3 * Copyright (c) 2003 The Libav Project
5 * This file is part of Libav.
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.
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.
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
22 #include "libavutil/dict.h"
23 #include "libavutil/intfloat.h"
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_VP6A, FLV_CODECID_VP6A },
39 { AV_CODEC_ID_H264, FLV_CODECID_H264 },
40 { AV_CODEC_ID_NONE, 0 }
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 }
57 typedef struct FLVContext {
59 int64_t duration_offset;
60 int64_t filesize_offset;
62 int64_t delay; ///< first dts delay (needed for AVC & Speex)
64 AVCodecContext *audio_enc;
65 AVCodecContext *video_enc;
67 AVCodecContext *data_enc;
70 typedef struct FLVStreamContext {
71 int64_t last_ts; ///< last timestamp for each stream
74 static int get_audio_flags(AVFormatContext *s, AVCodecContext *enc)
76 int flags = (enc->bits_per_coded_sample == 16) ? FLV_SAMPLESSIZE_16BIT
77 : FLV_SAMPLESSIZE_8BIT;
79 if (enc->codec_id == AV_CODEC_ID_AAC) // specs force these parameters
80 return FLV_CODECID_AAC | FLV_SAMPLERATE_44100HZ |
81 FLV_SAMPLESSIZE_16BIT | FLV_STEREO;
82 else if (enc->codec_id == AV_CODEC_ID_SPEEX) {
83 if (enc->sample_rate != 16000) {
84 av_log(s, AV_LOG_ERROR,
85 "flv only supports wideband (16kHz) Speex audio\n");
88 if (enc->channels != 1) {
89 av_log(s, AV_LOG_ERROR, "flv only supports mono Speex audio\n");
92 return FLV_CODECID_SPEEX | FLV_SAMPLERATE_11025HZ | FLV_SAMPLESSIZE_16BIT;
94 switch (enc->sample_rate) {
96 flags |= FLV_SAMPLERATE_44100HZ;
99 flags |= FLV_SAMPLERATE_22050HZ;
102 flags |= FLV_SAMPLERATE_11025HZ;
104 case 16000: // nellymoser only
105 case 8000: // nellymoser only
106 case 5512: // not MP3
107 if (enc->codec_id != AV_CODEC_ID_MP3) {
108 flags |= FLV_SAMPLERATE_SPECIAL;
112 av_log(s, AV_LOG_ERROR,
113 "flv does not support that sample rate, "
114 "choose from (44100, 22050, 11025).\n");
119 if (enc->channels > 1)
122 switch (enc->codec_id) {
123 case AV_CODEC_ID_MP3:
124 flags |= FLV_CODECID_MP3 | FLV_SAMPLESSIZE_16BIT;
126 case AV_CODEC_ID_PCM_U8:
127 flags |= FLV_CODECID_PCM | FLV_SAMPLESSIZE_8BIT;
129 case AV_CODEC_ID_PCM_S16BE:
130 flags |= FLV_CODECID_PCM | FLV_SAMPLESSIZE_16BIT;
132 case AV_CODEC_ID_PCM_S16LE:
133 flags |= FLV_CODECID_PCM_LE | FLV_SAMPLESSIZE_16BIT;
135 case AV_CODEC_ID_ADPCM_SWF:
136 flags |= FLV_CODECID_ADPCM | FLV_SAMPLESSIZE_16BIT;
138 case AV_CODEC_ID_NELLYMOSER:
139 if (enc->sample_rate == 8000)
140 flags |= FLV_CODECID_NELLYMOSER_8KHZ_MONO | FLV_SAMPLESSIZE_16BIT;
141 else if (enc->sample_rate == 16000)
142 flags |= FLV_CODECID_NELLYMOSER_16KHZ_MONO | FLV_SAMPLESSIZE_16BIT;
144 flags |= FLV_CODECID_NELLYMOSER | FLV_SAMPLESSIZE_16BIT;
146 case AV_CODEC_ID_PCM_MULAW:
147 flags = FLV_CODECID_PCM_MULAW | FLV_SAMPLERATE_SPECIAL | FLV_SAMPLESSIZE_16BIT;
149 case AV_CODEC_ID_PCM_ALAW:
150 flags = FLV_CODECID_PCM_ALAW | FLV_SAMPLERATE_SPECIAL | FLV_SAMPLESSIZE_16BIT;
153 flags |= enc->codec_tag << 4;
156 av_log(s, AV_LOG_ERROR, "codec not compatible with flv\n");
