FFmpeg
 All Data Structures Namespaces 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;
202  AVDictionaryEntry *tag = NULL;
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  if (video_enc) {
216  av_log(s, AV_LOG_ERROR,
217  "at most one video stream is supported in flv\n");
218  return AVERROR(EINVAL);
219  }
220  video_enc = enc;
221  if (enc->codec_tag == 0) {
222  av_log(s, AV_LOG_ERROR, "Video codec '%s' for stream %d is not compatible with FLV\n",
223  avcodec_get_name(enc->codec_id), i);
224  return AVERROR(EINVAL);
225  }
226  break;
227  case AVMEDIA_TYPE_AUDIO:
228  if (audio_enc) {
229  av_log(s, AV_LOG_ERROR,
230  "at most one audio stream is supported in flv\n");
231  return AVERROR(EINVAL);
232  }
233  audio_enc = enc;
234  if (get_audio_flags(s, enc) < 0)
235  return AVERROR_INVALIDDATA;
236  break;
237  case AVMEDIA_TYPE_DATA:
238  if (enc->codec_id != AV_CODEC_ID_TEXT) {
239  av_log(s, AV_LOG_ERROR, "Data codec '%s' for stream %d is not compatible with FLV\n",
240  avcodec_get_name(enc->codec_id), i);
241  return AVERROR_INVALIDDATA;
242  }
243  data_enc = enc;
244  break;
245  default:
246  av_log(s, AV_LOG_ERROR, "Codec type '%s' for stream %d is not compatible with FLV\n",
248  return AVERROR(EINVAL);
249  }
250  avpriv_set_pts_info(s->streams[i], 32, 1, 1000); /* 32 bit pts in ms */
251 
252  sc = av_mallocz(sizeof(FLVStreamContext));
253  if (!sc)
254  return AVERROR(ENOMEM);
255  s->streams[i]->priv_data = sc;
256  sc->last_ts = -1;
257  }
258 
259  flv->delay = AV_NOPTS_VALUE;
260 
261  avio_write(pb, "FLV", 3);
262  avio_w8(pb, 1);
263  avio_w8(pb, FLV_HEADER_FLAG_HASAUDIO * !!audio_enc +
264  FLV_HEADER_FLAG_HASVIDEO * !!video_enc);
265  avio_wb32(pb, 9);
266  avio_wb32(pb, 0);
267 
268  for (i = 0; i < s->nb_streams; i++)
269  if (s->streams[i]->codec->codec_tag == 5) {
270  avio_w8(pb, 8); // message type
271  avio_wb24(pb, 0); // include flags
272  avio_wb24(pb, 0); // time stamp
273  avio_wb32(pb, 0); // reserved
274  avio_wb32(pb, 11); // size
275  flv->reserved = 5;
276  }
277 
278  /* write meta_tag */
279  avio_w8(pb, 18); // tag type META
280  metadata_size_pos = avio_tell(pb);
281  avio_wb24(pb, 0); // size of data part (sum of all parts below)
282  avio_wb24(pb, 0); // timestamp
283  avio_wb32(pb, 0); // reserved
284 
285  /* now data of data_size size */
286 
287  /* first event name as a string */
289  put_amf_string(pb, "onMetaData"); // 12 bytes
290 
291  /* mixed array (hash) with size and string/type/data tuples */
293  metadata_count_pos = avio_tell(pb);
294  metadata_count = 5 * !!video_enc +
295  5 * !!audio_enc +
296  1 * !!data_enc +
297  2; // +2 for duration and file size
298 
299  avio_wb32(pb, metadata_count);
300 
301  put_amf_string(pb, "duration");
302  flv->duration_offset= avio_tell(pb);
303 
304  // fill in the guessed duration, it'll be corrected later if incorrect
