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