163 static void put_amf_string(AVIOContext *pb, const char *str)
165 size_t len = strlen(str);
167 avio_write(pb, str, len);
170 static void put_avc_eos_tag(AVIOContext *pb, unsigned ts)
172 avio_w8(pb, FLV_TAG_TYPE_VIDEO);
173 avio_wb24(pb, 5); /* Tag Data Size */
174 avio_wb24(pb, ts); /* lower 24 bits of timestamp in ms */
175 avio_w8(pb, (ts >> 24) & 0x7F); /* MSB of ts in ms */
176 avio_wb24(pb, 0); /* StreamId = 0 */
177 avio_w8(pb, 23); /* ub[4] FrameType = 1, ub[4] CodecId = 7 */
178 avio_w8(pb, 2); /* AVC end of sequence */
179 avio_wb24(pb, 0); /* Always 0 for AVC EOS. */
180 avio_wb32(pb, 16); /* Size of FLV tag */
183 static void put_amf_double(AVIOContext *pb, double d)
185 avio_w8(pb, AMF_DATA_TYPE_NUMBER);
186 avio_wb64(pb, av_double2int(d));
189 static void put_amf_bool(AVIOContext *pb, int b)
191 avio_w8(pb, AMF_DATA_TYPE_BOOL);
195 static void write_metadata(AVFormatContext *s, unsigned int ts)
197 AVIOContext *pb = s->pb;
198 FLVContext *flv = s->priv_data;
199 int metadata_count = 0;
200 int64_t metadata_size_pos, data_size, metadata_count_pos;
201 AVDictionaryEntry *tag = NULL;
204 avio_w8(pb, 18); // tag type META
205 metadata_size_pos = avio_tell(pb);
206 avio_wb24(pb, 0); // size of data part (sum of all parts below)
207 avio_wb24(pb, ts); // timestamp
208 avio_wb32(pb, 0); // reserved
210 /* now data of data_size size */
212 /* first event name as a string */
213 avio_w8(pb, AMF_DATA_TYPE_STRING);
214 put_amf_string(pb, "onMetaData"); // 12 bytes
216 /* mixed array (hash) with size and string/type/data tuples */
217 avio_w8(pb, AMF_DATA_TYPE_MIXEDARRAY);
218 metadata_count_pos = avio_tell(pb);
219 metadata_count = 4 * !!flv->video_enc +
220 5 * !!flv->audio_enc +
221 1 * !!flv->data_enc +
222 2; // +2 for duration and file size
224 avio_wb32(pb, metadata_count);
226 put_amf_string(pb, "duration");
227 flv->duration_offset = avio_tell(pb);
229 // fill in the guessed duration, it'll be corrected later if incorrect
230 put_amf_double(pb, s->duration / AV_TIME_BASE);
232 if (flv->video_enc) {
233 put_amf_string(pb, "width");
234 put_amf_double(pb, flv->video_enc->width);
236 put_amf_string(pb, "height");
237 put_amf_double(pb, flv->video_enc->height);
239 put_amf_string(pb, "videodatarate");
240 put_amf_double(pb, flv->video_enc->bit_rate / 1024.0);
242 if (flv->framerate != 0.0) {
243 put_amf_string(pb, "framerate");
244 put_amf_double(pb, flv->framerate);
248 put_amf_string(pb, "videocodecid");
249 put_amf_double(pb, flv->video_enc->codec_tag);
252 if (flv->audio_enc) {
253 put_amf_string(pb, "audiodatarate");
254 put_amf_double(pb, flv->audio_enc->bit_rate / 1024.0);
256 put_amf_string(pb, "audiosamplerate");
257 put_amf_double(pb, flv->audio_enc->sample_rate);
259 put_amf_string(pb, "audiosamplesize");
260 put_amf_double(pb, flv->audio_enc->codec_id == AV_CODEC_ID_PCM_U8 ? 8 : 16);
262 put_amf_string(pb, "stereo");
263 put_amf_bool(pb, flv->audio_enc->channels == 2);
265 put_amf_string(pb, "audiocodecid");
266 put_amf_double(pb, flv->audio_enc->codec_tag);
270 put_amf_string(pb, "datastream");
271 put_amf_double(pb, 0.0);
