839b9b0df48b825ffcf558b076e3910e87fe375e
[libav.git] / libavformat / flvenc.c
1 /*
2 * FLV muxer
3 * Copyright (c) 2003 The Libav Project
4 *
5 * This file is part of Libav.
6 *
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 #include "libavutil/dict.h"
23 #include "libavutil/intfloat.h"
24 #include "avc.h"
25 #include "avformat.h"
26 #include "flv.h"
27 #include "internal.h"
28 #include "metadata.h"
29
30 #undef NDEBUG
31 #include <assert.h>
32
33 static const AVCodecTag flv_video_codec_ids[] = {
34 { AV_CODEC_ID_FLV1, FLV_CODECID_H263 },
35 { AV_CODEC_ID_FLASHSV, FLV_CODECID_SCREEN },
36 { AV_CODEC_ID_FLASHSV2, FLV_CODECID_SCREEN2 },
37 { AV_CODEC_ID_VP6F, FLV_CODECID_VP6 },
38 { AV_CODEC_ID_VP6, FLV_CODECID_VP6 },
39 { AV_CODEC_ID_H264, FLV_CODECID_H264 },
40 { AV_CODEC_ID_NONE, 0 }
41 };
42
43 static const AVCodecTag flv_audio_codec_ids[] = {
44 { AV_CODEC_ID_MP3, FLV_CODECID_MP3 >> FLV_AUDIO_CODECID_OFFSET },
45 { AV_CODEC_ID_PCM_U8, FLV_CODECID_PCM >> FLV_AUDIO_CODECID_OFFSET },
46 { AV_CODEC_ID_PCM_S16BE, FLV_CODECID_PCM >> FLV_AUDIO_CODECID_OFFSET },
47 { AV_CODEC_ID_PCM_S16LE, FLV_CODECID_PCM_LE >> FLV_AUDIO_CODECID_OFFSET },
48 { AV_CODEC_ID_ADPCM_SWF, FLV_CODECID_ADPCM >> FLV_AUDIO_CODECID_OFFSET },
49 { AV_CODEC_ID_AAC, FLV_CODECID_AAC >> FLV_AUDIO_CODECID_OFFSET },
50 { AV_CODEC_ID_NELLYMOSER, FLV_CODECID_NELLYMOSER >> FLV_AUDIO_CODECID_OFFSET },
51 { AV_CODEC_ID_PCM_MULAW, FLV_CODECID_PCM_MULAW >> FLV_AUDIO_CODECID_OFFSET },
52 { AV_CODEC_ID_PCM_ALAW, FLV_CODECID_PCM_ALAW >> FLV_AUDIO_CODECID_OFFSET },
53 { AV_CODEC_ID_SPEEX, FLV_CODECID_SPEEX >> FLV_AUDIO_CODECID_OFFSET },
54 { AV_CODEC_ID_NONE, 0 }
55 };
56
57 typedef struct FLVContext {
58 int reserved;
59 int64_t duration_offset;
60 int64_t filesize_offset;
61 int64_t duration;
62 int64_t delay; ///< first dts delay (needed for AVC & Speex)
63 } FLVContext;
64
65 typedef struct FLVStreamContext {
66 int64_t last_ts; ///< last timestamp for each stream
67 } FLVStreamContext;
68
69 static int get_audio_flags(AVFormatContext *s, AVCodecContext *enc)
70 {
71 int flags = (enc->bits_per_coded_sample == 16) ? FLV_SAMPLESSIZE_16BIT
72 : FLV_SAMPLESSIZE_8BIT;
73
74 if (enc->codec_id == AV_CODEC_ID_AAC) // specs force these parameters
75 return FLV_CODECID_AAC | FLV_SAMPLERATE_44100HZ |
76 FLV_SAMPLESSIZE_16BIT | FLV_STEREO;
77 else if (enc->codec_id == AV_CODEC_ID_SPEEX) {
78 if (enc->sample_rate != 16000) {
79 av_log(s, AV_LOG_ERROR,
80 "flv only supports wideband (16kHz) Speex audio\n");
81 return -1;
82 }
83 if (enc->channels != 1) {
84 av_log(s, AV_LOG_ERROR, "flv only supports mono Speex audio\n");
85 return -1;
86 }
87 return FLV_CODECID_SPEEX | FLV_SAMPLERATE_11025HZ | FLV_SAMPLESSIZE_16BIT;
