src/platform/ffmpeg/ffmpeg-encoder.c (view raw)
1#include "ffmpeg-encoder.h"
2
3#include "gba-video.h"
4
5#include <libavutil/imgutils.h>
6#include <libavutil/opt.h>
7
8static void _ffmpegPostVideoFrame(struct GBAAVStream*, struct GBAVideoRenderer* renderer);
9static void _ffmpegPostAudioFrame(struct GBAAVStream*, int32_t left, int32_t right);
10
11enum {
12 PREFERRED_SAMPLE_RATE = 0x8000
13};
14
15void FFmpegEncoderInit(struct FFmpegEncoder* encoder) {
16 av_register_all();
17
18 encoder->d.postVideoFrame = _ffmpegPostVideoFrame;
19 encoder->d.postAudioFrame = _ffmpegPostAudioFrame;
20
21 encoder->audioCodec = 0;
22 encoder->videoCodec = 0;
23 encoder->containerFormat = 0;
24 FFmpegEncoderSetAudio(encoder, "flac", 0);
25 FFmpegEncoderSetVideo(encoder, "png", 0);
26 FFmpegEncoderSetContainer(encoder, "matroska");
27 encoder->currentAudioSample = 0;
28 encoder->currentAudioFrame = 0;
29 encoder->currentVideoFrame = 0;
30 encoder->context = 0;
31}
32
33bool FFmpegEncoderSetAudio(struct FFmpegEncoder* encoder, const char* acodec, unsigned abr) {
34 static const struct {
35 int format;
36 int priority;
37 } priorities[] = {
38 { AV_SAMPLE_FMT_S16, 0 },
39 { AV_SAMPLE_FMT_S16P, 1 },
40 { AV_SAMPLE_FMT_S32, 2 },
41 { AV_SAMPLE_FMT_S32P, 2 },
42 { AV_SAMPLE_FMT_FLT, 3 },
43 { AV_SAMPLE_FMT_FLTP, 3 },
44 { AV_SAMPLE_FMT_DBL, 4 },
45 { AV_SAMPLE_FMT_DBLP, 4 }
46 };
47 AVCodec* codec = avcodec_find_encoder_by_name(acodec);
48 if (!codec) {
49 return false;
50 }
51
52 if (!codec->sample_fmts) {
53 return false;
54 }
55 size_t i;
56 size_t j;
57 int priority = INT_MAX;
58 encoder->sampleFormat = AV_SAMPLE_FMT_NONE;
59 for (i = 0; codec->sample_fmts[i] != AV_SAMPLE_FMT_NONE; ++i) {
60 for (j = 0; j < sizeof(priorities) / sizeof(*priorities); ++j) {
61 if (codec->sample_fmts[i] == priorities[j].format && priority > priorities[j].priority) {
62 priority = priorities[j].priority;
63 encoder->sampleFormat = codec->sample_fmts[i];
64 }
65 }
66 }
67 if (encoder->sampleFormat == AV_SAMPLE_FMT_NONE) {
68 return false;
69 }
70 encoder->sampleRate = PREFERRED_SAMPLE_RATE;
71 if (codec->supported_samplerates) {
72 for (i = 0; codec->supported_samplerates[i]; ++i) {
73 if (codec->supported_samplerates[i] < PREFERRED_SAMPLE_RATE) {
74 continue;
75 }
76 if (encoder->sampleRate == PREFERRED_SAMPLE_RATE || encoder->sampleRate > codec->supported_samplerates[i]) {
77 encoder->sampleRate = codec->supported_samplerates[i];
78 }
79 }
80 } else if (codec->id == AV_CODEC_ID_AAC) {
81 // HACK: AAC doesn't support 32768Hz (it rounds to 32000), but libfaac doesn't tell us that
82 encoder->sampleRate = 44100;
83 }
84 encoder->audioCodec = acodec;
85 encoder->audioBitrate = abr;
86 return true;
87}
88
89bool FFmpegEncoderSetVideo(struct FFmpegEncoder* encoder, const char* vcodec, unsigned vbr) {
90 static const struct {
91 enum AVPixelFormat format;
92 int priority;
93 } priorities[] = {
94 { AV_PIX_FMT_RGB24, 0 },
95 { AV_PIX_FMT_BGR0, 1 },
96 { AV_PIX_FMT_YUV422P, 2 },
97 { AV_PIX_FMT_YUV444P, 3 },
98 { AV_PIX_FMT_YUV420P, 4 }
99 };
100 AVCodec* codec = avcodec_find_encoder_by_name(vcodec);
101 if (!codec) {
102 return false;
103 }
104
105 size_t i;
106 size_t j;
107 int priority = INT_MAX;
