FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
flvenc.c
Go to the documentation of this file.
1 /*
2  * FLV muxer
3  * Copyright (c) 2003 The FFmpeg Project
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg 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  * FFmpeg 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 FFmpeg; 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/intreadwrite.h"
23 #include "libavutil/dict.h"
24 #include "libavutil/intfloat.h"
25 #include "libavutil/avassert.h"
26 #include "avc.h"
27 #include "avformat.h"
28 #include "flv.h"
29 #include "internal.h"
30 #include "metadata.h"
31 
32 
33 static const AVCodecTag flv_video_codec_ids[] = {
43  { AV_CODEC_ID_NONE, 0 }
44 };
45 
46 static const AVCodecTag flv_audio_codec_ids[] = {
57  { AV_CODEC_ID_NONE, 0 }
58 };
59 
60 typedef struct FLVContext {
61  int reserved;
62  int64_t duration_offset;
63  int64_t filesize_offset;
64  int64_t duration;
65  int64_t delay; ///< first dts delay (needed for AVC & Speex)
66 } FLVContext;
67 
68 typedef struct FLVStreamContext {
69  int64_t last_ts; ///< last timestamp for each stream
71 
73 {
76 
77  if (enc->codec_id == AV_CODEC_ID_AAC) // specs force these parameters
80  else if (enc->codec_id == AV_CODEC_ID_SPEEX) {
81  if (enc->sample_rate != 16000) {
83  "FLV only supports wideband (16kHz) Speex audio\n");
84  return AVERROR(EINVAL);
85  }
86  if (enc->channels != 1) {
87  av_log(s, AV_LOG_ERROR, "FLV only supports mono Speex audio\n");
88  return AVERROR(EINVAL);
89  }
91  } else {
92  switch (enc->sample_rate) {
93  case 44100:
94  flags |= FLV_SAMPLERATE_44100HZ;
95  break;
96  case 22050:
97  flags |= FLV_SAMPLERATE_22050HZ;
98  break;
99  case 11025:
100  flags |= FLV_SAMPLERATE_11025HZ;
101  break;
102  case 16000: // nellymoser only
103  case 8000: // nellymoser only
104  case 5512: // not MP3
105  if (enc->codec_id != AV_CODEC_ID_MP3) {
106  flags |= FLV_SAMPLERATE_SPECIAL;
107  break;
108  }
109  default:
110  av_log(s, AV_LOG_ERROR,
111  "FLV does not support sample rate %d, "
112  "choose from (44100, 22050, 11025)\n", enc->sample_rate);
113  return AVERROR(EINVAL);
114  }
115  }
116 
117  if (enc->channels > 1)
118  flags |= FLV_STEREO;
119 
120  switch (enc->codec_id) {
121  case AV_CODEC_ID_MP3:
123  break;
124  case AV_CODEC_ID_PCM_U8:
126  break;
129  break;
132  break;
135  break;
137  if (enc->sample_rate == 8000)
139  else if (enc->sample_rate == 16000)
141  else
143  break;
146  break;
149  break;
150  case 0:
151  flags |= enc->codec_tag << 4;
152  break;
153  default:
154  av_log(s, AV_LOG_ERROR, "Audio codec '%s' not compatible with FLV\n",
155  avcodec_get_name(enc->codec_id));
156  return AVERROR(EINVAL);
157  }
158 
159  return flags;
160 }
161 
162 static void put_amf_string(AVIOContext *pb, const char *str)
163 {
164  size_t len = strlen(str);
165  avio_wb16(pb, len);
166  avio_write(pb, str, len);
167 }
168 
169 static void put_avc_eos_tag(AVIOContext *pb, unsigned ts)
170 {
172  avio_wb24(pb, 5); /* Tag Data Size */
173  avio_wb24(pb, ts); /* lower 24 bits of timestamp in ms */