306 
307  if (video_enc) {
308  put_amf_string(pb, "width");
309  put_amf_double(pb, video_enc->width);
310 
311  put_amf_string(pb, "height");
312  put_amf_double(pb, video_enc->height);
313 
314  put_amf_string(pb, "videodatarate");
315  put_amf_double(pb, video_enc->bit_rate / 1024.0);
316 
317  put_amf_string(pb, "framerate");
318  put_amf_double(pb, framerate);
319 
320  put_amf_string(pb, "videocodecid");
321  put_amf_double(pb, video_enc->codec_tag);
322  }
323 
324  if (audio_enc) {
325  put_amf_string(pb, "audiodatarate");
326  put_amf_double(pb, audio_enc->bit_rate / 1024.0);
327 
328  put_amf_string(pb, "audiosamplerate");
329  put_amf_double(pb, audio_enc->sample_rate);
330 
331  put_amf_string(pb, "audiosamplesize");
332  put_amf_double(pb, audio_enc->codec_id == AV_CODEC_ID_PCM_U8 ? 8 : 16);
333 
334  put_amf_string(pb, "stereo");
335  put_amf_bool(pb, audio_enc->channels == 2);
336 
337  put_amf_string(pb, "audiocodecid");
338  put_amf_double(pb, audio_enc->codec_tag);
339  }
340 
341  if (data_enc) {
342  put_amf_string(pb, "datastream");
343  put_amf_double(pb, 0.0);
344  }
345 
346  while ((tag = av_dict_get(s->metadata, "", tag, AV_DICT_IGNORE_SUFFIX))) {
347  if( !strcmp(tag->key, "width")
348  ||!strcmp(tag->key, "height")
349  ||!strcmp(tag->key, "videodatarate")
350  ||!strcmp(tag->key, "framerate")
351  ||!strcmp(tag->key, "videocodecid")
352  ||!strcmp(tag->key, "audiodatarate")
353  ||!strcmp(tag->key, "audiosamplerate")
354  ||!strcmp(tag->key, "audiosamplesize")
355  ||!strcmp(tag->key, "stereo")
356  ||!strcmp(tag->key, "audiocodecid")
357  ||!strcmp(tag->key, "duration")
358  ||!strcmp(tag->key, "onMetaData")
359  ){
360  av_log(s, AV_LOG_DEBUG, "Ignoring metadata for %s\n", tag->key);
361  continue;
362  }
363  put_amf_string(pb, tag->key);
365  put_amf_string(pb, tag->value);
366  metadata_count++;
367  }
368 
369  put_amf_string(pb, "filesize");
370  flv->filesize_offset = avio_tell(pb);
371  put_amf_double(pb, 0); // delayed write
372 
373  put_amf_string(pb, "");
375 
376  /* write total size of tag */
377  data_size = avio_tell(pb) - metadata_size_pos - 10;
378 
379  avio_seek(pb, metadata_count_pos, SEEK_SET);
380  avio_wb32(pb, metadata_count);
381 
382  avio_seek(pb, metadata_size_pos, SEEK_SET);
383  avio_wb24(pb, data_size);
384  avio_skip(pb, data_size + 10 - 3);
385  avio_wb32(pb, data_size + 11);
386 
387  for (i = 0; i < s->nb_streams; i++) {
388  AVCodecContext *enc = s->streams[i]->codec;
389  if (enc->codec_id == AV_CODEC_ID_AAC || enc->codec_id == AV_CODEC_ID_H264 || enc->codec_id == AV_CODEC_ID_MPEG4) {
390  int64_t pos;
393  avio_wb24(pb, 0); // size patched later
394  avio_wb24(pb, 0); // ts
395  avio_w8(pb, 0); // ts ext
396  avio_wb24(pb, 0); // streamid
397  pos = avio_tell(pb);
398  if (enc->codec_id == AV_CODEC_ID_AAC) {
399  avio_w8(pb, get_audio_flags(s, enc));