88 } else {
89 switch (enc->sample_rate) {
90 case 44100:
91 flags |= FLV_SAMPLERATE_44100HZ;
92 break;
93 case 22050:
94 flags |= FLV_SAMPLERATE_22050HZ;
95 break;
96 case 11025:
97 flags |= FLV_SAMPLERATE_11025HZ;
98 break;
99 case 16000: // nellymoser only
100 case 8000: // nellymoser only
101 case 5512: // not MP3
102 if (enc->codec_id != AV_CODEC_ID_MP3) {
103 flags |= FLV_SAMPLERATE_SPECIAL;
104 break;
105 }
106 default:
107 av_log(s, AV_LOG_ERROR,
108 "flv does not support that sample rate, "
109 "choose from (44100, 22050, 11025).\n");
110 return -1;
111 }
112 }
113
114 if (enc->channels > 1)
115 flags |= FLV_STEREO;
116
117 switch (enc->codec_id) {
118 case AV_CODEC_ID_MP3:
119 flags |= FLV_CODECID_MP3 | FLV_SAMPLESSIZE_16BIT;
120 break;
121 case AV_CODEC_ID_PCM_U8:
122 flags |= FLV_CODECID_PCM | FLV_SAMPLESSIZE_8BIT;
123 break;
124 case AV_CODEC_ID_PCM_S16BE:
125 flags |= FLV_CODECID_PCM | FLV_SAMPLESSIZE_16BIT;
126 break;
127 case AV_CODEC_ID_PCM_S16LE:
128 flags |= FLV_CODECID_PCM_LE | FLV_SAMPLESSIZE_16BIT;
129 break;
130 case AV_CODEC_ID_ADPCM_SWF:
131 flags |= FLV_CODECID_ADPCM | FLV_SAMPLESSIZE_16BIT;
132 break;
133 case AV_CODEC_ID_NELLYMOSER:
134 if (enc->sample_rate == 8000)
135 flags |= FLV_CODECID_NELLYMOSER_8KHZ_MONO | FLV_SAMPLESSIZE_16BIT;
136 else if (enc->sample_rate == 16000)
137 flags |= FLV_CODECID_NELLYMOSER_16KHZ_MONO | FLV_SAMPLESSIZE_16BIT;
138 else
139 flags |= FLV_CODECID_NELLYMOSER | FLV_SAMPLESSIZE_16BIT;
140 break;
141 case AV_CODEC_ID_PCM_MULAW:
142 flags = FLV_CODECID_PCM_MULAW | FLV_SAMPLERATE_SPECIAL | FLV_SAMPLESSIZE_16BIT;
143 break;
144 case AV_CODEC_ID_PCM_ALAW:
145 flags = FLV_CODECID_PCM_ALAW | FLV_SAMPLERATE_SPECIAL | FLV_SAMPLESSIZE_16BIT;
146 break;
147 case 0:
148 flags |= enc->codec_tag << 4;
149 break;
150 default:
151 av_log(s, AV_LOG_ERROR, "codec not compatible with flv\n");
152 return -1;
153 }
154
155 return flags;
156 }
157
158 static void put_amf_string(AVIOContext *pb, const char *str)
159 {
160 size_t len = strlen(str);
161 avio_wb16(pb, len);
162 avio_write(pb, str, len);
163 }
164
165 static void put_avc_eos_tag(AVIOContext *pb, unsigned ts)
166 {
167 avio_w8(pb, FLV_TAG_TYPE_VIDEO);
168 avio_wb24(pb, 5); /* Tag Data Size */
169 avio_wb24(pb, ts); /* lower 24 bits of timestamp in ms */
170 avio_w8(pb, (ts >> 24) & 0x7F); /* MSB of ts in ms */
171 avio_wb24(pb, 0); /* StreamId = 0 */
172 avio_w8(pb, 23); /* ub[4] FrameType = 1, ub[4] CodecId = 7 */
173 avio_w8(pb, 2); /* AVC end of sequence */
174 avio_wb24(pb, 0); /* Always 0 for AVC EOS. */
175 avio_wb32(pb, 16); /* Size of FLV tag */
176 }
177
178 static void put_amf_double(AVIOContext *pb, double d)