174  avio_w8(pb, (ts >> 24) & 0x7F); /* MSB of ts in ms */
175  avio_wb24(pb, 0); /* StreamId = 0 */
176  avio_w8(pb, 23); /* ub[4] FrameType = 1, ub[4] CodecId = 7 */
177  avio_w8(pb, 2); /* AVC end of sequence */
178  avio_wb24(pb, 0); /* Always 0 for AVC EOS. */
179  avio_wb32(pb, 16); /* Size of FLV tag */
180 }
181 
182 static void put_amf_double(AVIOContext *pb, double d)
183 {
185  avio_wb64(pb, av_double2int(d));
186 }
187 
188 static void put_amf_bool(AVIOContext *pb, int b)
189 {
191  avio_w8(pb, !!b);
192 }
193 
195 {
196  AVIOContext *pb = s->pb;
197  FLVContext *flv = s->priv_data;
198  AVCodecContext *audio_enc = NULL, *video_enc = NULL, *data_enc = NULL;
199  int i, metadata_count = 0;
200  double framerate = 0.0;
201  int64_t metadata_size_pos, data_size, metadata_count_pos;
203 
204  for (i = 0; i < s->nb_streams; i++) {
205  AVCodecContext *enc = s->streams[i]->codec;
206  FLVStreamContext *sc;
207  switch (enc->codec_type) {
208  case AVMEDIA_TYPE_VIDEO:
209  if (s->streams[i]->avg_frame_rate.den &&
210  s->streams[i]->avg_frame_rate.num) {
211  framerate = av_q2d(s->streams[i]->avg_frame_rate);
212  } else {
213  framerate = 1 / av_q2d(s->streams[i]->codec->time_base);
214  }
215  video_enc = enc;
216  if (enc->codec_tag == 0) {
217  av_log(s, AV_LOG_ERROR, "Video codec '%s' for stream %d is not compatible with FLV\n",
218  avcodec_get_name(enc->codec_id), i);
219  return AVERROR(EINVAL);
220  }
221  break;
222  case AVMEDIA_TYPE_AUDIO:
223  audio_enc = enc;
224  if (get_audio_flags(s, enc) < 0)
225  return AVERROR_INVALIDDATA;
226  break;
227  case AVMEDIA_TYPE_DATA:
228  if (enc->codec_id != AV_CODEC_ID_TEXT) {
229  av_log(s, AV_LOG_ERROR, "Data codec '%s' for stream %d is not compatible with FLV\n",
230  avcodec_get_name(enc->codec_id), i);
231  return AVERROR_INVALIDDATA;
232  }
233  data_enc = enc;
234  break;
235  default:
236  av_log(s, AV_LOG_ERROR, "Codec type '%s' for stream %d is not compatible with FLV\n",
238  return AVERROR(EINVAL);
239  }
240  avpriv_set_pts_info(s->streams[i], 32, 1, 1000); /* 32 bit pts in ms */
241 
242  sc = av_mallocz(sizeof(FLVStreamContext));
243  if (!sc)
244  return AVERROR(ENOMEM);
245  s->streams[i]->priv_data = sc;
246  sc->last_ts = -1;
247  }
248 
249  flv->delay = AV_NOPTS_VALUE;
250 
251  avio_write(pb, "FLV", 3);
252  avio_w8(pb, 1);
253  avio_w8(pb, FLV_HEADER_FLAG_HASAUDIO * !!audio_enc +
254  FLV_HEADER_FLAG_HASVIDEO * !!video_enc);
255  avio_wb32(pb, 9);
256  avio_wb32(pb, 0);
257 
258  for (i = 0; i < s->nb_streams; i++)
259  if (s->streams[i]->codec->codec_tag == 5) {
260  avio_w8(pb, 8); // message type
261  avio_wb24(pb, 0); // include flags
262  avio_wb24(pb, 0); // time stamp
263  avio_wb32(pb, 0); // reserved
264  avio_wb32(pb, 11); // size
265  flv->reserved = 5;
266  }
267 
268  /* write meta_tag */
269  avio_w8(pb, 18); // tag type META
270  metadata_size_pos = avio_tell(pb);
271  avio_wb24(pb, 0); // size of data part (sum of all parts below)
272  avio_wb24(pb, 0); // timestamp
273  avio_wb32(pb, 0); // reserved
274 
