]> rtime.felk.cvut.cz Git - frescor/ffmpeg.git/blobdiff - libavformat/flvenc.c
Put FRSH support as a separate protocol
[frescor/ffmpeg.git] / libavformat / flvenc.c
index 1bc0f9f63be6bd4826db3d01069ca7f89a482e31..63accbb0cb00adad29cfa99aec998252b5639599 100644 (file)
@@ -1,6 +1,6 @@
 /*
  * FLV muxer
- * Copyright (c) 2003 The FFmpeg Project.
+ * Copyright (c) 2003 The FFmpeg Project
  *
  * This file is part of FFmpeg.
  *
@@ -21,6 +21,7 @@
 #include "avformat.h"
 #include "flv.h"
 #include "riff.h"
+#include "avc.h"
 
 #undef NDEBUG
 #include <assert.h>
@@ -30,30 +31,35 @@ static const AVCodecTag flv_video_codec_ids[] = {
     {CODEC_ID_FLASHSV, FLV_CODECID_SCREEN},
     {CODEC_ID_VP6F,    FLV_CODECID_VP6   },
     {CODEC_ID_VP6,     FLV_CODECID_VP6   },
+    {CODEC_ID_H264,    FLV_CODECID_H264  },
     {CODEC_ID_NONE,    0}
 };
 
 static const AVCodecTag flv_audio_codec_ids[] = {
     {CODEC_ID_MP3,       FLV_CODECID_MP3    >> FLV_AUDIO_CODECID_OFFSET},
-    {CODEC_ID_PCM_S8,    FLV_CODECID_PCM_BE >> FLV_AUDIO_CODECID_OFFSET},
-    {CODEC_ID_PCM_S16BE, FLV_CODECID_PCM_BE >> FLV_AUDIO_CODECID_OFFSET},
+    {CODEC_ID_PCM_S8,    FLV_CODECID_PCM    >> FLV_AUDIO_CODECID_OFFSET},
+    {CODEC_ID_PCM_S16BE, FLV_CODECID_PCM    >> FLV_AUDIO_CODECID_OFFSET},
     {CODEC_ID_PCM_S16LE, FLV_CODECID_PCM_LE >> FLV_AUDIO_CODECID_OFFSET},
     {CODEC_ID_ADPCM_SWF, FLV_CODECID_ADPCM  >> FLV_AUDIO_CODECID_OFFSET},
+    {CODEC_ID_AAC,       FLV_CODECID_AAC    >> FLV_AUDIO_CODECID_OFFSET},
+    {CODEC_ID_NELLYMOSER, FLV_CODECID_NELLYMOSER >> FLV_AUDIO_CODECID_OFFSET},
     {CODEC_ID_NONE,      0}
 };
 
 typedef struct FLVContext {
-    int hasAudio;
-    int hasVideo;
     int reserved;
-    offset_t duration_offset;
-    offset_t filesize_offset;
+    int64_t duration_offset;
+    int64_t filesize_offset;
     int64_t duration;
+    int delay; ///< first dts delay for AVC
 } FLVContext;
 
 static int get_audio_flags(AVCodecContext *enc){
-    int flags = (enc->bits_per_sample == 16) ? FLV_SAMPLESSIZE_16BIT : FLV_SAMPLESSIZE_8BIT;
+    int flags = (enc->bits_per_coded_sample == 16) ? FLV_SAMPLESSIZE_16BIT : FLV_SAMPLESSIZE_8BIT;
 
+    if (enc->codec_id == CODEC_ID_AAC) // specs force these parameters
+        return FLV_CODECID_AAC | FLV_SAMPLERATE_44100HZ | FLV_SAMPLESSIZE_16BIT | FLV_STEREO;
+    else {
     switch (enc->sample_rate) {
         case    44100:
             flags |= FLV_SAMPLERATE_44100HZ;
@@ -66,12 +72,15 @@ static int get_audio_flags(AVCodecContext *enc){
             break;
         case     8000: //nellymoser only
         case     5512: //not mp3
-            flags |= FLV_SAMPLERATE_SPECIAL;
-            break;
+            if(enc->codec_id != CODEC_ID_MP3){
+                flags |= FLV_SAMPLERATE_SPECIAL;
+                break;
+            }
         default:
-            av_log(enc, AV_LOG_ERROR, "flv doesnt support that sample rate, choose from (44100, 22050, 11025)\n");
+            av_log(enc, AV_LOG_ERROR, "flv does not support that sample rate, choose from (44100, 22050, 11025).\n");
             return -1;
     }
+    }
 