179 {
180 avio_w8(pb, AMF_DATA_TYPE_NUMBER);
181 avio_wb64(pb, av_double2int(d));
182 }
183
184 static void put_amf_bool(AVIOContext *pb, int b)
185 {
186 avio_w8(pb, AMF_DATA_TYPE_BOOL);
187 avio_w8(pb, !!b);
188 }
189
190 static int flv_write_header(AVFormatContext *s)
191 {
192 AVIOContext *pb = s->pb;
193 FLVContext *flv = s->priv_data;
194 AVCodecContext *audio_enc = NULL, *video_enc = NULL, *data_enc = NULL;
195 int i, metadata_count = 0;
196 double framerate = 0.0;
197 int64_t metadata_size_pos, data_size, metadata_count_pos;
198 AVDictionaryEntry *tag = NULL;
199
200 for (i = 0; i < s->nb_streams; i++) {
201 AVCodecContext *enc = s->streams[i]->codec;
202 FLVStreamContext *sc;
203 switch (enc->codec_type) {
204 case AVMEDIA_TYPE_VIDEO:
205 if (s->streams[i]->avg_frame_rate.den &&
206 s->streams[i]->avg_frame_rate.num) {
207 framerate = av_q2d(s->streams[i]->avg_frame_rate);
208 } else {
209 framerate = 1 / av_q2d(s->streams[i]->codec->time_base);
210 }
211 video_enc = enc;
212 if (enc->codec_tag == 0) {
213 av_log(s, AV_LOG_ERROR, "video codec not compatible with flv\n");
214 return -1;
215 }
216 break;
217 case AVMEDIA_TYPE_AUDIO:
218 audio_enc = enc;
219 if (get_audio_flags(s, enc) < 0)
220 return AVERROR_INVALIDDATA;
221 break;
222 case AVMEDIA_TYPE_DATA:
223 if (enc->codec_id != AV_CODEC_ID_TEXT) {
224 av_log(s, AV_LOG_ERROR, "codec not compatible with flv\n");
225 return AVERROR_INVALIDDATA;
226 }
227 data_enc = enc;
228 break;
229 default:
230 av_log(s, AV_LOG_ERROR, "codec not compatible with flv\n");
231 return -1;
232 }
233 avpriv_set_pts_info(s->streams[i], 32, 1, 1000); /* 32 bit pts in ms */
234
235 sc = av_mallocz(sizeof(FLVStreamContext));
236 if (!sc)
237 return AVERROR(ENOMEM);
238 s->streams[i]->priv_data = sc;
239 sc->last_ts = -1;
240 }
241
242 flv->delay = AV_NOPTS_VALUE;
243
244 avio_write(pb, "FLV", 3);
245 avio_w8(pb, 1);
246 avio_w8(pb, FLV_HEADER_FLAG_HASAUDIO * !!audio_enc +
247 FLV_HEADER_FLAG_HASVIDEO * !!video_enc);
248 avio_wb32(pb, 9);
249 avio_wb32(pb, 0);
250
251 for (i = 0; i < s->nb_streams; i++)
252 if (s->streams[i]->codec->codec_tag == 5) {
253 avio_w8(pb, 8); // message type
254 avio_wb24(pb, 0); // include flags
255 avio_wb24(pb, 0); // time stamp
256 avio_wb32(pb, 0); // reserved
257 avio_wb32(pb, 11); // size
258 flv->reserved = 5;
259 }
260
261 /* write meta_tag */
262 avio_w8(pb, 18); // tag type META
263 metadata_size_pos = avio_tell(pb);
264 avio_wb24(pb, 0); // size of data part (sum of all parts below)
265 avio_wb24(pb, 0); // timestamp
266 avio_wb32(pb, 0); // reserved
267
268 /* now data of data_size size */
269
270 /* first event name as a string */
