FFmpeg: libavformat/flvenc.c Source File

00001 /*
00002  * FLV muxer
00003  * Copyright (c) 2003 The FFmpeg Project
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00022 #include "libavutil/intreadwrite.h"
00023 #include "avformat.h"
00024 #include "flv.h"
00025 #include "internal.h"
00026 #include "avc.h"
00027 #include "metadata.h"
00028 #include "libavutil/dict.h"
00029 
00030 #undef NDEBUG
00031 #include <assert.h>
00032 
00033 static const AVCodecTag flv_video_codec_ids[] = {
00034     {CODEC_ID_FLV1,    FLV_CODECID_H263  },
00035     {CODEC_ID_FLASHSV, FLV_CODECID_SCREEN},
00036     {CODEC_ID_FLASHSV2, FLV_CODECID_SCREEN2},
00037     {CODEC_ID_VP6F,    FLV_CODECID_VP6   },
00038     {CODEC_ID_VP6,     FLV_CODECID_VP6   },
00039     {CODEC_ID_H264,    FLV_CODECID_H264  },
00040     {CODEC_ID_NONE,    0}
00041 };
00042 
00043 static const AVCodecTag flv_audio_codec_ids[] = {
00044     {CODEC_ID_MP3,       FLV_CODECID_MP3    >> FLV_AUDIO_CODECID_OFFSET},
00045     {CODEC_ID_PCM_U8,    FLV_CODECID_PCM    >> FLV_AUDIO_CODECID_OFFSET},
00046     {CODEC_ID_PCM_S16BE, FLV_CODECID_PCM    >> FLV_AUDIO_CODECID_OFFSET},
00047     {CODEC_ID_PCM_S16LE, FLV_CODECID_PCM_LE >> FLV_AUDIO_CODECID_OFFSET},
00048     {CODEC_ID_ADPCM_SWF, FLV_CODECID_ADPCM  >> FLV_AUDIO_CODECID_OFFSET},
00049     {CODEC_ID_AAC,       FLV_CODECID_AAC    >> FLV_AUDIO_CODECID_OFFSET},
00050     {CODEC_ID_NELLYMOSER, FLV_CODECID_NELLYMOSER >> FLV_AUDIO_CODECID_OFFSET},
00051     {CODEC_ID_SPEEX,     FLV_CODECID_SPEEX  >> FLV_AUDIO_CODECID_OFFSET},
00052     {CODEC_ID_NONE,      0}
00053 };
00054 
00055 typedef struct FLVContext {
00056     int reserved;
00057     int64_t duration_offset;
00058     int64_t filesize_offset;
00059     int64_t duration;
00060     int delay; 
00061     int64_t last_video_ts;
00062 } FLVContext;
00063 
00064 static int get_audio_flags(AVCodecContext *enc){
00065     int flags = (enc->bits_per_coded_sample == 16) ? FLV_SAMPLESSIZE_16BIT : FLV_SAMPLESSIZE_8BIT;
00066 
00067     if (enc->codec_id == CODEC_ID_AAC) // specs force these parameters
00068         return FLV_CODECID_AAC | FLV_SAMPLERATE_44100HZ | FLV_SAMPLESSIZE_16BIT | FLV_STEREO;
00069     else if (enc->codec_id == CODEC_ID_SPEEX) {
00070         if (enc->sample_rate != 16000) {
00071             av_log(enc, AV_LOG_ERROR, "flv only supports wideband (16kHz) Speex audio\n");
00072             return -1;
00073         }
00074         if (enc->channels != 1) {
00075             av_log(enc, AV_LOG_ERROR, "flv only supports mono Speex audio\n");
00076             return -1;
00077         }
00078         if (enc->frame_size / 320 > 8) {
00079             av_log(enc, AV_LOG_WARNING, "Warning: Speex stream has more than "
00080                                         "8 frames per packet. Adobe Flash "
00081                                         "Player cannot handle this!\n");
00082         }
00083         return FLV_CODECID_SPEEX | FLV_SAMPLERATE_11025HZ | FLV_SAMPLESSIZE_16BIT;
00084     } else {
00085     switch (enc->sample_rate) {
00086         case    44100:
00087             flags |= FLV_SAMPLERATE_44100HZ;
00088             break;
00089         case    22050:
00090             flags |= FLV_SAMPLERATE_22050HZ;
00091             break;
00092         case    11025:
00093             flags |= FLV_SAMPLERATE_11025HZ;
00094             break;
00095         case     8000: //nellymoser only
00096         case     5512: //not mp3
00097             if(enc->codec_id != CODEC_ID_MP3){
00098                 flags |= FLV_SAMPLERATE_SPECIAL;
00099                 break;
00100             }
00101         default:
00102             av_log(enc, AV_LOG_ERROR, "flv does not support that sample rate, choose from (44100, 22050, 11025).\n");
00103             return -1;
00104     }
