FFmpeg: libavformat/rmenc.c Source File

00001 /*
00002  * "Real" compatible muxer.
00003  * Copyright (c) 2000, 2001 Fabrice Bellard
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 "rm.h"
00023 
00024 typedef struct {
00025     int nb_packets;
00026     int packet_total_size;
00027     int packet_max_size;
00028     /* codec related output */
00029     int bit_rate;
00030     float frame_rate;
00031     int nb_frames;    /* current frame number */
00032     int total_frames; /* total number of frames */
00033     int num;
00034     AVCodecContext *enc;
00035 } StreamInfo;
00036 
00037 typedef struct {
00038     StreamInfo streams[2];
00039     StreamInfo *audio_stream, *video_stream;
00040     int data_pos; /* position of the data after the header */
00041 } RMMuxContext;
00042 
00043 /* in ms */
00044 #define BUFFER_DURATION 0
00045 
00046 
00047 static void put_str(ByteIOContext *s, const char *tag)
00048 {
00049     put_be16(s,strlen(tag));
00050     while (*tag) {
00051         put_byte(s, *tag++);
00052     }
00053 }
00054 
00055 static void put_str8(ByteIOContext *s, const char *tag)
00056 {
00057     put_byte(s, strlen(tag));
00058     while (*tag) {
00059         put_byte(s, *tag++);
00060     }
00061 }
00062 
00063 static void rv10_write_header(AVFormatContext *ctx,
00064                               int data_size, int index_pos)
00065 {
00066     RMMuxContext *rm = ctx->priv_data;
00067     ByteIOContext *s = ctx->pb;
00068     StreamInfo *stream;
00069     unsigned char *data_offset_ptr, *start_ptr;
00070     const char *desc, *mimetype;
00071     int nb_packets, packet_total_size, packet_max_size, size, packet_avg_size, i;
00072     int bit_rate, v, duration, flags, data_pos;
00073     AVMetadataTag *tag;
00074 
00075     start_ptr = s->buf_ptr;
00076 
00077     put_tag(s, ".RMF");
00078     put_be32(s,18); /* header size */
00079     put_be16(s,0);
00080     put_be32(s,0);
00081     put_be32(s,4 + ctx->nb_streams); /* num headers */
00082 
00083     put_tag(s,"PROP");
00084     put_be32(s, 50);
00085     put_be16(s, 0);
00086     packet_max_size = 0;
00087     packet_total_size = 0;
00088     nb_packets = 0;
00089     bit_rate = 0;
00090     duration = 0;
00091     for(i=0;i<ctx->nb_streams;i++) {
00092         StreamInfo *stream = &rm->streams[i];
00093         bit_rate += stream->bit_rate;
00094         if (stream->packet_max_size > packet_max_size)
00095             packet_max_size = stream->packet_max_size;
00096         nb_packets += stream->nb_packets;
00097         packet_total_size += stream->packet_total_size;
00098         /* select maximum duration */
00099         v = (int) (1000.0 * (float)stream->total_frames / stream->frame_rate);
00100         if (v > duration)
00101             duration = v;
00102     }
00103     put_be32(s, bit_rate); /* max bit rate */
00104     put_be32(s, bit_rate); /* avg bit rate */
00105     put_be32(s, packet_max_size);        /* max packet size */
00106     if (nb_packets > 0)
00107         packet_avg_size = packet_total_size / nb_packets;
00108     else
00109         packet_avg_size = 0;
00110     put_be32(s, packet_avg_size);        /* avg packet size */
00111     put_be32(s, nb_packets);  /* num packets */
00112     put_be32(s, duration); /* duration */
00113     put_be32(s, BUFFER_DURATION);           /* preroll */
00114     put_be32(s, index_pos);           /* index offset */
00115     /* computation of data the data offset */
00116     data_offset_ptr = s->buf_ptr;
00117     put_be32(s, 0);           /* data offset : will be patched after */
00118     put_be16(s, ctx->nb_streams);    /* num streams */
00119     flags = 1 | 2; /* save allowed & perfect play */
00120     if (url_is_streamed(s))
00121         flags |= 4; /* live broadcast */
00122     put_be16(s, flags);
00123 
00124     /* comments */
00125 
00126     put_tag(s,"CONT");
00127     size =  4 * 2 + 10;
00128     for(i=0; i<FF_ARRAY_ELEMS(ff_rm_metadata); i++) {
00129         tag = av_metadata_get(ctx->metadata, ff_rm_metadata[i], NULL, 0);
00130         if(tag) size += strlen(tag->value);
00131     }
00132     put_be32(s,size);
00133     put_be16(s,0);
00134     for(i=0; i<FF_ARRAY_ELEMS(ff_rm_metadata); i++) {
00135         tag = av_metadata_get(ctx->metadata, ff_rm_metadata[i], NULL, 0);