400  avio_w8(pb, 0); // AAC sequence header
401  avio_write(pb, enc->extradata, enc->extradata_size);
402  } else {
403  avio_w8(pb, enc->codec_tag | FLV_FRAME_KEY); // flags
404  avio_w8(pb, 0); // AVC sequence header
405  avio_wb24(pb, 0); // composition time
407  }
408  data_size = avio_tell(pb) - pos;
409  avio_seek(pb, -data_size - 10, SEEK_CUR);
410  avio_wb24(pb, data_size);
411  avio_skip(pb, data_size + 10 - 3);
412  avio_wb32(pb, data_size + 11); // previous tag size
413  }
414  }
415 
416  return 0;
417 }
418 
420 {
421  int64_t file_size;
422 
423  AVIOContext *pb = s->pb;
424  FLVContext *flv = s->priv_data;
425  int i;
426 
427  /* Add EOS tag */
428  for (i = 0; i < s->nb_streams; i++) {
429  AVCodecContext *enc = s->streams[i]->codec;
430  FLVStreamContext *sc = s->streams[i]->priv_data;
431  if (enc->codec_type == AVMEDIA_TYPE_VIDEO &&
433  put_avc_eos_tag(pb, sc->last_ts);
434  }
435 
436  file_size = avio_tell(pb);
437 
438  /* update information */
439  if (avio_seek(pb, flv->duration_offset, SEEK_SET) < 0)
440  av_log(s, AV_LOG_WARNING, "Failed to update header with correct duration.\n");
441  else
442  put_amf_double(pb, flv->duration / (double)1000);
443  if (avio_seek(pb, flv->filesize_offset, SEEK_SET) < 0)
444  av_log(s, AV_LOG_WARNING, "Failed to update header with correct filesize.\n");
445  else
446  put_amf_double(pb, file_size);
447 
448  avio_seek(pb, file_size, SEEK_SET);
449  return 0;
450 }
451 
453 {
454  AVIOContext *pb = s->pb;
455  AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
456  FLVContext *flv = s->priv_data;
458  unsigned ts;
459  int size = pkt->size;
460  uint8_t *data = NULL;
461  int flags = -1, flags_size, ret;
462 
463  if (enc->codec_id == AV_CODEC_ID_VP6 || enc->codec_id == AV_CODEC_ID_VP6F ||
465  flags_size = 2;
466  else if (enc->codec_id == AV_CODEC_ID_H264 || enc->codec_id == AV_CODEC_ID_MPEG4)
467  flags_size = 5;
468  else
469  flags_size = 1;
470 
471  switch (enc->codec_type) {
472  case AVMEDIA_TYPE_VIDEO:
474 
475  flags = enc->codec_tag;
476  if (flags == 0) {
477  av_log(s, AV_LOG_ERROR,
478  "Video codec '%s' is not compatible with FLV\n",
479  avcodec_get_name(enc->codec_id));
480  return AVERROR(EINVAL);
481  }
482 
484  break;
485  case AVMEDIA_TYPE_AUDIO:
486  flags = get_audio_flags(s, enc);
487 
488  av_assert0(size);
489 
491  break;
492  case AVMEDIA_TYPE_DATA:
494  break;
495  default:
496  return AVERROR(EINVAL);
497  }
498 
499  if (enc->codec_id == AV_CODEC_ID_H264 || enc->codec_id == AV_CODEC_ID_MPEG4) {
500  /* check if extradata looks like mp4 formated */
501  if (enc->extradata_size > 0 && *(uint8_t*)enc->extradata != 1)
502  if ((ret = ff_avc_parse_nal_units_buf(pkt->data, &data, &size)) < 0)
503  return ret;
504  } else if (enc->codec_id == AV_CODEC_ID_AAC && pkt->size > 2 &&
505  (AV_RB16(pkt->data) & 0xfff0) == 0xfff0) {