00105     }
00106 
00107     if (enc->channels > 1) {
00108         flags |= FLV_STEREO;
00109     }
00110 
00111     switch(enc->codec_id){
00112     case CODEC_ID_MP3:
00113         flags |= FLV_CODECID_MP3    | FLV_SAMPLESSIZE_16BIT;
00114         break;
00115     case CODEC_ID_PCM_U8:
00116         flags |= FLV_CODECID_PCM    | FLV_SAMPLESSIZE_8BIT;
00117         break;
00118     case CODEC_ID_PCM_S16BE:
00119         flags |= FLV_CODECID_PCM    | FLV_SAMPLESSIZE_16BIT;
00120         break;
00121     case CODEC_ID_PCM_S16LE:
00122         flags |= FLV_CODECID_PCM_LE | FLV_SAMPLESSIZE_16BIT;
00123         break;
00124     case CODEC_ID_ADPCM_SWF:
00125         flags |= FLV_CODECID_ADPCM | FLV_SAMPLESSIZE_16BIT;
00126         break;
00127     case CODEC_ID_NELLYMOSER:
00128         if (enc->sample_rate == 8000) {
00129             flags |= FLV_CODECID_NELLYMOSER_8KHZ_MONO | FLV_SAMPLESSIZE_16BIT;
00130         } else {
00131             flags |= FLV_CODECID_NELLYMOSER | FLV_SAMPLESSIZE_16BIT;
00132         }
00133         break;
00134     case 0:
00135         flags |= enc->codec_tag<<4;
00136         break;
00137     default:
00138         av_log(enc, AV_LOG_ERROR, "codec not compatible with flv\n");
00139         return -1;
00140     }
00141 
00142     return flags;
00143 }
00144 
00145 static void put_amf_string(AVIOContext *pb, const char *str)
00146 {
00147     size_t len = strlen(str);
00148     avio_wb16(pb, len);
00149     avio_write(pb, str, len);
00150 }
00151 
00152 static void put_avc_eos_tag(AVIOContext *pb, unsigned ts) {
00153     avio_w8(pb, FLV_TAG_TYPE_VIDEO);
00154     avio_wb24(pb, 5);  /* Tag Data Size */
00155     avio_wb24(pb, ts);  /* lower 24 bits of timestamp in ms*/
00156     avio_w8(pb, (ts >> 24) & 0x7F);  /* MSB of ts in ms*/
00157     avio_wb24(pb, 0);  /* StreamId = 0 */
00158     avio_w8(pb, 23);  /* ub[4] FrameType = 1, ub[4] CodecId = 7 */
00159     avio_w8(pb, 2);  /* AVC end of sequence */
00160     avio_wb24(pb, 0);  /* Always 0 for AVC EOS. */
00161     avio_wb32(pb, 16);  /* Size of FLV tag */
00162 }
00163 
00164 static void put_amf_double(AVIOContext *pb, double d)
00165 {
00166     avio_w8(pb, AMF_DATA_TYPE_NUMBER);
00167     avio_wb64(pb, av_dbl2int(d));
00168 }
00169 
00170 static void put_amf_bool(AVIOContext *pb, int b) {
00171     avio_w8(pb, AMF_DATA_TYPE_BOOL);
00172     avio_w8(pb, !!b);
00173 }
00174 
00175 static int flv_write_header(AVFormatContext *s)
00176 {
00177     AVIOContext *pb = s->pb;
00178     FLVContext *flv = s->priv_data;
00179     AVCodecContext *audio_enc = NULL, *video_enc = NULL;
00180     int i;
00181     double framerate = 0.0;
00182     int64_t metadata_size_pos, data_size;
00183     AVDictionaryEntry *tag = NULL;
00184 
00185     for(i=0; i<s->nb_streams; i++){
00186         AVCodecContext *enc = s->streams[i]->codec;
00187         if (enc->codec_type == AVMEDIA_TYPE_VIDEO) {
00188             if (s->streams[i]->r_frame_rate.den && s->streams[i]->r_frame_rate.num) {
00189                 framerate = av_q2d(s->streams[i]->r_frame_rate);
00190             } else {
00191                 framerate = 1/av_q2d(s->streams[i]->codec->time_base);
00192             }
00193             video_enc = enc;
00194             if(enc->codec_tag == 0) {
00195                 av_log(enc, AV_LOG_ERROR, "video codec not compatible with flv\n");
00196                 return -1;
00197             }
00198         } else {
00199             audio_enc = enc;
00200             if(get_audio_flags(enc)<0)
00201                 return -1;
00202         }
00203         av_set_pts_info(s->streams[i], 32, 1, 1000); /* 32 bit pts in ms */
00204     }
00205     avio_write(pb, "FLV", 3);
00206     avio_w8(pb,1);
00207     avio_w8(pb,   FLV_HEADER_FLAG_HASAUDIO * !!audio_enc