00136         put_str(s, tag ? tag->value : "");
00137     }
00138 
00139     for(i=0;i<ctx->nb_streams;i++) {
00140         int codec_data_size;
00141 
00142         stream = &rm->streams[i];
00143 
00144         if (stream->enc->codec_type == CODEC_TYPE_AUDIO) {
00145             desc = "The Audio Stream";
00146             mimetype = "audio/x-pn-realaudio";
00147             codec_data_size = 73;
00148         } else {
00149             desc = "The Video Stream";
00150             mimetype = "video/x-pn-realvideo";
00151             codec_data_size = 34;
00152         }
00153 
00154         put_tag(s,"MDPR");
00155         size = 10 + 9 * 4 + strlen(desc) + strlen(mimetype) + codec_data_size;
00156         put_be32(s, size);
00157         put_be16(s, 0);
00158 
00159         put_be16(s, i); /* stream number */
00160         put_be32(s, stream->bit_rate); /* max bit rate */
00161         put_be32(s, stream->bit_rate); /* avg bit rate */
00162         put_be32(s, stream->packet_max_size);        /* max packet size */
00163         if (stream->nb_packets > 0)
00164             packet_avg_size = stream->packet_total_size /
00165                 stream->nb_packets;
00166         else
00167             packet_avg_size = 0;
00168         put_be32(s, packet_avg_size);        /* avg packet size */
00169         put_be32(s, 0);           /* start time */
00170         put_be32(s, BUFFER_DURATION);           /* preroll */
00171         /* duration */
00172         if (url_is_streamed(s) || !stream->total_frames)
00173             put_be32(s, (int)(3600 * 1000));
00174         else
00175             put_be32(s, (int)(stream->total_frames * 1000 / stream->frame_rate));
00176         put_str8(s, desc);
00177         put_str8(s, mimetype);
00178         put_be32(s, codec_data_size);
00179 
00180         if (stream->enc->codec_type == CODEC_TYPE_AUDIO) {
00181             int coded_frame_size, fscode, sample_rate;
00182             sample_rate = stream->enc->sample_rate;
00183             coded_frame_size = (stream->enc->bit_rate *
00184                                 stream->enc->frame_size) / (8 * sample_rate);
00185             /* audio codec info */
00186             put_tag(s, ".ra");
00187             put_byte(s, 0xfd);
00188             put_be32(s, 0x00040000); /* version */
00189             put_tag(s, ".ra4");
00190             put_be32(s, 0x01b53530); /* stream length */
00191             put_be16(s, 4); /* unknown */
00192             put_be32(s, 0x39); /* header size */
00193 
00194             switch(sample_rate) {
00195             case 48000:
00196             case 24000:
00197             case 12000:
00198                 fscode = 1;
00199                 break;
00200             default:
00201             case 44100:
00202             case 22050:
00203             case 11025:
00204                 fscode = 2;
00205                 break;
00206             case 32000:
00207             case 16000:
00208             case 8000:
00209                 fscode = 3;
00210             }
00211             put_be16(s, fscode); /* codec additional info, for AC-3, seems
00212                                      to be a frequency code */
00213             /* special hack to compensate rounding errors... */
00214             if (coded_frame_size == 557)
00215                 coded_frame_size--;
00216             put_be32(s, coded_frame_size); /* frame length */
00217             put_be32(s, 0x51540); /* unknown */
00218             put_be32(s, 0x249f0); /* unknown */
00219             put_be32(s, 0x249f0); /* unknown */
00220             put_be16(s, 0x01);
00221             /* frame length : seems to be very important */
00222             put_be16(s, coded_frame_size);
00223             put_be32(s, 0); /* unknown */
00224             put_be16(s, stream->enc->sample_rate); /* sample rate */
00225             put_be32(s, 0x10); /* unknown */
00226             put_be16(s, stream->enc->channels);
00227             put_str8(s, "Int0"); /* codec name */
00228             put_str8(s, "dnet"); /* codec name */
00229             put_be16(s, 0); /* title length */
00230             put_be16(s, 0); /* author length */
00231             put_be16(s, 0); /* copyright length */
00232             put_byte(s, 0); /* end of header */
00233         } else {