108 encoder->pixFormat = AV_PIX_FMT_NONE;
109 for (i = 0; codec->pix_fmts[i] != AV_PIX_FMT_NONE; ++i) {
110 for (j = 0; j < sizeof(priorities) / sizeof(*priorities); ++j) {
111 if (codec->pix_fmts[i] == priorities[j].format && priority > priorities[j].priority) {
112 priority = priorities[j].priority;
113 encoder->pixFormat = codec->pix_fmts[i];
114 }
115 }
116 }
117 if (encoder->pixFormat == AV_PIX_FMT_NONE) {
118 return false;
119 }
120 encoder->videoCodec = vcodec;
121 encoder->videoBitrate = vbr;
122 return true;
123}
124
125bool FFmpegEncoderSetContainer(struct FFmpegEncoder* encoder, const char* container) {
126 AVOutputFormat* oformat = av_guess_format(container, 0, 0);
127 if (!oformat) {
128 return false;
129 }
130 encoder->containerFormat = container;
131 return true;
132}
133
134bool FFmpegEncoderVerifyContainer(struct FFmpegEncoder* encoder) {
135 AVOutputFormat* oformat = av_guess_format(encoder->containerFormat, 0, 0);
136 AVCodec* acodec = avcodec_find_encoder_by_name(encoder->audioCodec);
137 AVCodec* vcodec = avcodec_find_encoder_by_name(encoder->videoCodec);
138 if (!acodec || !vcodec || !oformat) {
139 return false;
140 }
141 if (!avformat_query_codec(oformat, acodec->id, FF_COMPLIANCE_EXPERIMENTAL)) {
142 return false;
143 }
144 if (!avformat_query_codec(oformat, vcodec->id, FF_COMPLIANCE_EXPERIMENTAL)) {
145 return false;
146 }
147 return true;
148}
149
150bool FFmpegEncoderOpen(struct FFmpegEncoder* encoder, const char* outfile) {
151 AVCodec* acodec = avcodec_find_encoder_by_name(encoder->audioCodec);
152 AVCodec* vcodec = avcodec_find_encoder_by_name(encoder->videoCodec);
153 if (!acodec || !vcodec || !FFmpegEncoderVerifyContainer(encoder)) {
154 return false;
155 }
156
157 avformat_alloc_output_context2(&encoder->context, 0, 0, outfile);
158
159 encoder->context->oformat = av_guess_format(encoder->containerFormat, 0, 0);
160
161 encoder->audioStream = avformat_new_stream(encoder->context, acodec);
162 encoder->audio = encoder->audioStream->codec;
163 encoder->audio->bit_rate = encoder->audioBitrate;
164 encoder->audio->channels = 2;
165 encoder->audio->channel_layout = AV_CH_LAYOUT_STEREO;
166 encoder->audio->sample_rate = encoder->sampleRate;
167 encoder->audio->sample_fmt = encoder->sampleFormat;
168 AVDictionary* opts = 0;
169 av_dict_set_int(&opts, "strict", FF_COMPLIANCE_EXPERIMENTAL, 0);
170 avcodec_open2(encoder->audio, acodec, &opts);
171 av_dict_free(&opts);
172 encoder->audioFrame = av_frame_alloc();
173 encoder->audioFrame->nb_samples = encoder->audio->frame_size;
174 encoder->audioFrame->format = encoder->audio->sample_fmt;
175 encoder->audioFrame->pts = 0;
176 encoder->resampleContext = avresample_alloc_context();
177 av_opt_set_int(encoder->resampleContext, "in_channel_layout", AV_CH_LAYOUT_STEREO, 0);
178 av_opt_set_int(encoder->resampleContext, "out_channel_layout", AV_CH_LAYOUT_STEREO, 0);
179 av_opt_set_int(encoder->resampleContext, "in_sample_rate", PREFERRED_SAMPLE_RATE, 0);
180 av_opt_set_int(encoder->resampleContext, "out_sample_rate", encoder->sampleRate, 0);
181 av_opt_set_int(encoder->resampleContext, "in_sample_fmt", AV_SAMPLE_FMT_S16, 0);
182 av_opt_set_int(encoder->resampleContext, "out_sample_fmt", encoder->sampleFormat, 0);
