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 #include "avformat.h"
00022 #include "flv.h"
00023 #include "riff.h"
00024 #include "avc.h"
00025 
00026 #undef NDEBUG
00027 #include <assert.h>
00028 
00029 static const AVCodecTag flv_video_codec_ids[] = {
00030     {CODEC_ID_FLV1,    FLV_CODECID_H263  },
00031     {CODEC_ID_FLASHSV, FLV_CODECID_SCREEN},
00032     {CODEC_ID_FLASHSV2, FLV_CODECID_SCREEN2},
00033     {CODEC_ID_VP6F,    FLV_CODECID_VP6   },
00034     {CODEC_ID_VP6,     FLV_CODECID_VP6   },
00035     {CODEC_ID_H264,    FLV_CODECID_H264  },
00036     {CODEC_ID_NONE,    0}
00037 };
00038 
00039 static const AVCodecTag flv_audio_codec_ids[] = {
00040     {CODEC_ID_MP3,       FLV_CODECID_MP3    >> FLV_AUDIO_CODECID_OFFSET},
00041     {CODEC_ID_PCM_U8,    FLV_CODECID_PCM    >> FLV_AUDIO_CODECID_OFFSET},
00042     {CODEC_ID_PCM_S16BE, FLV_CODECID_PCM    >> FLV_AUDIO_CODECID_OFFSET},
00043     {CODEC_ID_PCM_S16LE, FLV_CODECID_PCM_LE >> FLV_AUDIO_CODECID_OFFSET},
00044     {CODEC_ID_ADPCM_SWF, FLV_CODECID_ADPCM  >> FLV_AUDIO_CODECID_OFFSET},
00045     {CODEC_ID_AAC,       FLV_CODECID_AAC    >> FLV_AUDIO_CODECID_OFFSET},
00046     {CODEC_ID_NELLYMOSER, FLV_CODECID_NELLYMOSER >> FLV_AUDIO_CODECID_OFFSET},
00047     {CODEC_ID_SPEEX,     FLV_CODECID_SPEEX  >> FLV_AUDIO_CODECID_OFFSET},
00048     {CODEC_ID_NONE,      0}
00049 };
00050 
00051 typedef struct FLVContext {
00052     int reserved;
00053     int64_t duration_offset;
00054     int64_t filesize_offset;
00055     int64_t duration;
00056     int delay; 
00057 } FLVContext;
00058 
00059 static int get_audio_flags(AVCodecContext *enc){
00060     int flags = (enc->bits_per_coded_sample == 16) ? FLV_SAMPLESSIZE_16BIT : FLV_SAMPLESSIZE_8BIT;
00061 
00062     if (enc->codec_id == CODEC_ID_AAC) // specs force these parameters
00063         return FLV_CODECID_AAC | FLV_SAMPLERATE_44100HZ | FLV_SAMPLESSIZE_16BIT | FLV_STEREO;
00064     else if (enc->codec_id == CODEC_ID_SPEEX) {
00065         if (enc->sample_rate != 16000) {
00066             av_log(enc, AV_LOG_ERROR, "flv only supports wideband (16kHz) Speex audio\n");
00067             return -1;
00068         }
00069         if (enc->channels != 1) {
00070             av_log(enc, AV_LOG_ERROR, "flv only supports mono Speex audio\n");
00071             return -1;
00072         }
00073         if (enc->frame_size / 320 > 8) {
00074             av_log(enc, AV_LOG_WARNING, "Warning: Speex stream has more than "
00075                                         "8 frames per packet. Adobe Flash "
00076                                         "Player cannot handle this!\n");
00077         }
00078         return FLV_CODECID_SPEEX | FLV_SAMPLERATE_11025HZ | FLV_SAMPLESSIZE_16BIT;
00079     } else {
00080     switch (enc->sample_rate) {
00081         case    44100:
00082             flags |= FLV_SAMPLERATE_44100HZ;
00083             break;
00084         case    22050:
00085             flags |= FLV_SAMPLERATE_22050HZ;
00086             break;
00087         case    11025:
00088             flags |= FLV_SAMPLERATE_11025HZ;
00089             break;
00090         case     8000: //nellymoser only
00091         case     5512: //not mp3
00092             if(enc->codec_id != CODEC_ID_MP3){
00093                 flags |= FLV_SAMPLERATE_SPECIAL;
00094                 break;
00095             }
00096         default:
00097             av_log(enc, AV_LOG_ERROR, "flv does not support that sample rate, choose from (44100, 22050, 11025).\n");
00098             return -1;
00099     }
00100     }
00101 
00102     if (enc->channels > 1) {
00103         flags |= FLV_STEREO;
00104     }
00105 
00106     switch(enc->codec_id){
00107     case CODEC_ID_MP3:
00108         flags |= FLV_CODECID_MP3    | FLV_SAMPLESSIZE_16BIT;
00109         break;
00110     case CODEC_ID_PCM_U8:
00111         flags |= FLV_CODECID_PCM    | FLV_SAMPLESSIZE_8BIT;
