec = 0;
28 encoder->containerFormat = 0;
29 FFmpegEncoderSetAudio(encoder, "flac", 0);
30 FFmpegEncoderSetVideo(encoder, "png", 0);
31 FFmpegEncoderSetContainer(encoder, "matroska");
32 FFmpegEncoderSetDimensions(encoder, VIDEO_HORIZONTAL_PIXELS, VIDEO_VERTICAL_PIXELS);
33 encoder->resampleContext = 0;
34 encoder->absf = 0;
35 encoder->context = 0;
36}
37
38bool FFmpegEncoderSetAudio(struct FFmpegEncoder* encoder, const char* acodec, unsigned abr) {
39 static const struct {
40 int format;
41 int priority;
42 } priorities[] = {
43 { AV_SAMPLE_FMT_S16, 0 },
44 { AV_SAMPLE_FMT_S16P, 1 },
45 { AV_SAMPLE_FMT_S32, 2 },
46 { AV_SAMPLE_FMT_S32P, 2 },
47 { AV_SAMPLE_FMT_FLT, 3 },
48 { AV_SAMPLE_FMT_FLTP, 3 },
49 { AV_SAMPLE_FMT_DBL, 4 },
50 { AV_SAMPLE_FMT_DBLP, 4 }
51 };
52 AVCodec* codec = avcodec_find_encoder_by_name(acodec);
53 if (!codec) {
54 return false;
55 }
56
57 if (!codec->sample_fmts) {
58 return false;
59 }
60 size_t i;
61 size_t j;
62 int priority = INT_MAX;
63 encoder->sampleFormat = AV_SAMPLE_FMT_NONE;
64 for (i = 0; codec->sample_fmts[i] != AV_SAMPLE_FMT_NONE; ++i) {
65 for (j = 0; j < sizeof(priorities) / sizeof(*priorities); ++j) {
66 if (codec->sample_fmts[i] == priorities[j].format && priority > priorities[j].priority) {
67 priority = priorities[j].priority;
68 encoder->sampleFormat = codec->sample_fmts[i];
69 }
70 }
71 }
72 if (encoder->sampleFormat == AV_SAMPLE_FMT_NONE) {
73 return false;
74 }
75 encoder->sampleRate = PREFERRED_SAMPLE_RATE;
76 if (codec->supported_samplerates) {
77 for (i = 0; codec->supported_samplerates[i]; ++i) {
78 if (codec->supported_samplerates[i] < PREFERRED_SAMPLE_RATE) {
79 continue;
80 }
81 if (encoder->sampleRate == PREFERRED_SAMPLE_RATE || encoder->sampleRate > codec->supported_samplerates[i]) {
82 encoder->sampleRate = codec->supported_samplerates[i];
83 }
84 }
85 } else if (codec->id == AV_CODEC_ID_AAC) {
86 // HACK: AAC doesn't support 32768Hz (it rounds to 32000), but libfaac doesn't tell us that
87 encoder->sampleRate = 44100;
88 }
89 encoder->audioCodec = acodec;
90 encoder->audioBitrate = abr;
91 return true;
92}
93
94bool FFmpegEncoderSetVideo(struct FFmpegEncoder* encoder, const char* vcodec, unsigned vbr) {
95 static const struct {
96 enum AVPixelFormat format;
97 int priority;
98 } priorities[] = {
99 { AV_PIX_FMT_RGB555, 0 },
100 { AV_PIX_FMT_BGR555, 0 },
101 { AV_PIX_FMT_RGB565, 1 },
102 { AV_PIX_FMT_BGR565, 1 },
103 { AV_PIX_FMT_RGB24, 2 },
104 { AV_PIX_FMT_BGR24, 2 },
105 { AV_PIX_FMT_BGR0, 3 },
106 { AV_PIX_FMT_RGB0, 3 },
107 { AV_PIX_FMT_0BGR, 3 },
108 { AV_PIX_FMT_0RGB, 3 },
109 { AV_PIX_FMT_YUV422P, 4 },
110 { AV_PIX_FMT_YUV444P, 5 },
111 { AV_PIX_FMT_YUV420P, 6 }
112 };
113 AVCodec* codec = avcodec_find_encoder_by_name(vcodec);
114 if (!codec) {
115 return false;
116 }
117
118 size_t i;
119 size_t j;
120 int priority = INT_MAX;
121 encoder->pixFormat = AV_PIX_FMT_NONE;
122 for (i = 0; codec->pix_fmts[i] != AV_PIX_FMT_NONE; ++i) {
123 for (j = 0; j < sizeof(priorities) / sizeof(*priorities); ++j) {
124 if (codec->pix_fmts[i] == priorities[j].format && priority > priorities[j].priority) {
125 priority = priorities[j].priority;
126 encoder->pixFormat = codec->pix_fmts[i];
127 }
128 }
129 }
130 if (encoder->pixFormat == AV_PIX_FMT_NONE) {
131 return false;
132 }
133 encoder->videoCodec = vcodec;