274 while ((tag = av_dict_get(s->metadata, "", tag, AV_DICT_IGNORE_SUFFIX))) {
275 put_amf_string(pb, tag->key);
276 avio_w8(pb, AMF_DATA_TYPE_STRING);
277 put_amf_string(pb, tag->value);
281 put_amf_string(pb, "filesize");
282 flv->filesize_offset = avio_tell(pb);
283 put_amf_double(pb, 0); // delayed write
285 put_amf_string(pb, "");
286 avio_w8(pb, AMF_END_OF_OBJECT);
288 /* write total size of tag */
289 data_size = avio_tell(pb) - metadata_size_pos - 10;
291 avio_seek(pb, metadata_count_pos, SEEK_SET);
292 avio_wb32(pb, metadata_count);
294 avio_seek(pb, metadata_size_pos, SEEK_SET);
295 avio_wb24(pb, data_size);
296 avio_skip(pb, data_size + 10 - 3);
297 avio_wb32(pb, data_size + 11);
300 static int unsupported_codec(AVFormatContext *s,
301 const char* type, int codec_id)
303 const AVCodecDescriptor *desc = avcodec_descriptor_get(codec_id);
304 av_log(s, AV_LOG_ERROR,
305 "%s codec %s not compatible with flv\n",
307 desc ? desc->name : "unknown");
308 return AVERROR(ENOSYS);
311 static int flv_write_header(AVFormatContext *s)
314 AVIOContext *pb = s->pb;
315 FLVContext *flv = s->priv_data;
318 for (i = 0; i < s->nb_streams; i++) {
319 AVCodecContext *enc = s->streams[i]->codec;
320 FLVStreamContext *sc;
321 switch (enc->codec_type) {
322 case AVMEDIA_TYPE_VIDEO:
323 if (s->streams[i]->avg_frame_rate.den &&
324 s->streams[i]->avg_frame_rate.num) {
325 flv->framerate = av_q2d(s->streams[i]->avg_frame_rate);
327 if (flv->video_enc) {
328 av_log(s, AV_LOG_ERROR,
329 "at most one video stream is supported in flv\n");
330 return AVERROR(EINVAL);
332 flv->video_enc = enc;
333 if (!ff_codec_get_tag(flv_video_codec_ids, enc->codec_id))
334 return unsupported_codec(s, "Video", enc->codec_id);
336 case AVMEDIA_TYPE_AUDIO:
337 if (flv->audio_enc) {
338 av_log(s, AV_LOG_ERROR,
339 "at most one audio stream is supported in flv\n");
340 return AVERROR(EINVAL);
342 flv->audio_enc = enc;
343 if (get_audio_flags(s, enc) < 0)
344 return unsupported_codec(s, "Audio", enc->codec_id);
346 case AVMEDIA_TYPE_DATA:
347 if (enc->codec_id != AV_CODEC_ID_TEXT)
348 return unsupported_codec(s, "Data", enc->codec_id);
352 av_log(s, AV_LOG_ERROR, "codec not compatible with flv\n");
355 avpriv_set_pts_info(s->streams[i], 32, 1, 1000); /* 32 bit pts in ms */
357 sc = av_mallocz(sizeof(FLVStreamContext));
359 return AVERROR(ENOMEM);
360 s->streams[i]->priv_data = sc;
364 flv->delay = AV_NOPTS_VALUE;
366 avio_write(pb, "FLV", 3);
368 avio_w8(pb, FLV_HEADER_FLAG_HASAUDIO * !!flv->audio_enc +
369 FLV_HEADER_FLAG_HASVIDEO * !!flv->video_enc);
373 for (i = 0; i < s->nb_streams; i++)
374 if (s->streams[i]->codec->codec_tag == 5) {
375 avio_w8(pb, 8); // message type
376 avio_wb24(pb, 0); // include flags
377 avio_wb24(pb, 0); // time stamp
378 avio_wb32(pb, 0); // reserved
379 avio_wb32(pb, 11); // size
383 write_metadata(s, 0);
385 for (i = 0; i < s->nb_streams; i++) {
386 AVCodecContext *enc = s->streams[i]->codec;
387 if (enc->codec_id == AV_CODEC_ID_AAC || enc->codec_id == AV_CODEC_ID_H264) {
389 avio_w8(pb, enc->codec_type == AVMEDIA_TYPE_VIDEO ?