00112         break;
00113     case CODEC_ID_PCM_S16BE:
00114         flags |= FLV_CODECID_PCM    | FLV_SAMPLESSIZE_16BIT;
00115         break;
00116     case CODEC_ID_PCM_S16LE:
00117         flags |= FLV_CODECID_PCM_LE | FLV_SAMPLESSIZE_16BIT;
00118         break;
00119     case CODEC_ID_ADPCM_SWF:
00120         flags |= FLV_CODECID_ADPCM | FLV_SAMPLESSIZE_16BIT;
00121         break;
00122     case CODEC_ID_NELLYMOSER:
00123         if (enc->sample_rate == 8000) {
00124             flags |= FLV_CODECID_NELLYMOSER_8KHZ_MONO | FLV_SAMPLESSIZE_16BIT;
00125         } else {
00126             flags |= FLV_CODECID_NELLYMOSER | FLV_SAMPLESSIZE_16BIT;
00127         }
00128         break;
00129     case 0:
00130         flags |= enc->codec_tag<<4;
00131         break;
00132     default:
00133         av_log(enc, AV_LOG_ERROR, "codec not compatible with flv\n");
00134         return -1;
00135     }
00136 
00137     return flags;
00138 }
00139 
00140 static void put_amf_string(ByteIOContext *pb, const char *str)
00141 {
00142     size_t len = strlen(str);
00143     put_be16(pb, len);
00144     put_buffer(pb, str, len);
00145 }
00146 
00147 static void put_amf_double(ByteIOContext *pb, double d)
00148 {
00149     put_byte(pb, AMF_DATA_TYPE_NUMBER);
00150     put_be64(pb, av_dbl2int(d));
00151 }
00152 
00153 static void put_amf_bool(ByteIOContext *pb, int b) {
00154     put_byte(pb, AMF_DATA_TYPE_BOOL);
00155     put_byte(pb, !!b);
00156 }
00157 
00158 static int flv_write_header(AVFormatContext *s)
00159 {
00160     ByteIOContext *pb = s->pb;
00161     FLVContext *flv = s->priv_data;
00162     AVCodecContext *audio_enc = NULL, *video_enc = NULL;
00163     int i;
00164     double framerate = 0.0;
00165     int64_t metadata_size_pos, data_size;
00166 
00167     for(i=0; i<s->nb_streams; i++){
00168         AVCodecContext *enc = s->streams[i]->codec;
00169         if (enc->codec_type == AVMEDIA_TYPE_VIDEO) {
00170             if (s->streams[i]->r_frame_rate.den && s->streams[i]->r_frame_rate.num) {
00171                 framerate = av_q2d(s->streams[i]->r_frame_rate);
00172             } else {
00173                 framerate = 1/av_q2d(s->streams[i]->codec->time_base);
00174             }
00175             video_enc = enc;
00176             if(enc->codec_tag == 0) {
00177                 av_log(enc, AV_LOG_ERROR, "video codec not compatible with flv\n");
00178                 return -1;
00179             }
00180         } else {
00181             audio_enc = enc;
00182             if(get_audio_flags(enc)<0)
00183                 return -1;
00184         }
00185         av_set_pts_info(s->streams[i], 32, 1, 1000); /* 32 bit pts in ms */
00186     }
00187     put_tag(pb,"FLV");
00188     put_byte(pb,1);
00189     put_byte(pb,   FLV_HEADER_FLAG_HASAUDIO * !!audio_enc
00190                  + FLV_HEADER_FLAG_HASVIDEO * !!video_enc);
00191     put_be32(pb,9);
00192     put_be32(pb,0);
00193 
00194     for(i=0; i<s->nb_streams; i++){
00195         if(s->streams[i]->codec->codec_tag == 5){
00196             put_byte(pb,8); // message type
00197             put_be24(pb,0); // include flags
00198             put_be24(pb,0); // time stamp
00199             put_be32(pb,0); // reserved
00200             put_be32(pb,11); // size
00201             flv->reserved=5;
00202         }
00203     }
00204 
00205     /* write meta_tag */
00206     put_byte(pb, 18);         // tag type META
00207     metadata_size_pos= url_ftell(pb);
00208     put_be24(pb, 0);          // size of data part (sum of all parts below)
00209     put_be24(pb, 0);          // time stamp
00210     put_be32(pb, 0);          // reserved
00211 
00212     /* now data of data_size size */
00213 
00214     /* first event name as a string */
00215     put_byte(pb, AMF_DATA_TYPE_STRING);
00216     put_amf_string(pb, "onMetaData"); // 12 bytes
00217 