183 avresample_open(encoder->resampleContext);
184 encoder->audioBufferSize = (encoder->audioFrame->nb_samples * PREFERRED_SAMPLE_RATE / encoder->sampleRate) * 4;
185 encoder->audioBuffer = av_malloc(encoder->audioBufferSize);
186 encoder->postaudioBufferSize = av_samples_get_buffer_size(0, encoder->audio->channels, encoder->audio->frame_size, encoder->audio->sample_fmt, 0);
187 encoder->postaudioBuffer = av_malloc(encoder->postaudioBufferSize);
188 avcodec_fill_audio_frame(encoder->audioFrame, encoder->audio->channels, encoder->audio->sample_fmt, (const uint8_t*) encoder->postaudioBuffer, encoder->postaudioBufferSize, 0);
189
190 encoder->videoStream = avformat_new_stream(encoder->context, vcodec);
191 encoder->video = encoder->videoStream->codec;
192 encoder->video->bit_rate = encoder->videoBitrate;
193 encoder->video->width = VIDEO_HORIZONTAL_PIXELS;
194 encoder->video->height = VIDEO_VERTICAL_PIXELS;
195 encoder->video->time_base = (AVRational) { VIDEO_TOTAL_LENGTH, GBA_ARM7TDMI_FREQUENCY };
196 encoder->video->pix_fmt = encoder->pixFormat;
197 encoder->video->gop_size = 15;
198 encoder->video->max_b_frames = 0;
199 avcodec_open2(encoder->video, vcodec, 0);
200 encoder->videoFrame = av_frame_alloc();
201 encoder->videoFrame->format = encoder->video->pix_fmt;
202 encoder->videoFrame->width = encoder->video->width;
203 encoder->videoFrame->height = encoder->video->height;
204 encoder->videoFrame->pts = 0;
205 av_image_alloc(encoder->videoFrame->data, encoder->videoFrame->linesize, encoder->video->width, encoder->video->height, encoder->video->pix_fmt, 32);
206
207 if (encoder->context->oformat->flags & AVFMT_GLOBALHEADER) {
208 encoder->audio->flags |= CODEC_FLAG_GLOBAL_HEADER;
209 encoder->video->flags |= CODEC_FLAG_GLOBAL_HEADER;
210 }
211
212 avio_open(&encoder->context->pb, outfile, AVIO_FLAG_WRITE);
213 avformat_write_header(encoder->context, 0);
214
215 return true;
216}
217
218void FFmpegEncoderClose(struct FFmpegEncoder* encoder) {
219 if (!encoder->context) {
220 return;
221 }
222 av_write_trailer(encoder->context);
223 avio_close(encoder->context->pb);
224
225 av_free(encoder->postaudioBuffer);
226 if (encoder->audioBuffer) {
227 av_free(encoder->audioBuffer);
228 }
229 av_frame_free(&encoder->audioFrame);
230 avcodec_close(encoder->audio);
231
232 av_frame_free(&encoder->videoFrame);
233 avcodec_close(encoder->video);
234
235 if (encoder->resampleContext) {
236 avresample_close(encoder->resampleContext);
237 }
238
239 avformat_free_context(encoder->context);
240 encoder->context = 0;
241
242 encoder->currentAudioSample = 0;
243 encoder->currentAudioFrame = 0;
244 encoder->nextAudioPts = 0;
245 encoder->currentVideoFrame = 0;
246}
247
248void _ffmpegPostAudioFrame(struct GBAAVStream* stream, int32_t left, int32_t right) {
249 struct FFmpegEncoder* encoder = (struct FFmpegEncoder*) stream;
250 if (!encoder->context) {
251 return;
252 }
253
254 av_frame_make_writable(encoder->audioFrame);
255 encoder->audioBuffer[encoder->currentAudioSample * 2] = left;
256 encoder->audioBuffer[encoder->currentAudioSample * 2 + 1] = right;
257
258 ++encoder->currentAudioFrame;
259 ++encoder->currentAudioSample;
260
261 if ((encoder->currentAudioSample * 4) < encoder->audioBufferSize) {