275  /* now data of data_size size */
276 
277  /* first event name as a string */
279  put_amf_string(pb, "onMetaData"); // 12 bytes
280 
281  /* mixed array (hash) with size and string/type/data tuples */
283  metadata_count_pos = avio_tell(pb);
284  metadata_count = 5 * !!video_enc +
285  5 * !!audio_enc +
286  1 * !!data_enc +
287  2; // +2 for duration and file size
288 
289  avio_wb32(pb, metadata_count);
290 
291  put_amf_string(pb, "duration");
292  flv->duration_offset= avio_tell(pb);
293 
294  // fill in the guessed duration, it'll be corrected later if incorrect
296 
297  if (video_enc) {
298  put_amf_string(pb, "width");
299  put_amf_double(pb, video_enc->width);
300 
301  put_amf_string(pb, "height");
302  put_amf_double(pb, video_enc->height);
303 
304  put_amf_string(pb, "videodatarate");
305  put_amf_double(pb, video_enc->bit_rate / 1024.0);
306 
307  put_amf_string(pb, "framerate");
308  put_amf_double(pb, framerate);
309 
310  put_amf_string(pb, "videocodecid");
311  put_amf_double(pb, video_enc->codec_tag);
312  }
313 
314  if (audio_enc) {
315  put_amf_string(pb, "audiodatarate");
316  put_amf_double(pb, audio_enc->bit_rate / 1024.0);
317 
318  put_amf_string(pb, "audiosamplerate");
319  put_amf_double(pb, audio_enc->sample_rate);
320 
321  put_amf_string(pb, "audiosamplesize");
322  put_amf_double(pb, audio_enc->codec_id == AV_CODEC_ID_PCM_U8 ? 8 : 16);
323 
324  put_amf_string(pb, "stereo");
325  put_amf_bool(pb, audio_enc->channels == 2);
326 
327  put_amf_string(pb, "audiocodecid");
328  put_amf_double(pb, audio_enc->codec_tag);
329  }
330 
331  if (data_enc) {
332  put_amf_string(pb, "datastream");
333  put_amf_double(pb, 0.0);
334  }
335 
336  while ((tag = av_dict_get(s->metadata, "", tag, AV_DICT_IGNORE_SUFFIX))) {
337  if( !strcmp(tag->key, "width")
338  ||!strcmp(tag->key, "height")
339  ||!strcmp(tag->key, "videodatarate")
340  ||!strcmp(tag->key, "framerate")
341  ||!strcmp(tag->key, "videocodecid")
342  ||!strcmp(tag->key, "audiodatarate")
343  ||!strcmp(tag->key, "audiosamplerate")
344  ||!strcmp(tag->key, "audiosamplesize")
345  ||!strcmp(tag->key, "stereo")
346  ||!strcmp(tag->key, "audiocodecid")
347  ||!strcmp(tag->key, "duration")
348  ||!strcmp(tag->key, "onMetaData")
349  ){
350  av_log(s, AV_LOG_DEBUG, "Ignoring metadata for %s\n", tag->key);
351  continue;
352  }
353  put_amf_string(pb, tag->key);
355  put_amf_string(pb, tag->value);
356  metadata_count++;
357  }
358 
359  put_amf_string(pb, "filesize");
360  flv->filesize_offset = avio_tell(pb);
361  put_amf_double(pb, 0); // delayed write
362 
363  put_amf_string(pb, "");
365 
366  /* write total size of tag */
367  data_size = avio_tell(pb) - metadata_size_pos - 10;
368 
369  avio_seek(pb, metadata_count_pos, SEEK_SET);
370  avio_wb32(pb, metadata_count);
371 
372  avio_seek(pb, metadata_size_pos, SEEK_SET);
373  avio_wb24(pb, data_size);
374  avio_skip(pb, data_size + 10 - 3);
375  avio_wb32(pb, data_size + 11);
376 
377  for (i = 0; i < s->nb_streams; i++) {
378  AVCodecContext *enc = s->streams[i]->codec;