00218     /* mixed array (hash) with size and string/type/data tuples */
00219     put_byte(pb, AMF_DATA_TYPE_MIXEDARRAY);
00220     put_be32(pb, 5*!!video_enc + 5*!!audio_enc + 2); // +2 for duration and file size
00221 
00222     put_amf_string(pb, "duration");
00223     flv->duration_offset= url_ftell(pb);
00224     put_amf_double(pb, s->duration / AV_TIME_BASE); // fill in the guessed duration, it'll be corrected later if incorrect
00225 
00226     if(video_enc){
00227         put_amf_string(pb, "width");
00228         put_amf_double(pb, video_enc->width);
00229 
00230         put_amf_string(pb, "height");
00231         put_amf_double(pb, video_enc->height);
00232 
00233         put_amf_string(pb, "videodatarate");
00234         put_amf_double(pb, video_enc->bit_rate / 1024.0);
00235 
00236         put_amf_string(pb, "framerate");
00237         put_amf_double(pb, framerate);
00238 
00239         put_amf_string(pb, "videocodecid");
00240         put_amf_double(pb, video_enc->codec_tag);
00241     }
00242 
00243     if(audio_enc){
00244         put_amf_string(pb, "audiodatarate");
00245         put_amf_double(pb, audio_enc->bit_rate / 1024.0);
00246 
00247         put_amf_string(pb, "audiosamplerate");
00248         put_amf_double(pb, audio_enc->sample_rate);
00249 
00250         put_amf_string(pb, "audiosamplesize");
00251         put_amf_double(pb, audio_enc->codec_id == CODEC_ID_PCM_U8 ? 8 : 16);
00252 
00253         put_amf_string(pb, "stereo");
00254         put_amf_bool(pb, audio_enc->channels == 2);
00255 
00256         put_amf_string(pb, "audiocodecid");
00257         put_amf_double(pb, audio_enc->codec_tag);
00258     }
00259 
00260     put_amf_string(pb, "filesize");
00261     flv->filesize_offset= url_ftell(pb);
00262     put_amf_double(pb, 0); // delayed write
00263 
00264     put_amf_string(pb, "");
00265     put_byte(pb, AMF_END_OF_OBJECT);
00266 
00267     /* write total size of tag */
00268     data_size= url_ftell(pb) - metadata_size_pos - 10;
00269     url_fseek(pb, metadata_size_pos, SEEK_SET);
00270     put_be24(pb, data_size);
00271     url_fseek(pb, data_size + 10 - 3, SEEK_CUR);
00272     put_be32(pb, data_size + 11);
00273 
00274     for (i = 0; i < s->nb_streams; i++) {
00275         AVCodecContext *enc = s->streams[i]->codec;
00276         if (enc->codec_id == CODEC_ID_AAC || enc->codec_id == CODEC_ID_H264) {
00277             int64_t pos;
00278             put_byte(pb, enc->codec_type == AVMEDIA_TYPE_VIDEO ?
00279                      FLV_TAG_TYPE_VIDEO : FLV_TAG_TYPE_AUDIO);
00280             put_be24(pb, 0); // size patched later
00281             put_be24(pb, 0); // ts
00282             put_byte(pb, 0); // ts ext
00283             put_be24(pb, 0); // streamid
00284             pos = url_ftell(pb);
00285             if (enc->codec_id == CODEC_ID_AAC) {
00286                 put_byte(pb, get_audio_flags(enc));
00287                 put_byte(pb, 0); // AAC sequence header
00288                 put_buffer(pb, enc->extradata, enc->extradata_size);
00289             } else {
00290                 put_byte(pb, enc->codec_tag | FLV_FRAME_KEY); // flags
00291                 put_byte(pb, 0); // AVC sequence header
00292                 put_be24(pb, 0); // composition time
00293                 ff_isom_write_avcc(pb, enc->extradata, enc->extradata_size);
00294             }
00295             data_size = url_ftell(pb) - pos;
00296             url_fseek(pb, -data_size - 10, SEEK_CUR);
00297             put_be24(pb, data_size);
00298             url_fseek(pb, data_size + 10 - 3, SEEK_CUR);
00299             put_be32(pb, data_size + 11); // previous tag size
00300         }
00301     }
00302 
00303     return 0;
00304 }
00305 
00306 static int flv_write_trailer(AVFormatContext *s)
00307 {
00308     int64_t file_size;
00309 
00310     ByteIOContext *pb = s->pb;
00311     FLVContext *flv = s->priv_data;