00234             /* video codec info */
00235             put_be32(s,34); /* size */
00236             if(stream->enc->codec_id == CODEC_ID_RV10)
00237                 put_tag(s,"VIDORV10");
00238             else
00239                 put_tag(s,"VIDORV20");
00240             put_be16(s, stream->enc->width);
00241             put_be16(s, stream->enc->height);
00242             put_be16(s, (int) stream->frame_rate); /* frames per seconds ? */
00243             put_be32(s,0);     /* unknown meaning */
00244             put_be16(s, (int) stream->frame_rate);  /* unknown meaning */
00245             put_be32(s,0);     /* unknown meaning */
00246             put_be16(s, 8);    /* unknown meaning */
00247             /* Seems to be the codec version: only use basic H263. The next
00248                versions seems to add a diffential DC coding as in
00249                MPEG... nothing new under the sun */
00250             if(stream->enc->codec_id == CODEC_ID_RV10)
00251                 put_be32(s,0x10000000);
00252             else
00253                 put_be32(s,0x20103001);
00254             //put_be32(s,0x10003000);
00255         }
00256     }
00257 
00258     /* patch data offset field */
00259     data_pos = s->buf_ptr - start_ptr;
00260     rm->data_pos = data_pos;
00261     data_offset_ptr[0] = data_pos >> 24;
00262     data_offset_ptr[1] = data_pos >> 16;
00263     data_offset_ptr[2] = data_pos >> 8;
00264     data_offset_ptr[3] = data_pos;
00265 
00266     /* data stream */
00267     put_tag(s,"DATA");
00268     put_be32(s,data_size + 10 + 8);
00269     put_be16(s,0);
00270 
00271     put_be32(s, nb_packets); /* number of packets */
00272     put_be32(s,0); /* next data header */
00273 }
00274 
00275 static void write_packet_header(AVFormatContext *ctx, StreamInfo *stream,
00276                                 int length, int key_frame)
00277 {
00278     int timestamp;
00279     ByteIOContext *s = ctx->pb;
00280 
00281     stream->nb_packets++;
00282     stream->packet_total_size += length;
00283     if (length > stream->packet_max_size)
00284         stream->packet_max_size =  length;
00285 
00286     put_be16(s,0); /* version */
00287     put_be16(s,length + 12);
00288     put_be16(s, stream->num); /* stream number */
00289     timestamp = (1000 * (float)stream->nb_frames) / stream->frame_rate;
00290     put_be32(s, timestamp); /* timestamp */
00291     put_byte(s, 0); /* reserved */
00292     put_byte(s, key_frame ? 2 : 0); /* flags */
00293 }
00294 
00295 static int rm_write_header(AVFormatContext *s)
00296 {
00297     RMMuxContext *rm = s->priv_data;
00298     StreamInfo *stream;
00299     int n;
00300     AVCodecContext *codec;
00301 
00302     for(n=0;n<s->nb_streams;n++) {
00303         s->streams[n]->id = n;
00304         codec = s->streams[n]->codec;
00305         stream = &rm->streams[n];
00306         memset(stream, 0, sizeof(StreamInfo));
00307         stream->num = n;
00308         stream->bit_rate = codec->bit_rate;
00309         stream->enc = codec;
00310 
00311         switch(codec->codec_type) {
00312         case CODEC_TYPE_AUDIO:
00313             rm->audio_stream = stream;
00314             stream->frame_rate = (float)codec->sample_rate / (float)codec->frame_size;
00315             /* XXX: dummy values */
00316             stream->packet_max_size = 1024;
00317             stream->nb_packets = 0;
00318             stream->total_frames = stream->nb_packets;
00319             break;
00320         case CODEC_TYPE_VIDEO:
00321             rm->video_stream = stream;
00322             stream->frame_rate = (float)codec->time_base.den / (float)codec->time_base.num;
00323             /* XXX: dummy values */
00324             stream->packet_max_size = 4096;
00325             stream->nb_packets = 0;
00326             stream->total_frames = stream->nb_packets;
00327             break;
00328         default:
00329             return -1;
00330         }
00331     }
00332 
00333     rv10_write_header(s, 0, 0);
00334     put_flush_packet(s->pb);
00335     return 0;
00336 }
00337 
00338 static int rm_write_audio(AVFormatContext *s, const uint8_t *buf, int size, int flags)
00339 {
00340     uint8_t *buf1;
00341     RMMuxContext *rm = s->priv_data;
00342     ByteIOContext *pb = s->pb;
00343     StreamInfo *stream = rm->audio_stream;