390 FLV_TAG_TYPE_VIDEO : FLV_TAG_TYPE_AUDIO);
391 avio_wb24(pb, 0); // size patched later
392 avio_wb24(pb, 0); // ts
393 avio_w8(pb, 0); // ts ext
394 avio_wb24(pb, 0); // streamid
396 if (enc->codec_id == AV_CODEC_ID_AAC) {
397 avio_w8(pb, get_audio_flags(s, enc));
398 avio_w8(pb, 0); // AAC sequence header
399 avio_write(pb, enc->extradata, enc->extradata_size);
401 avio_w8(pb, enc->codec_tag | FLV_FRAME_KEY); // flags
402 avio_w8(pb, 0); // AVC sequence header
403 avio_wb24(pb, 0); // composition time
404 ff_isom_write_avcc(pb, enc->extradata, enc->extradata_size);
406 data_size = avio_tell(pb) - pos;
407 avio_seek(pb, -data_size - 10, SEEK_CUR);
408 avio_wb24(pb, data_size);
409 avio_skip(pb, data_size + 10 - 3);
410 avio_wb32(pb, data_size + 11); // previous tag size
417 static int flv_write_trailer(AVFormatContext *s)
421 AVIOContext *pb = s->pb;
422 FLVContext *flv = s->priv_data;
426 for (i = 0; i < s->nb_streams; i++) {
427 AVCodecContext *enc = s->streams[i]->codec;
428 FLVStreamContext *sc = s->streams[i]->priv_data;
429 if (enc->codec_type == AVMEDIA_TYPE_VIDEO &&
430 enc->codec_id == AV_CODEC_ID_H264)
431 put_avc_eos_tag(pb, sc->last_ts);
434 file_size = avio_tell(pb);
436 /* update information */
437 if (avio_seek(pb, flv->duration_offset, SEEK_SET) < 0)
438 av_log(s, AV_LOG_WARNING, "Failed to update header with correct duration.\n");
440 put_amf_double(pb, flv->duration / (double)1000);
441 if (avio_seek(pb, flv->filesize_offset, SEEK_SET) < 0)
442 av_log(s, AV_LOG_WARNING, "Failed to update header with correct filesize.\n");
444 put_amf_double(pb, file_size);
446 avio_seek(pb, file_size, SEEK_SET);
450 static int flv_write_packet(AVFormatContext *s, AVPacket *pkt)
452 AVIOContext *pb = s->pb;
453 AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
454 FLVContext *flv = s->priv_data;
455 FLVStreamContext *sc = s->streams[pkt->stream_index]->priv_data;
457 int size = pkt->size;
458 uint8_t *data = NULL;
459 int flags = 0, flags_size;
461 if (enc->codec_id == AV_CODEC_ID_VP6F || enc->codec_id == AV_CODEC_ID_VP6A ||
462 enc->codec_id == AV_CODEC_ID_AAC)
464 else if (enc->codec_id == AV_CODEC_ID_H264)
469 if (flv->delay == AV_NOPTS_VALUE)
470 flv->delay = -pkt->dts;
472 if (pkt->dts < -flv->delay) {
473 av_log(s, AV_LOG_WARNING,
474 "Packets are not in the proper order with respect to DTS\n");
475 return AVERROR(EINVAL);
478 ts = pkt->dts + flv->delay; // add delay to force positive dts
480 if (s->event_flags & AVSTREAM_EVENT_FLAG_METADATA_UPDATED) {
481 write_metadata(s, ts);
482 s->event_flags &= ~AVSTREAM_EVENT_FLAG_METADATA_UPDATED;
485 switch (enc->codec_type) {
486 case AVMEDIA_TYPE_VIDEO:
487 avio_w8(pb, FLV_TAG_TYPE_VIDEO);
489 flags = ff_codec_get_tag(flv_video_codec_ids, enc->codec_id);
491 flags |= pkt->flags & AV_PKT_FLAG_KEY ? FLV_FRAME_KEY : FLV_FRAME_INTER;