506  if (!s->streams[pkt->stream_index]->nb_frames) {
507  av_log(s, AV_LOG_ERROR, "Malformed AAC bitstream detected: "
508  "use audio bitstream filter 'aac_adtstoasc' to fix it "
509  "('-bsf:a aac_adtstoasc' option with ffmpeg)\n");
510  return AVERROR_INVALIDDATA;
511  }
512  av_log(s, AV_LOG_WARNING, "aac bitstream error\n");
513  }
514 
515  if (flv->delay == AV_NOPTS_VALUE)
516  flv->delay = -pkt->dts;
517 
518  if (pkt->dts < -flv->delay) {
520  "Packets are not in the proper order with respect to DTS\n");
521  return AVERROR(EINVAL);
522  }
523 
524  ts = pkt->dts + flv->delay; // add delay to force positive dts
525 
526  /* check Speex packet duration */
527  if (enc->codec_id == AV_CODEC_ID_SPEEX && ts - sc->last_ts > 160)
528  av_log(s, AV_LOG_WARNING, "Warning: Speex stream has more than "
529  "8 frames per packet. Adobe Flash "
530  "Player cannot handle this!\n");
531 
532  if (sc->last_ts < ts)
533  sc->last_ts = ts;
534 
535  avio_wb24(pb, size + flags_size);
536  avio_wb24(pb, ts);
537  avio_w8(pb, (ts >> 24) & 0x7F); // timestamps are 32 bits _signed_
538  avio_wb24(pb, flv->reserved);
539 
540  if (enc->codec_type == AVMEDIA_TYPE_DATA) {
541  int data_size;
542  int metadata_size_pos = avio_tell(pb);
544  put_amf_string(pb, "onTextData");
546  avio_wb32(pb, 2);
547  put_amf_string(pb, "type");
549  put_amf_string(pb, "Text");
550  put_amf_string(pb, "text");
552  put_amf_string(pb, pkt->data);
553  put_amf_string(pb, "");
555  /* write total size of tag */
556  data_size = avio_tell(pb) - metadata_size_pos;
557  avio_seek(pb, metadata_size_pos - 10, SEEK_SET);
558  avio_wb24(pb, data_size);
559  avio_seek(pb, data_size + 10 - 3, SEEK_CUR);
560  avio_wb32(pb, data_size + 11);
561  } else {
562  av_assert1(flags>=0);
563  avio_w8(pb,flags);
564  if (enc->codec_id == AV_CODEC_ID_VP6)
565  avio_w8(pb,0);
566  if (enc->codec_id == AV_CODEC_ID_VP6F || enc->codec_id == AV_CODEC_ID_VP6A)
567  avio_w8(pb, enc->extradata_size ? enc->extradata[0] : 0);
568  else if (enc->codec_id == AV_CODEC_ID_AAC)
569  avio_w8(pb,1); // AAC raw
570  else if (enc->codec_id == AV_CODEC_ID_H264 || enc->codec_id == AV_CODEC_ID_MPEG4) {
571  avio_w8(pb,1); // AVC NALU
572  avio_wb24(pb,pkt->pts - pkt->dts);
573  }
574 
575  avio_write(pb, data ? data : pkt->data, size);
576 
577  avio_wb32(pb, size + flags_size + 11); // previous tag size
578  flv->duration = FFMAX(flv->duration,
579  pkt->pts + flv->delay + pkt->duration);
580  }
581 
582  av_free(data);
583 
584  return pb->error;
585 }
586 
588  .name = "flv",
589  .long_name = NULL_IF_CONFIG_SMALL("FLV (Flash Video)"),
590  .mime_type = "video/x-flv",
591  .extensions = "flv",
592  .priv_data_size = sizeof(FLVContext),
593  .audio_codec = CONFIG_LIBMP3LAME ? AV_CODEC_ID_MP3 : AV_CODEC_ID_ADPCM_SWF,
594  .video_codec = AV_CODEC_ID_FLV1,
598  .codec_tag = (const AVCodecTag* const []) {
600  },
603 };