134 encoder->videoBitrate = vbr;
135 return true;
136}
137
138bool FFmpegEncoderSetContainer(struct FFmpegEncoder* encoder, const char* container) {
139 AVOutputFormat* oformat = av_guess_format(container, 0, 0);
140 if (!oformat) {
141 return false;
142 }
143 encoder->containerFormat = container;
144 return true;
145}
146
147void FFmpegEncoderSetDimensions(struct FFmpegEncoder* encoder, int width, int height) {
148 encoder->width = width > 0 ? width : VIDEO_HORIZONTAL_PIXELS;
149 encoder->height = height > 0 ? height : VIDEO_VERTICAL_PIXELS;
150}
151
152bool FFmpegEncoderVerifyContainer(struct FFmpegEncoder* encoder) {
153 AVOutputFormat* oformat = av_guess_format(encoder->containerFormat, 0, 0);
154 AVCodec* acodec = avcodec_find_encoder_by_name(encoder->audioCodec);
155 AVCodec* vcodec = avcodec_find_encoder_by_name(encoder->videoCodec);
156 if (!acodec || !vcodec || !oformat) {
157 return false;
158 }
159 if (!avformat_query_codec(oformat, acodec->id, FF_COMPLIANCE_EXPERIMENTAL)) {
160 return false;
161 }
162 if (!avformat_query_codec(oformat, vcodec->id, FF_COMPLIANCE_EXPERIMENTAL)) {
163 return false;
164 }
165 return true;
166}
167
168bool FFmpegEncoderOpen(struct FFmpegEncoder* encoder, const char* outfile) {
169 AVCodec* acodec = avcodec_find_encoder_by_name(encoder->audioCodec);
170 AVCodec* vcodec = avcodec_find_encoder_by_name(encoder->videoCodec);
171 if (!acodec || !vcodec || !FFmpegEncoderVerifyContainer(encoder)) {
172 return false;
173 }
174
175 encoder->currentAudioSample = 0;
176 encoder->currentAudioFrame = 0;
177 encoder->currentVideoFrame = 0;
178 encoder->nextAudioPts = 0;
179
180 avformat_alloc_output_context2(&encoder->context, 0, 0, outfile);
181
182 encoder->context->oformat = av_guess_format(encoder->containerFormat, 0, 0);
183
184 encoder->audioStream = avformat_new_stream(encoder->context, acodec);
185 encoder->audio = encoder->audioStream->codec;
186 encoder->audio->bit_rate = encoder->audioBitrate;
187 encoder->audio->channels = 2;
188 encoder->audio->channel_layout = AV_CH_LAYOUT_STEREO;
189 encoder->audio->sample_rate = encoder->sampleRate;
190 encoder->audio->sample_fmt = encoder->sampleFormat;
191 AVDictionary* opts = 0;
192 av_dict_set(&opts, "strict", "-2", 0);
193 if (encoder->context->oformat->flags & AVFMT_GLOBALHEADER) {
194 encoder->audio->flags |= CODEC_FLAG_GLOBAL_HEADER;
195 }
196 avcodec_open2(encoder->audio, acodec, &opts);
197 av_dict_free(&opts);
198 encoder->audioFrame = av_frame_alloc();
199 encoder->audioFrame->nb_samples = encoder->audio->frame_size;
200 encoder->audioFrame->format = encoder->audio->sample_fmt;
201 encoder->audioFrame->pts = 0;
202 encoder->resampleContext = avresample_alloc_context();
203 av_opt_set_int(encoder->resampleContext, "in_channel_layout", AV_CH_LAYOUT_STEREO, 0);
204 av_opt_set_int(encoder->resampleContext, "out_channel_layout", AV_CH_LAYOUT_STEREO, 0);
205 av_opt_set_int(encoder->resampleContext, "in_sample_rate", PREFERRED_SAMPLE_RATE, 0);
206 av_opt_set_int(encoder->resampleContext, "out_sample_rate", encoder->sampleRate, 0);
207 av_opt_set_int(encoder->resampleContext, "in_sample_fmt", AV_SAMPLE_FMT_S16, 0);
208 av_opt_set_int(encoder->resampleContext, "out_sample_fmt", encoder->sampleFormat, 0);
209 avresample_open(encoder->resampleContext);