271 avio_w8(pb, AMF_DATA_TYPE_STRING);
272 put_amf_string(pb, "onMetaData"); // 12 bytes
273
274 /* mixed array (hash) with size and string/type/data tuples */
275 avio_w8(pb, AMF_DATA_TYPE_MIXEDARRAY);
276 metadata_count_pos = avio_tell(pb);
277 metadata_count = 5 * !!video_enc +
278 5 * !!audio_enc +
279 1 * !!data_enc +
280 2; // +2 for duration and file size
281
282 avio_wb32(pb, metadata_count);
283
284 put_amf_string(pb, "duration");
285 flv->duration_offset= avio_tell(pb);
286
287 // fill in the guessed duration, it'll be corrected later if incorrect
288 put_amf_double(pb, s->duration / AV_TIME_BASE);
289
290 if (video_enc) {
291 put_amf_string(pb, "width");
292 put_amf_double(pb, video_enc->width);
293
294 put_amf_string(pb, "height");
295 put_amf_double(pb, video_enc->height);
296
297 put_amf_string(pb, "videodatarate");
298 put_amf_double(pb, video_enc->bit_rate / 1024.0);
299
300 put_amf_string(pb, "framerate");
301 put_amf_double(pb, framerate);
302
303 put_amf_string(pb, "videocodecid");
304 put_amf_double(pb, video_enc->codec_tag);
305 }
306
307 if (audio_enc) {
308 put_amf_string(pb, "audiodatarate");
309 put_amf_double(pb, audio_enc->bit_rate / 1024.0);
310
311 put_amf_string(pb, "audiosamplerate");
312 put_amf_double(pb, audio_enc->sample_rate);
313
314 put_amf_string(pb, "audiosamplesize");
315 put_amf_double(pb, audio_enc->codec_id == AV_CODEC_ID_PCM_U8 ? 8 : 16);
316
317 put_amf_string(pb, "stereo");
318 put_amf_bool(pb, audio_enc->channels == 2);
319
320 put_amf_string(pb, "audiocodecid");
321 put_amf_double(pb, audio_enc->codec_tag);
322 }
323
324 if (data_enc) {
325 put_amf_string(pb, "datastream");
326 put_amf_double(pb, 0.0);
327 }
328
329 while ((tag = av_dict_get(s->metadata, "", tag, AV_DICT_IGNORE_SUFFIX))) {
330 put_amf_string(pb, tag->key);
331 avio_w8(pb, AMF_DATA_TYPE_STRING);
332 put_amf_string(pb, tag->value);
333 metadata_count++;
334 }
335
336 put_amf_string(pb, "filesize");
337 flv->filesize_offset = avio_tell(pb);
338 put_amf_double(pb, 0); // delayed write
339
340 put_amf_string(pb, "");
341 avio_w8(pb, AMF_END_OF_OBJECT);
342
343 /* write total size of tag */
344 data_size = avio_tell(pb) - metadata_size_pos - 10;
345
346 avio_seek(pb, metadata_count_pos, SEEK_SET);
347 avio_wb32(pb, metadata_count);
348
349 avio_seek(pb, metadata_size_pos, SEEK_SET);
350 avio_wb24(pb, data_size);
351 avio_skip(pb, data_size + 10 - 3);
352 avio_wb32(pb, data_size + 11);
353
354 for (i = 0; i < s->nb_streams; i++) {
355 AVCodecContext *enc = s->streams[i]->codec;
356 if (enc->codec_id == AV_CODEC_ID_AAC || enc->codec_id == AV_CODEC_ID_H264) {
357 int64_t pos;
358 avio_w8(pb, enc->codec_type == AVMEDIA_TYPE_VIDEO ?