00208                  + FLV_HEADER_FLAG_HASVIDEO * !!video_enc);
00209     avio_wb32(pb,9);
00210     avio_wb32(pb,0);
00211 
00212     for(i=0; i<s->nb_streams; i++){
00213         if(s->streams[i]->codec->codec_tag == 5){
00214             avio_w8(pb,8); // message type
00215             avio_wb24(pb,0); // include flags
00216             avio_wb24(pb,0); // time stamp
00217             avio_wb32(pb,0); // reserved
00218             avio_wb32(pb,11); // size
00219             flv->reserved=5;
00220         }
00221     }
00222 
00223     flv->last_video_ts = -1;
00224 
00225     /* write meta_tag */
00226     avio_w8(pb, 18);         // tag type META
00227     metadata_size_pos= avio_tell(pb);
00228     avio_wb24(pb, 0);          // size of data part (sum of all parts below)
00229     avio_wb24(pb, 0);          // time stamp
00230     avio_wb32(pb, 0);          // reserved
00231 
00232     /* now data of data_size size */
00233 
00234     /* first event name as a string */
00235     avio_w8(pb, AMF_DATA_TYPE_STRING);
00236     put_amf_string(pb, "onMetaData"); // 12 bytes
00237 
00238     /* mixed array (hash) with size and string/type/data tuples */
00239     avio_w8(pb, AMF_DATA_TYPE_MIXEDARRAY);
00240     avio_wb32(pb, 5*!!video_enc + 5*!!audio_enc + 2); // +2 for duration and file size
00241 
00242     put_amf_string(pb, "duration");
00243     flv->duration_offset= avio_tell(pb);
00244     put_amf_double(pb, s->duration / AV_TIME_BASE); // fill in the guessed duration, it'll be corrected later if incorrect
00245 
00246     if(video_enc){
00247         put_amf_string(pb, "width");
00248         put_amf_double(pb, video_enc->width);
00249 
00250         put_amf_string(pb, "height");
00251         put_amf_double(pb, video_enc->height);
00252 
00253         put_amf_string(pb, "videodatarate");
00254         put_amf_double(pb, video_enc->bit_rate / 1024.0);
00255 
00256         put_amf_string(pb, "framerate");
00257         put_amf_double(pb, framerate);
00258 
00259         put_amf_string(pb, "videocodecid");
00260         put_amf_double(pb, video_enc->codec_tag);
00261     }
00262 
00263     if(audio_enc){
00264         put_amf_string(pb, "audiodatarate");
00265         put_amf_double(pb, audio_enc->bit_rate / 1024.0);
00266 
00267         put_amf_string(pb, "audiosamplerate");
00268         put_amf_double(pb, audio_enc->sample_rate);
00269 
00270         put_amf_string(pb, "audiosamplesize");
00271         put_amf_double(pb, audio_enc->codec_id == CODEC_ID_PCM_U8 ? 8 : 16);
00272 
00273         put_amf_string(pb, "stereo");
00274         put_amf_bool(pb, audio_enc->channels == 2);
00275 
00276         put_amf_string(pb, "audiocodecid");
00277         put_amf_double(pb, audio_enc->codec_tag);
00278     }
00279 
00280     while ((tag = av_dict_get(s->metadata, "", tag, AV_DICT_IGNORE_SUFFIX))) {
00281         put_amf_string(pb, tag->key);
00282         avio_w8(pb, AMF_DATA_TYPE_STRING);
00283         put_amf_string(pb, tag->value);
00284     }
00285 
00286     put_amf_string(pb, "filesize");
00287     flv->filesize_offset= avio_tell(pb);
00288     put_amf_double(pb, 0); // delayed write
00289 
00290     put_amf_string(pb, "");
00291     avio_w8(pb, AMF_END_OF_OBJECT);
00292 
00293     /* write total size of tag */
00294     data_size= avio_tell(pb) - metadata_size_pos - 10;
00295     avio_seek(pb, metadata_size_pos, SEEK_SET);
00296     avio_wb24(pb, data_size);
00297     avio_skip(pb, data_size + 10 - 3);
00298     avio_wb32(pb, data_size + 11);
00299 
00300     for (i = 0; i < s->nb_streams; i++) {
00301         AVCodecContext *enc = s->streams[i]->codec;
00302         if (enc->codec_id == CODEC_ID_AAC || enc->codec_id == CODEC_ID_H264) {
00303             int64_t pos;
00304             avio_w8(pb, enc->codec_type == AVMEDIA_TYPE_VIDEO ?