210 encoder->audioBufferSize = (encoder->audioFrame->nb_samples * PREFERRED_SAMPLE_RATE / encoder->sampleRate) * 4;
211 encoder->audioBuffer = av_malloc(encoder->audioBufferSize);
212 encoder->postaudioBufferSize = av_samples_get_buffer_size(0, encoder->audio->channels, encoder->audio->frame_size, encoder->audio->sample_fmt, 0);
213 encoder->postaudioBuffer = av_malloc(encoder->postaudioBufferSize);
214 avcodec_fill_audio_frame(encoder->audioFrame, encoder->audio->channels, encoder->audio->sample_fmt, (const uint8_t*) encoder->postaudioBuffer, encoder->postaudioBufferSize, 0);
215
216 if (encoder->audio->codec->id == AV_CODEC_ID_AAC &&
217 (strcasecmp(encoder->containerFormat, "mp4") ||
218 strcasecmp(encoder->containerFormat, "m4v") ||
219 strcasecmp(encoder->containerFormat, "mov"))) {
220 // MP4 container doesn't support the raw ADTS AAC format that the encoder spits out
221 encoder->absf = av_bitstream_filter_init("aac_adtstoasc");
222 }
223
224 encoder->videoStream = avformat_new_stream(encoder->context, vcodec);
225 encoder->video = encoder->videoStream->codec;
226 encoder->video->bit_rate = encoder->videoBitrate;
227 encoder->video->width = encoder->width;
228 encoder->video->height = encoder->height;
229 encoder->video->time_base = (AVRational) { VIDEO_TOTAL_LENGTH, GBA_ARM7TDMI_FREQUENCY };
230 encoder->video->pix_fmt = encoder->pixFormat;
231 encoder->video->gop_size = 15;
232 encoder->video->max_b_frames = 0;
233 if (encoder->context->oformat->flags & AVFMT_GLOBALHEADER) {
234 encoder->video->flags |= CODEC_FLAG_GLOBAL_HEADER;
235 }
236 avcodec_open2(encoder->video, vcodec, 0);
237 encoder->videoFrame = av_frame_alloc();
238 encoder->videoFrame->format = encoder->video->pix_fmt;
239 encoder->videoFrame->width = encoder->video->width;
240 encoder->videoFrame->height = encoder->video->height;
241 encoder->videoFrame->pts = 0;
242 encoder->scaleContext = sws_getContext(VIDEO_HORIZONTAL_PIXELS, VIDEO_VERTICAL_PIXELS, AV_PIX_FMT_0BGR32,
243 encoder->videoFrame->width, encoder->videoFrame->height, encoder->video->pix_fmt,
244 SWS_POINT, 0, 0, 0);
245 av_image_alloc(encoder->videoFrame->data, encoder->videoFrame->linesize, encoder->video->width, encoder->video->height, encoder->video->pix_fmt, 32);
246
247 avio_open(&encoder->context->pb, outfile, AVIO_FLAG_WRITE);
248 avformat_write_header(encoder->context, 0);
249
250 return true;
251}
252
253void FFmpegEncoderClose(struct FFmpegEncoder* encoder) {
254 if (!encoder->context) {
255 return;
256 }
257 av_write_trailer(encoder->context);
258 avio_close(encoder->context->pb);
259
260 av_free(encoder->postaudioBuffer);
261 if (encoder->audioBuffer) {
262 av_free(encoder->audioBuffer);
263 }
264 av_frame_free(&encoder->audioFrame);
265 avcodec_close(encoder->audio);
266
267 av_frame_free(&encoder->videoFrame);
268 avcodec_close(encoder->video);
269
270 if (encoder->resampleContext) {
271 avresample_close(encoder->resampleContext);
272 }
273
274 if (encoder->absf) {
275 av_bitstream_filter_close(encoder->absf);
276 encoder->absf = 0;
277 }
278
279 sws_freeContext(encoder->scaleContext);
280
281 avformat_free_context(encoder->context);
282 encoder->context = 0;
283}
284
285bool FFmpegEncoderIsOpen(struct FFmpegEncoder* encoder) {
286 return !!encoder->context;
287}
288