379  if (enc->codec_id == AV_CODEC_ID_AAC || enc->codec_id == AV_CODEC_ID_H264 || enc->codec_id == AV_CODEC_ID_MPEG4) {
380  int64_t pos;
383  avio_wb24(pb, 0); // size patched later
384  avio_wb24(pb, 0); // ts
385  avio_w8(pb, 0); // ts ext
386  avio_wb24(pb, 0); // streamid
387  pos = avio_tell(pb);
388  if (enc->codec_id == AV_CODEC_ID_AAC) {
389  avio_w8(pb, get_audio_flags(s, enc));
390  avio_w8(pb, 0); // AAC sequence header
391  avio_write(pb, enc->extradata, enc->extradata_size);
392  } else {
393  avio_w8(pb, enc->codec_tag | FLV_FRAME_KEY); // flags
394  avio_w8(pb, 0); // AVC sequence header
395  avio_wb24(pb, 0); // composition time
397  }
398  data_size = avio_tell(pb) - pos;
399  avio_seek(pb, -data_size - 10, SEEK_CUR);
400  avio_wb24(pb, data_size);
401  avio_skip(pb, data_size + 10 - 3);
402  avio_wb32(pb, data_size + 11); // previous tag size
403  }
404  }
405 
406  return 0;
407 }
408 
410 {
411  int64_t file_size;
412 
413  AVIOContext *pb = s->pb;
414  FLVContext *flv = s->priv_data;
415  int i;
416 
417  /* Add EOS tag */
418  for (i = 0; i < s->nb_streams; i++) {
419  AVCodecContext *enc = s->streams[i]->codec;
420  FLVStreamContext *sc = s->streams[i]->priv_data;
421  if (enc->codec_type == AVMEDIA_TYPE_VIDEO &&
423  put_avc_eos_tag(pb, sc->last_ts);
424  }
425 
426  file_size = avio_tell(pb);
427 
428  /* update information */
429  if (avio_seek(pb, flv->duration_offset, SEEK_SET) < 0)
430  av_log(s, AV_LOG_WARNING, "Failed to update header with correct duration.\n");
431  else
432  put_amf_double(pb, flv->duration / (double)1000);
433  if (avio_seek(pb, flv->filesize_offset, SEEK_SET) < 0)
434  av_log(s, AV_LOG_WARNING, "Failed to update header with correct filesize.\n");
435  else
436  put_amf_double(pb, file_size);
437 
438  avio_seek(pb, file_size, SEEK_SET);
439  return 0;
440 }
441 
443 {
444  AVIOContext *pb = s->pb;
445  AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
446  FLVContext *flv = s->priv_data;
448  unsigned ts;
449  int size = pkt->size;
450  uint8_t *data = NULL;
451  int flags = -1, flags_size, ret;
452 
453  if (enc->codec_id == AV_CODEC_ID_VP6 || enc->codec_id == AV_CODEC_ID_VP6F ||
455  flags_size = 2;
456  else if (enc->codec_id == AV_CODEC_ID_H264 || enc->codec_id == AV_CODEC_ID_MPEG4)
457  flags_size = 5;
458  else
459  flags_size = 1;
460 
461  switch (enc->codec_type) {
462  case AVMEDIA_TYPE_VIDEO:
464 
465  flags = enc->codec_tag;
466  if (flags == 0) {
467  av_log(s, AV_LOG_ERROR,
468  "Video codec '%s' is not compatible with FLV\n",
469  avcodec_get_name(enc->codec_id));
470  return AVERROR(EINVAL);
471  }
472 
474  break;
475  case AVMEDIA_TYPE_AUDIO:
476  flags = get_audio_flags(s, enc);
477 
478  av_assert0(size);
479 
481  break;
482  case AVMEDIA_TYPE_DATA:
484  break;
485  default:
486  return AVERROR(EINVAL);
487  }
488 
489  if (enc->codec_id == AV_CODEC_ID_H264 || enc->codec_id == AV_CODEC_ID_MPEG4) {
490  /* check if extradata looks like mp4 formated */
491  if (enc->extradata_size > 0 && *(uint8_t*)enc->extradata != 1)
492  if ((ret = ff_avc_parse_nal_units_buf(pkt->data, &data, &size)) < 0)