00305                      FLV_TAG_TYPE_VIDEO : FLV_TAG_TYPE_AUDIO);
00306             avio_wb24(pb, 0); // size patched later
00307             avio_wb24(pb, 0); // ts
00308             avio_w8(pb, 0); // ts ext
00309             avio_wb24(pb, 0); // streamid
00310             pos = avio_tell(pb);
00311             if (enc->codec_id == CODEC_ID_AAC) {
00312                 avio_w8(pb, get_audio_flags(enc));
00313                 avio_w8(pb, 0); // AAC sequence header
00314                 avio_write(pb, enc->extradata, enc->extradata_size);
00315             } else {
00316                 avio_w8(pb, enc->codec_tag | FLV_FRAME_KEY); // flags
00317                 avio_w8(pb, 0); // AVC sequence header
00318                 avio_wb24(pb, 0); // composition time
00319                 ff_isom_write_avcc(pb, enc->extradata, enc->extradata_size);
00320             }
00321             data_size = avio_tell(pb) - pos;
00322             avio_seek(pb, -data_size - 10, SEEK_CUR);
00323             avio_wb24(pb, data_size);
00324             avio_skip(pb, data_size + 10 - 3);
00325             avio_wb32(pb, data_size + 11); // previous tag size
00326         }
00327     }
00328 
00329     return 0;
00330 }
00331 
00332 static int flv_write_trailer(AVFormatContext *s)
00333 {
00334     int64_t file_size;
00335 
00336     AVIOContext *pb = s->pb;
00337     FLVContext *flv = s->priv_data;
00338     int i;
00339 
00340     /* Add EOS tag */
00341     for (i = 0; i < s->nb_streams; i++) {
00342         AVCodecContext *enc = s->streams[i]->codec;
00343         if (enc->codec_type == AVMEDIA_TYPE_VIDEO &&
00344                 enc->codec_id == CODEC_ID_H264) {
00345             put_avc_eos_tag(pb, flv->last_video_ts);
00346         }
00347     }
00348 
00349     file_size = avio_tell(pb);
00350 
00351     /* update informations */
00352     avio_seek(pb, flv->duration_offset, SEEK_SET);
00353     put_amf_double(pb, flv->duration / (double)1000);
00354     avio_seek(pb, flv->filesize_offset, SEEK_SET);
00355     put_amf_double(pb, file_size);
00356 
00357     avio_seek(pb, file_size, SEEK_SET);
00358     return 0;
00359 }
00360 
00361 static int flv_write_packet(AVFormatContext *s, AVPacket *pkt)
00362 {
00363     AVIOContext *pb = s->pb;
00364     AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
00365     FLVContext *flv = s->priv_data;
00366     unsigned ts;
00367     int size= pkt->size;
00368     uint8_t *data= NULL;
00369     int flags, flags_size;
00370 
00371 //    av_log(s, AV_LOG_DEBUG, "type:%d pts: %"PRId64" size:%d\n", enc->codec_type, timestamp, size);
00372 
00373     if(enc->codec_id == CODEC_ID_VP6 || enc->codec_id == CODEC_ID_VP6F ||
00374        enc->codec_id == CODEC_ID_AAC)
00375         flags_size= 2;
00376     else if(enc->codec_id == CODEC_ID_H264)
00377         flags_size= 5;
00378     else
00379         flags_size= 1;
00380 
00381     if (enc->codec_type == AVMEDIA_TYPE_VIDEO) {
00382         avio_w8(pb, FLV_TAG_TYPE_VIDEO);
00383 