00344     int i;
00345 
00346     /* XXX: suppress this malloc */
00347     buf1= (uint8_t*) av_malloc( size * sizeof(uint8_t) );
00348 
00349     write_packet_header(s, stream, size, !!(flags & PKT_FLAG_KEY));
00350 
00351     /* for AC-3, the words seem to be reversed */
00352     for(i=0;i<size;i+=2) {
00353         buf1[i] = buf[i+1];
00354         buf1[i+1] = buf[i];
00355     }
00356     put_buffer(pb, buf1, size);
00357     put_flush_packet(pb);
00358     stream->nb_frames++;
00359     av_free(buf1);
00360     return 0;
00361 }
00362 
00363 static int rm_write_video(AVFormatContext *s, const uint8_t *buf, int size, int flags)
00364 {
00365     RMMuxContext *rm = s->priv_data;
00366     ByteIOContext *pb = s->pb;
00367     StreamInfo *stream = rm->video_stream;
00368     int key_frame = !!(flags & PKT_FLAG_KEY);
00369 
00370     /* XXX: this is incorrect: should be a parameter */
00371 
00372     /* Well, I spent some time finding the meaning of these bits. I am
00373        not sure I understood everything, but it works !! */
00374 #if 1
00375     write_packet_header(s, stream, size + 7 + (size >= 0x4000)*4, key_frame);
00376     /* bit 7: '1' if final packet of a frame converted in several packets */
00377     put_byte(pb, 0x81);
00378     /* bit 7: '1' if I frame. bits 6..0 : sequence number in current
00379        frame starting from 1 */
00380     if (key_frame) {
00381         put_byte(pb, 0x81);
00382     } else {
00383         put_byte(pb, 0x01);
00384     }
00385     if(size >= 0x4000){
00386         put_be32(pb, size); /* total frame size */
00387         put_be32(pb, size); /* offset from the start or the end */
00388     }else{
00389         put_be16(pb, 0x4000 | size); /* total frame size */
00390         put_be16(pb, 0x4000 | size); /* offset from the start or the end */
00391     }
00392 #else
00393     /* full frame */
00394     write_packet_header(s, size + 6);
00395     put_byte(pb, 0xc0);
00396     put_be16(pb, 0x4000 + size); /* total frame size */
00397     put_be16(pb, 0x4000 + packet_number * 126); /* position in stream */
00398 #endif
00399     put_byte(pb, stream->nb_frames & 0xff);
00400 
00401     put_buffer(pb, buf, size);
00402     put_flush_packet(pb);
00403 
00404     stream->nb_frames++;
00405     return 0;
00406 }
00407 
00408 static int rm_write_packet(AVFormatContext *s, AVPacket *pkt)
00409 {
00410     if (s->streams[pkt->stream_index]->codec->codec_type ==
00411         CODEC_TYPE_AUDIO)
00412         return rm_write_audio(s, pkt->data, pkt->size, pkt->flags);
00413     else
00414         return rm_write_video(s, pkt->data, pkt->size, pkt->flags);
00415 }
00416 
00417 static int rm_write_trailer(AVFormatContext *s)
00418 {
00419     RMMuxContext *rm = s->priv_data;
00420     int data_size, index_pos, i;
00421     ByteIOContext *pb = s->pb;
00422 
00423     if (!url_is_streamed(s->pb)) {
00424         /* end of file: finish to write header */
00425         index_pos = url_fseek(pb, 0, SEEK_CUR);
00426         data_size = index_pos - rm->data_pos;
00427 
00428         /* index */
00429         put_tag(pb, "INDX");
00430         put_be32(pb, 10 + 10 * s->nb_streams);
00431         put_be16(pb, 0);
00432 
00433         for(i=0;i<s->nb_streams;i++) {
00434             put_be32(pb, 0); /* zero indexes */
00435             put_be16(pb, i); /* stream number */
00436             put_be32(pb, 0); /* next index */
00437         }
00438         /* undocumented end header */
00439         put_be32(pb, 0);
00440         put_be32(pb, 0);
00441 
00442         url_fseek(pb, 0, SEEK_SET);
00443         for(i=0;i<s->nb_streams;i++)
00444             rm->streams[i].total_frames = rm->streams[i].nb_frames;
00445         rv10_write_header(s, data_size, index_pos);
00446     } else {
00447         /* undocumented end header */
00448         put_be32(pb, 0);
00449         put_be32(pb, 0);
00450     }
00451     put_flush_packet(pb);
00452     return 0;
00453 }
00454 
00455 
00456 AVOutputFormat rm_muxer = {
00457     "rm",
00458     NULL_IF_CONFIG_SMALL("RealMedia format"),
00459     "application/vnd.rn-realmedia",
00460     "rm,ra",
00461     sizeof(RMMuxContext),
00462     CODEC_ID_AC3,
00463     CODEC_ID_RV10,
00464     rm_write_header,
00465     rm_write_packet,
00466     rm_write_trailer,
00467 };