493 case AVMEDIA_TYPE_AUDIO:
494 flags = get_audio_flags(s, enc);
498 avio_w8(pb, FLV_TAG_TYPE_AUDIO);
500 case AVMEDIA_TYPE_DATA:
501 avio_w8(pb, FLV_TAG_TYPE_META);
504 return AVERROR(EINVAL);
507 if (enc->codec_id == AV_CODEC_ID_H264)
508 /* check if extradata looks like MP4 */
509 if (enc->extradata_size > 0 && *(uint8_t*)enc->extradata != 1)
510 if (ff_avc_parse_nal_units_buf(pkt->data, &data, &size) < 0)
513 /* check Speex packet duration */
514 if (enc->codec_id == AV_CODEC_ID_SPEEX && ts - sc->last_ts > 160)
515 av_log(s, AV_LOG_WARNING, "Warning: Speex stream has more than "
516 "8 frames per packet. Adobe Flash "
517 "Player cannot handle this!\n");
519 if (sc->last_ts < ts)
522 avio_wb24(pb, size + flags_size);
524 avio_w8(pb, (ts >> 24) & 0x7F); // timestamps are 32 bits _signed_
525 avio_wb24(pb, flv->reserved);
527 if (enc->codec_type == AVMEDIA_TYPE_DATA) {
529 int64_t metadata_size_pos = avio_tell(pb);
530 avio_w8(pb, AMF_DATA_TYPE_STRING);
531 put_amf_string(pb, "onTextData");
532 avio_w8(pb, AMF_DATA_TYPE_MIXEDARRAY);
534 put_amf_string(pb, "type");
535 avio_w8(pb, AMF_DATA_TYPE_STRING);
536 put_amf_string(pb, "Text");
537 put_amf_string(pb, "text");
538 avio_w8(pb, AMF_DATA_TYPE_STRING);
539 put_amf_string(pb, pkt->data);
540 put_amf_string(pb, "");
541 avio_w8(pb, AMF_END_OF_OBJECT);
542 /* write total size of tag */
543 data_size = avio_tell(pb) - metadata_size_pos;
544 avio_seek(pb, metadata_size_pos - 10, SEEK_SET);
545 avio_wb24(pb, data_size);
546 avio_seek(pb, data_size + 10 - 3, SEEK_CUR);
547 avio_wb32(pb, data_size + 11);
550 if (enc->codec_id == AV_CODEC_ID_VP6F || enc->codec_id == AV_CODEC_ID_VP6A) {
551 if (enc->extradata_size)
552 avio_w8(pb, enc->extradata[0]);
554 avio_w8(pb, ((FFALIGN(enc->width, 16) - enc->width) << 4) |
555 (FFALIGN(enc->height, 16) - enc->height));
556 } else if (enc->codec_id == AV_CODEC_ID_AAC)
557 avio_w8(pb, 1); // AAC raw
558 else if (enc->codec_id == AV_CODEC_ID_H264) {
559 avio_w8(pb, 1); // AVC NALU
560 avio_wb24(pb, pkt->pts - pkt->dts);
563 avio_write(pb, data ? data : pkt->data, size);
565 avio_wb32(pb, size + flags_size + 11); // previous tag size
566 flv->duration = FFMAX(flv->duration,
567 pkt->pts + flv->delay + pkt->duration);
575 AVOutputFormat ff_flv_muxer = {
577 .long_name = NULL_IF_CONFIG_SMALL("FLV (Flash Video)"),
578 .mime_type = "video/x-flv",
580 .priv_data_size = sizeof(FLVContext),
581 .audio_codec = CONFIG_LIBMP3LAME ? AV_CODEC_ID_MP3 : AV_CODEC_ID_ADPCM_SWF,
582 .video_codec = AV_CODEC_ID_FLV1,
583 .write_header = flv_write_header,
584 .write_packet = flv_write_packet,
585 .write_trailer = flv_write_trailer,
586 .codec_tag = (const AVCodecTag* const []) {
587 flv_video_codec_ids, flv_audio_codec_ids, 0
589 .flags = AVFMT_GLOBALHEADER | AVFMT_VARIABLE_FPS |