289void _ffmpegPostAudioFrame(struct GBAAVStream* stream, int32_t left, int32_t right) {
290 struct FFmpegEncoder* encoder = (struct FFmpegEncoder*) stream;
291 if (!encoder->context) {
292 return;
293 }
294
295 av_frame_make_writable(encoder->audioFrame);
296 encoder->audioBuffer[encoder->currentAudioSample * 2] = left;
297 encoder->audioBuffer[encoder->currentAudioSample * 2 + 1] = right;
298
299 ++encoder->currentAudioFrame;
300 ++encoder->currentAudioSample;
301
302 if ((encoder->currentAudioSample * 4) < encoder->audioBufferSize) {
303 return;
304 }
305 encoder->currentAudioSample = 0;
306
307 int channelSize = 2 * av_get_bytes_per_sample(encoder->audio->sample_fmt);
308 avresample_convert(encoder->resampleContext,
309 0, 0, encoder->postaudioBufferSize / channelSize,
310 (uint8_t**) &encoder->audioBuffer, 0, encoder->audioBufferSize / 4);
311 if ((ssize_t) avresample_available(encoder->resampleContext) < (ssize_t) encoder->postaudioBufferSize / channelSize) {
312 return;
313 }
314 avresample_read(encoder->resampleContext, encoder->audioFrame->data, encoder->postaudioBufferSize / channelSize);
315
316 AVRational timeBase = { 1, PREFERRED_SAMPLE_RATE };
317 encoder->audioFrame->pts = encoder->nextAudioPts;
318 encoder->nextAudioPts = av_rescale_q(encoder->currentAudioFrame, timeBase, encoder->audioStream->time_base);
319
320 AVPacket packet;
321 av_init_packet(&packet);
322 packet.data = 0;
323 packet.size = 0;
324 int gotData;
325 avcodec_encode_audio2(encoder->audio, &packet, encoder->audioFrame, &gotData);
326 if (gotData) {
327 if (encoder->absf) {
328 AVPacket tempPacket = packet;
329 int success = av_bitstream_filter_filter(encoder->absf, encoder->audio, 0,
330 &tempPacket.data, &tempPacket.size,
331 packet.data, packet.size, 0);
332 if (success > 0) {
333 tempPacket.buf = av_buffer_create(tempPacket.data, tempPacket.size, av_buffer_default_free, 0, 0);
334 av_free_packet(&packet);
335 }
336 packet = tempPacket;
337 }
338 packet.stream_index = encoder->audioStream->index;
339 av_interleaved_write_frame(encoder->context, &packet);
340 }
341 av_free_packet(&packet);
342}
343
344void _ffmpegPostVideoFrame(struct GBAAVStream* stream, struct GBAVideoRenderer* renderer) {
345 struct FFmpegEncoder* encoder = (struct FFmpegEncoder*) stream;
346 if (!encoder->context) {
347 return;
348 }
349 uint8_t* pixels;
350 unsigned stride;
351 renderer->getPixels(renderer, &stride, (void**) &pixels);
352 stride *= 4;
353
354 AVPacket packet;
355
356 av_init_packet(&packet);
357 packet.data = 0;
358 packet.size = 0;
359 av_frame_make_writable(encoder->videoFrame);
360 encoder->videoFrame->pts = av_rescale_q(encoder->currentVideoFrame, encoder->video->time_base, encoder->videoStream->time_base);
361 ++encoder->currentVideoFrame;
362
363 sws_scale(encoder->scaleContext, (const uint8_t* const*) &pixels, (const int*) &stride, 0, VIDEO_VERTICAL_PIXELS, encoder->videoFrame->data, encoder->videoFrame->linesize);
364
365 int gotData;
366 avcodec_encode_video2(encoder->video, &packet, encoder->videoFrame, &gotData);
367 if (gotData) {
368 if (encoder->videoStream->codec->coded_frame->key_frame) {
369 packet.flags |= AV_PKT_FLAG_KEY;
370 }
371 packet.stream_index = encoder->videoStream->index;
372 av_interleaved_write_frame(encoder->context, &packet);
373 }
374 av_free_packet(&packet);
375}