359 FLV_TAG_TYPE_VIDEO : FLV_TAG_TYPE_AUDIO);
360 avio_wb24(pb, 0); // size patched later
361 avio_wb24(pb, 0); // ts
362 avio_w8(pb, 0); // ts ext
363 avio_wb24(pb, 0); // streamid
364 pos = avio_tell(pb);
365 if (enc->codec_id == AV_CODEC_ID_AAC) {
366 avio_w8(pb, get_audio_flags(s, enc));
367 avio_w8(pb, 0); // AAC sequence header
368 avio_write(pb, enc->extradata, enc->extradata_size);
369 } else {
370 avio_w8(pb, enc->codec_tag | FLV_FRAME_KEY); // flags
371 avio_w8(pb, 0); // AVC sequence header
372 avio_wb24(pb, 0); // composition time
373 ff_isom_write_avcc(pb, enc->extradata, enc->extradata_size);
374 }
375 data_size = avio_tell(pb) - pos;
376 avio_seek(pb, -data_size - 10, SEEK_CUR);
377 avio_wb24(pb, data_size);
378 avio_skip(pb, data_size + 10 - 3);
379 avio_wb32(pb, data_size + 11); // previous tag size
380 }
381 }
382
383 return 0;
384 }
385
386 static int flv_write_trailer(AVFormatContext *s)
387 {
388 int64_t file_size;
389
390 AVIOContext *pb = s->pb;
391 FLVContext *flv = s->priv_data;
392 int i;
393
394 /* Add EOS tag */
395 for (i = 0; i < s->nb_streams; i++) {
396 AVCodecContext *enc = s->streams[i]->codec;
397 FLVStreamContext *sc = s->streams[i]->priv_data;
398 if (enc->codec_type == AVMEDIA_TYPE_VIDEO &&
399 enc->codec_id == AV_CODEC_ID_H264)
400 put_avc_eos_tag(pb, sc->last_ts);
401 }
402
403 file_size = avio_tell(pb);
404
405 /* update information */
406 avio_seek(pb, flv->duration_offset, SEEK_SET);
407 put_amf_double(pb, flv->duration / (double)1000);
408 avio_seek(pb, flv->filesize_offset, SEEK_SET);
409 put_amf_double(pb, file_size);
410
411 avio_seek(pb, file_size, SEEK_SET);
412 return 0;
413 }
414
415 static int flv_write_packet(AVFormatContext *s, AVPacket *pkt)
416 {
417 AVIOContext *pb = s->pb;
418 AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
419 FLVContext *flv = s->priv_data;
420 FLVStreamContext *sc = s->streams[pkt->stream_index]->priv_data;
421 unsigned ts;
422 int size = pkt->size;
423 uint8_t *data = NULL;
424 int flags = 0, flags_size;
425
426 // av_log(s, AV_LOG_DEBUG, "type:%d pts: %"PRId64" size:%d\n",
427 // enc->codec_type, timestamp, size);
428
429 if (enc->codec_id == AV_CODEC_ID_VP6 || enc->codec_id == AV_CODEC_ID_VP6F ||
430 enc->codec_id == AV_CODEC_ID_AAC)
431 flags_size = 2;
432 else if (enc->codec_id == AV_CODEC_ID_H264)
433 flags_size = 5;
434 else
435 flags_size = 1;
436
437 switch (enc->codec_type) {
438 case AVMEDIA_TYPE_VIDEO:
439 avio_w8(pb, FLV_TAG_TYPE_VIDEO);
440
441 flags = enc->codec_tag;
442 if (flags == 0) {
443 av_log(s, AV_LOG_ERROR,
444 "video codec %X not compatible with flv\n",
445 enc->codec_id);
446 return -1;
447 }
448
449 flags |= pkt->flags & AV_PKT_FLAG_KEY ? FLV_FRAME_KEY : FLV_FRAME_INTER;
450 break;
451 case AVMEDIA_TYPE_AUDIO:
452 flags = get_audio_flags(s, enc);
453
454 assert(size);
455
456 avio_w8(pb, FLV_TAG_TYPE_AUDIO);
457 break;
458 case AVMEDIA_TYPE_DATA:
459 avio_w8(pb, FLV_TAG_TYPE_META);
460 break;
461 default:
462 return AVERROR(EINVAL);
463 }
464
465 if (enc->codec_id == AV_CODEC_ID_H264)
466 /* check if extradata looks like MP4 */