262 return;
263 }
264 encoder->currentAudioSample = 0;
265
266 int channelSize = 2 * av_get_bytes_per_sample(encoder->audio->sample_fmt);
267 avresample_convert(encoder->resampleContext,
268 0, 0, encoder->postaudioBufferSize / channelSize,
269 (uint8_t**) &encoder->audioBuffer, 0, encoder->audioBufferSize / 4);
270 if ((ssize_t) avresample_available(encoder->resampleContext) < (ssize_t) encoder->postaudioBufferSize / channelSize) {
271 return;
272 }
273 avresample_read(encoder->resampleContext, encoder->audioFrame->data, encoder->postaudioBufferSize / channelSize);
274
275 AVRational timeBase = { 1, PREFERRED_SAMPLE_RATE };
276 encoder->audioFrame->pts = encoder->nextAudioPts;
277 encoder->nextAudioPts = av_rescale_q(encoder->currentAudioFrame, timeBase, encoder->audioStream->time_base);
278
279 AVPacket packet;
280 av_init_packet(&packet);
281 packet.data = 0;
282 packet.size = 0;
283 int gotData;
284 avcodec_encode_audio2(encoder->audio, &packet, encoder->audioFrame, &gotData);
285 if (gotData) {
286 packet.stream_index = encoder->audioStream->index;
287 av_interleaved_write_frame(encoder->context, &packet);
288 }
289 av_free_packet(&packet);
290}
291
292void _ffmpegPostVideoFrame(struct GBAAVStream* stream, struct GBAVideoRenderer* renderer) {
293 struct FFmpegEncoder* encoder = (struct FFmpegEncoder*) stream;
294 if (!encoder->context) {
295 return;
296 }
297 uint32_t* pixels;
298 unsigned stride;
299 renderer->getPixels(renderer, &stride, (void**) &pixels);
300
301 AVPacket packet;
302
303 av_init_packet(&packet);
304 packet.data = 0;
305 packet.size = 0;
306 av_frame_make_writable(encoder->videoFrame);
307 encoder->videoFrame->pts = av_rescale_q(encoder->currentVideoFrame, encoder->video->time_base, encoder->videoStream->time_base);
308 ++encoder->currentVideoFrame;
309
310 unsigned x, y;
311 if (encoder->videoFrame->format == AV_PIX_FMT_BGR0) {
312 for (y = 0; y < VIDEO_VERTICAL_PIXELS; ++y) {
313 for (x = 0; x < VIDEO_HORIZONTAL_PIXELS; ++x) {
314 uint32_t pixel = pixels[stride * y + x];
315 encoder->videoFrame->data[0][y * encoder->videoFrame->linesize[0] + x * 4] = pixel >> 16;
316 encoder->videoFrame->data[0][y * encoder->videoFrame->linesize[0] + x * 4 + 1] = pixel >> 8;
317 encoder->videoFrame->data[0][y * encoder->videoFrame->linesize[0] + x * 4 + 2] = pixel;
318 }
319 }
320 } else if (encoder->videoFrame->format == AV_PIX_FMT_RGB24) {
321 for (y = 0; y < VIDEO_VERTICAL_PIXELS; ++y) {
322 for (x = 0; x < VIDEO_HORIZONTAL_PIXELS; ++x) {
323 uint32_t pixel = pixels[stride * y + x];
324 encoder->videoFrame->data[0][y * encoder->videoFrame->linesize[0] + x * 3] = pixel;
325 encoder->videoFrame->data[0][y * encoder->videoFrame->linesize[0] + x * 3 + 1] = pixel >> 8;
326 encoder->videoFrame->data[0][y * encoder->videoFrame->linesize[0] + x * 3 + 2] = pixel >> 16;
327 }
328 }
329 }
330
331 int gotData;
332 avcodec_encode_video2(encoder->video, &packet, encoder->videoFrame, &gotData);
333 if (gotData) {
334 if (encoder->videoStream->codec->coded_frame->key_frame) {
335 packet.flags |= AV_PKT_FLAG_KEY;
336 }
337 packet.stream_index = encoder->videoStream->index;
338 av_interleaved_write_frame(encoder->context, &packet);
339 }
340 av_free_packet(&packet);
341}