00384         flags = enc->codec_tag;
00385         if(flags == 0) {
00386             av_log(enc, AV_LOG_ERROR, "video codec %X not compatible with flv\n",enc->codec_id);
00387             return -1;
00388         }
00389 
00390         flags |= pkt->flags & AV_PKT_FLAG_KEY ? FLV_FRAME_KEY : FLV_FRAME_INTER;
00391     } else {
00392         assert(enc->codec_type == AVMEDIA_TYPE_AUDIO);
00393         flags = get_audio_flags(enc);
00394 
00395         assert(size);
00396 
00397         avio_w8(pb, FLV_TAG_TYPE_AUDIO);
00398     }
00399 
00400     if (enc->codec_id == CODEC_ID_H264) {
00401         /* check if extradata looks like mp4 formated */
00402         if (enc->extradata_size > 0 && *(uint8_t*)enc->extradata != 1) {
00403             if (ff_avc_parse_nal_units_buf(pkt->data, &data, &size) < 0)
00404                 return -1;
00405         }
00406         if (!flv->delay && pkt->dts < 0)
00407             flv->delay = -pkt->dts;
00408     } else if (enc->codec_id == CODEC_ID_AAC && pkt->size > 2 &&
00409                (AV_RB16(pkt->data) & 0xfff0) == 0xfff0) {
00410         av_log(s, AV_LOG_ERROR, "malformated aac bitstream, use -absf aac_adtstoasc\n");
00411         return -1;
00412     }
00413 
00414     ts = pkt->dts + flv->delay; // add delay to force positive dts
00415     if (enc->codec_type == AVMEDIA_TYPE_VIDEO) {
00416         if (flv->last_video_ts < ts)
00417             flv->last_video_ts = ts;
00418     }
00419     avio_wb24(pb,size + flags_size);
00420     avio_wb24(pb,ts);
00421     avio_w8(pb,(ts >> 24) & 0x7F); // timestamps are 32bits _signed_
00422     avio_wb24(pb,flv->reserved);
00423     avio_w8(pb,flags);
00424     if (enc->codec_id == CODEC_ID_VP6)
00425         avio_w8(pb,0);
00426     if (enc->codec_id == CODEC_ID_VP6F)
00427         avio_w8(pb, enc->extradata_size ? enc->extradata[0] : 0);
00428     else if (enc->codec_id == CODEC_ID_AAC)
00429         avio_w8(pb,1); // AAC raw
00430     else if (enc->codec_id == CODEC_ID_H264) {
00431         avio_w8(pb,1); // AVC NALU
00432         avio_wb24(pb,pkt->pts - pkt->dts);
00433     }
00434 
00435     avio_write(pb, data ? data : pkt->data, size);
00436 
00437     avio_wb32(pb,size+flags_size+11); // previous tag size
00438     flv->duration = FFMAX(flv->duration, pkt->pts + flv->delay + pkt->duration);
00439 
00440     avio_flush(pb);
00441 
00442     av_free(data);
00443 
00444     return pb->error;
00445 }
00446 
00447 AVOutputFormat ff_flv_muxer = {
00448     "flv",
00449     NULL_IF_CONFIG_SMALL("FLV format"),
00450     "video/x-flv",
00451     "flv",
00452     sizeof(FLVContext),
00453 #if CONFIG_LIBMP3LAME
00454     CODEC_ID_MP3,
00455 #else // CONFIG_LIBMP3LAME
00456     CODEC_ID_ADPCM_SWF,
00457 #endif // CONFIG_LIBMP3LAME
00458     CODEC_ID_FLV1,
00459     flv_write_header,
00460     flv_write_packet,
00461     flv_write_trailer,
00462     .codec_tag= (const AVCodecTag* const []){flv_video_codec_ids, flv_audio_codec_ids, 0},
00463     .flags= AVFMT_GLOBALHEADER | AVFMT_VARIABLE_FPS,
00464 };