467 if (enc->extradata_size > 0 && *(uint8_t*)enc->extradata != 1)
468 if (ff_avc_parse_nal_units_buf(pkt->data, &data, &size) < 0)
469 return -1;
470
471 if (flv->delay == AV_NOPTS_VALUE)
472 flv->delay = -pkt->dts;
473
474 if (pkt->dts < -flv->delay) {
475 av_log(s, AV_LOG_WARNING,
476 "Packets are not in the proper order with respect to DTS\n");
477 return AVERROR(EINVAL);
478 }
479
480 ts = pkt->dts + flv->delay; // add delay to force positive dts
481
482 /* check Speex packet duration */
483 if (enc->codec_id == AV_CODEC_ID_SPEEX && ts - sc->last_ts > 160)
484 av_log(s, AV_LOG_WARNING, "Warning: Speex stream has more than "
485 "8 frames per packet. Adobe Flash "
486 "Player cannot handle this!\n");
487
488 if (sc->last_ts < ts)
489 sc->last_ts = ts;
490
491 avio_wb24(pb, size + flags_size);
492 avio_wb24(pb, ts);
493 avio_w8(pb, (ts >> 24) & 0x7F); // timestamps are 32 bits _signed_
494 avio_wb24(pb, flv->reserved);
495
496 if (enc->codec_type == AVMEDIA_TYPE_DATA) {
497 int data_size;
498 int metadata_size_pos = avio_tell(pb);
499 avio_w8(pb, AMF_DATA_TYPE_STRING);
500 put_amf_string(pb, "onTextData");
501 avio_w8(pb, AMF_DATA_TYPE_MIXEDARRAY);
502 avio_wb32(pb, 2);
503 put_amf_string(pb, "type");
504 avio_w8(pb, AMF_DATA_TYPE_STRING);
505 put_amf_string(pb, "Text");
506 put_amf_string(pb, "text");
507 avio_w8(pb, AMF_DATA_TYPE_STRING);
508 put_amf_string(pb, pkt->data);
509 put_amf_string(pb, "");
510 avio_w8(pb, AMF_END_OF_OBJECT);
511 /* write total size of tag */
512 data_size = avio_tell(pb) - metadata_size_pos;
513 avio_seek(pb, metadata_size_pos - 10, SEEK_SET);
514 avio_wb24(pb, data_size);
515 avio_seek(pb, data_size + 10 - 3, SEEK_CUR);
516 avio_wb32(pb, data_size + 11);
517 } else {
518 avio_w8(pb,flags);
519 if (enc->codec_id == AV_CODEC_ID_VP6)
520 avio_w8(pb, 0);
521 if (enc->codec_id == AV_CODEC_ID_VP6F)
522 avio_w8(pb, enc->extradata_size ? enc->extradata[0] : 0);
523 else if (enc->codec_id == AV_CODEC_ID_AAC)
524 avio_w8(pb, 1); // AAC raw
525 else if (enc->codec_id == AV_CODEC_ID_H264) {
526 avio_w8(pb, 1); // AVC NALU
527 avio_wb24(pb, pkt->pts - pkt->dts);
528 }
529
530 avio_write(pb, data ? data : pkt->data, size);
531
532 avio_wb32(pb, size + flags_size + 11); // previous tag size
533 flv->duration = FFMAX(flv->duration,
534 pkt->pts + flv->delay + pkt->duration);
535 }
536
537 avio_flush(pb);
538 av_free(data);
539
540 return pb->error;
541 }
542
543 AVOutputFormat ff_flv_muxer = {
544 .name = "flv",
545 .long_name = NULL_IF_CONFIG_SMALL("FLV (Flash Video)"),
546 .mime_type = "video/x-flv",
547 .extensions = "flv",
548 .priv_data_size = sizeof(FLVContext),
549 .audio_codec = CONFIG_LIBMP3LAME ? AV_CODEC_ID_MP3 : AV_CODEC_ID_ADPCM_SWF,
550 .video_codec = AV_CODEC_ID_FLV1,
551 .write_header = flv_write_header,
552 .write_packet = flv_write_packet,
553 .write_trailer = flv_write_trailer,
554 .codec_tag = (const AVCodecTag* const []) {
555 flv_video_codec_ids, flv_audio_codec_ids, 0
556 },
557 .flags = AVFMT_GLOBALHEADER | AVFMT_VARIABLE_FPS |
558 AVFMT_TS_NONSTRICT,
559 };