493  return ret;
494  } else if (enc->codec_id == AV_CODEC_ID_AAC && pkt->size > 2 &&
495  (AV_RB16(pkt->data) & 0xfff0) == 0xfff0) {
496  if (!s->streams[pkt->stream_index]->nb_frames) {
497  av_log(s, AV_LOG_ERROR, "Malformed AAC bitstream detected: "
498  "use audio bitstream filter 'aac_adtstoasc' to fix it "
499  "('-bsf:a aac_adtstoasc' option with ffmpeg)\n");
500  return AVERROR_INVALIDDATA;
501  }
502  av_log(s, AV_LOG_WARNING, "aac bitstream error\n");
503  }
504 
505  if (flv->delay == AV_NOPTS_VALUE)
506  flv->delay = -pkt->dts;
507 
508  if (pkt->dts < -flv->delay) {
510  "Packets are not in the proper order with respect to DTS\n");
511  return AVERROR(EINVAL);
512  }
513 
514  ts = pkt->dts + flv->delay; // add delay to force positive dts
515 
516  /* check Speex packet duration */
517  if (enc->codec_id == AV_CODEC_ID_SPEEX && ts - sc->last_ts > 160)
518  av_log(s, AV_LOG_WARNING, "Warning: Speex stream has more than "
519  "8 frames per packet. Adobe Flash "
520  "Player cannot handle this!\n");
521 
522  if (sc->last_ts < ts)
523  sc->last_ts = ts;
524 
525  avio_wb24(pb, size + flags_size);
526  avio_wb24(pb, ts);
527  avio_w8(pb, (ts >> 24) & 0x7F); // timestamps are 32 bits _signed_
528  avio_wb24(pb, flv->reserved);
529 
530  if (enc->codec_type == AVMEDIA_TYPE_DATA) {
531  int data_size;
532  int metadata_size_pos = avio_tell(pb);
534  put_amf_string(pb, "onTextData");
536  avio_wb32(pb, 2);
537  put_amf_string(pb, "type");
539  put_amf_string(pb, "Text");
540  put_amf_string(pb, "text");
542  put_amf_string(pb, pkt->data);
543  put_amf_string(pb, "");
545  /* write total size of tag */
546  data_size = avio_tell(pb) - metadata_size_pos;
547  avio_seek(pb, metadata_size_pos - 10, SEEK_SET);
548  avio_wb24(pb, data_size);
549  avio_seek(pb, data_size + 10 - 3, SEEK_CUR);
550  avio_wb32(pb, data_size + 11);
551  } else {
552  av_assert1(flags>=0);
553  avio_w8(pb,flags);
554  if (enc->codec_id == AV_CODEC_ID_VP6)
555  avio_w8(pb,0);
556  if (enc->codec_id == AV_CODEC_ID_VP6F || enc->codec_id == AV_CODEC_ID_VP6A)
557  avio_w8(pb, enc->extradata_size ? enc->extradata[0] : 0);
558  else if (enc->codec_id == AV_CODEC_ID_AAC)
559  avio_w8(pb,1); // AAC raw
560  else if (enc->codec_id == AV_CODEC_ID_H264 || enc->codec_id == AV_CODEC_ID_MPEG4) {
561  avio_w8(pb,1); // AVC NALU
562  avio_wb24(pb,pkt->pts - pkt->dts);
563  }
564 
565  avio_write(pb, data ? data : pkt->data, size);
566 
567  avio_wb32(pb, size + flags_size + 11); // previous tag size
568  flv->duration = FFMAX(flv->duration,
569  pkt->pts + flv->delay + pkt->duration);
570  }
571 
572  avio_flush(pb);
573  av_free(data);
574 
575  return pb->error;
576 }
577 
579  .name = "flv",
580  .long_name = NULL_IF_CONFIG_SMALL("FLV (Flash Video)"),
581  .mime_type = "video/x-flv",
582  .extensions = "flv",
583  .priv_data_size = sizeof(FLVContext),
584  .audio_codec = CONFIG_LIBMP3LAME ? AV_CODEC_ID_MP3 : AV_CODEC_ID_ADPCM_SWF,
585  .video_codec = AV_CODEC_ID_FLV1,
589  .codec_tag = (const AVCodecTag* const []) {
591  },
594 };