00312 
00313     file_size = url_ftell(pb);
00314 
00315     /* update informations */
00316     url_fseek(pb, flv->duration_offset, SEEK_SET);
00317     put_amf_double(pb, flv->duration / (double)1000);
00318     url_fseek(pb, flv->filesize_offset, SEEK_SET);
00319     put_amf_double(pb, file_size);
00320 
00321     url_fseek(pb, file_size, SEEK_SET);
00322     return 0;
00323 }
00324 
00325 static int flv_write_packet(AVFormatContext *s, AVPacket *pkt)
00326 {
00327     ByteIOContext *pb = s->pb;
00328     AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
00329     FLVContext *flv = s->priv_data;
00330     unsigned ts;
00331     int size= pkt->size;
00332     uint8_t *data= NULL;
00333     int flags, flags_size;
00334 
00335 //    av_log(s, AV_LOG_DEBUG, "type:%d pts: %"PRId64" size:%d\n", enc->codec_type, timestamp, size);
00336 
00337     if(enc->codec_id == CODEC_ID_VP6 || enc->codec_id == CODEC_ID_VP6F ||
00338        enc->codec_id == CODEC_ID_AAC)
00339         flags_size= 2;
00340     else if(enc->codec_id == CODEC_ID_H264)
00341         flags_size= 5;
00342     else
00343         flags_size= 1;
00344 
00345     if (enc->codec_type == AVMEDIA_TYPE_VIDEO) {
00346         put_byte(pb, FLV_TAG_TYPE_VIDEO);
00347 
00348         flags = enc->codec_tag;
00349         if(flags == 0) {
00350             av_log(enc, AV_LOG_ERROR, "video codec %X not compatible with flv\n",enc->codec_id);
00351             return -1;
00352         }
00353 
00354         flags |= pkt->flags & AV_PKT_FLAG_KEY ? FLV_FRAME_KEY : FLV_FRAME_INTER;
00355     } else {
00356         assert(enc->codec_type == AVMEDIA_TYPE_AUDIO);
00357         flags = get_audio_flags(enc);
00358 
00359         assert(size);
00360 
00361         put_byte(pb, FLV_TAG_TYPE_AUDIO);
00362     }
00363 
00364     if (enc->codec_id == CODEC_ID_H264) {
00365         /* check if extradata looks like mp4 formated */
00366         if (enc->extradata_size > 0 && *(uint8_t*)enc->extradata != 1) {
00367             if (ff_avc_parse_nal_units_buf(pkt->data, &data, &size) < 0)
00368                 return -1;
00369         }
00370         if (!flv->delay && pkt->dts < 0)
00371             flv->delay = -pkt->dts;
00372     }
00373 
00374     ts = pkt->dts + flv->delay; // add delay to force positive dts
00375     put_be24(pb,size + flags_size);
00376     put_be24(pb,ts);
00377     put_byte(pb,(ts >> 24) & 0x7F); // timestamps are 32bits _signed_
00378     put_be24(pb,flv->reserved);
00379     put_byte(pb,flags);
00380     if (enc->codec_id == CODEC_ID_VP6)
00381         put_byte(pb,0);
00382     if (enc->codec_id == CODEC_ID_VP6F)
00383         put_byte(pb, enc->extradata_size ? enc->extradata[0] : 0);
00384     else if (enc->codec_id == CODEC_ID_AAC)
00385         put_byte(pb,1); // AAC raw
00386     else if (enc->codec_id == CODEC_ID_H264) {
00387         put_byte(pb,1); // AVC NALU
00388         put_be24(pb,pkt->pts - pkt->dts);
00389     }
00390 
00391     put_buffer(pb, data ? data : pkt->data, size);
00392 
00393     put_be32(pb,size+flags_size+11); // previous tag size
00394     flv->duration = FFMAX(flv->duration, pkt->pts + flv->delay + pkt->duration);
00395 
00396     put_flush_packet(pb);
00397 
00398     av_free(data);
00399 
00400     return 0;
00401 }
00402 
00403 AVOutputFormat flv_muxer = {
00404     "flv",
00405     NULL_IF_CONFIG_SMALL("FLV format"),
00406     "video/x-flv",
00407     "flv",
00408     sizeof(FLVContext),
00409 #if CONFIG_LIBMP3LAME
00410     CODEC_ID_MP3,
00411 #else // CONFIG_LIBMP3LAME
00412     CODEC_ID_ADPCM_SWF,
00413 #endif // CONFIG_LIBMP3LAME
00414     CODEC_ID_FLV1,
00415     flv_write_header,
00416     flv_write_packet,
00417     flv_write_trailer,
00418     .codec_tag= (const AVCodecTag* const []){flv_video_codec_ids, flv_audio_codec_ids, 0},
00419     .flags= AVFMT_GLOBALHEADER | AVFMT_VARIABLE_FPS,
00420 };