     if (enc->channels > 1) {
         flags |= FLV_STEREO;
@@ -82,16 +91,23 @@ static int get_audio_flags(AVCodecContext *enc){
         flags |= FLV_CODECID_MP3    | FLV_SAMPLESSIZE_16BIT;
         break;
     case CODEC_ID_PCM_S8:
-        flags |= FLV_CODECID_PCM_BE | FLV_SAMPLESSIZE_8BIT;
+        flags |= FLV_CODECID_PCM    | FLV_SAMPLESSIZE_8BIT;
         break;
     case CODEC_ID_PCM_S16BE:
-        flags |= FLV_CODECID_PCM_BE | FLV_SAMPLESSIZE_16BIT;
+        flags |= FLV_CODECID_PCM    | FLV_SAMPLESSIZE_16BIT;
         break;
     case CODEC_ID_PCM_S16LE:
         flags |= FLV_CODECID_PCM_LE | FLV_SAMPLESSIZE_16BIT;
         break;
     case CODEC_ID_ADPCM_SWF:
-        flags |= FLV_CODECID_ADPCM;
+        flags |= FLV_CODECID_ADPCM | FLV_SAMPLESSIZE_16BIT;
+        break;
+    case CODEC_ID_NELLYMOSER:
+        if (enc->sample_rate == 8000) {
+            flags |= FLV_CODECID_NELLYMOSER_8KHZ_MONO | FLV_SAMPLESSIZE_16BIT;
+        } else {
+            flags |= FLV_CODECID_NELLYMOSER | FLV_SAMPLESSIZE_16BIT;
+        }
         break;
     case 0:
         flags |= enc->codec_tag<<4;
@@ -124,49 +140,37 @@ static void put_amf_bool(ByteIOContext *pb, int b) {
 
 static int flv_write_header(AVFormatContext *s)
 {
-    ByteIOContext *pb = &s->pb;
+    ByteIOContext *pb = s->pb;
     FLVContext *flv = s->priv_data;
-    int i, width, height, samplerate, samplesize, channels, audiocodecid, videocodecid;
+    AVCodecContext *audio_enc = NULL, *video_enc = NULL;
+    int i;
     double framerate = 0.0;
     int metadata_size_pos, data_size;
 
-    flv->hasAudio = 0;
-    flv->hasVideo = 0;
-
     for(i=0; i<s->nb_streams; i++){
         AVCodecContext *enc = s->streams[i]->codec;
         if (enc->codec_type == CODEC_TYPE_VIDEO) {
-            width = enc->width;
-            height = enc->height;
             if (s->streams[i]->r_frame_rate.den && s->streams[i]->r_frame_rate.num) {
                 framerate = av_q2d(s->streams[i]->r_frame_rate);
             } else {
                 framerate = 1/av_q2d(s->streams[i]->codec->time_base);
             }
-            flv->hasVideo=1;
-
-            videocodecid = codec_get_tag(flv_video_codec_ids, enc->codec_id);
-            if(videocodecid == 0) {
+            video_enc = enc;
+            if(enc->codec_tag == 0) {
                 av_log(enc, AV_LOG_ERROR, "video codec not compatible with flv\n");
                 return -1;
             }
         } else {
-            flv->hasAudio=1;
-            samplerate = enc->sample_rate;
-            channels = enc->channels;
-
-            audiocodecid = codec_get_tag(flv_audio_codec_ids, enc->codec_id);
-            samplesize = (enc->codec_id == CODEC_ID_PCM_S8) ? 8 : 16;
-
+            audio_enc = enc;
             if(get_audio_flags(enc)<0)
                 return -1;
         }
-        av_set_pts_info(s->streams[i], 24, 1, 1000); /* 24 bit pts in ms */
+        av_set_pts_info(s->streams[i], 32, 1, 1000); /* 32 bit pts in ms */
     }
     put_tag(pb,"FLV");
     put_byte(pb,1);
-    put_byte(pb,   FLV_HEADER_FLAG_HASAUDIO * flv->hasAudio
-                 + FLV_HEADER_FLAG_HASVIDEO * flv->hasVideo);
+    put_byte(pb,   FLV_HEADER_FLAG_HASAUDIO * !!audio_enc
+                 + FLV_HEADER_FLAG_HASVIDEO * !!video_enc);
     put_be32(pb,9);
     put_be32(pb,0);
 
@@ -196,41 +200,44 @@ static int flv_write_header(AVFormatContext *s)
 
     /* mixed array (hash) with size and string/type/data tuples */
     put_byte(pb, AMF_DATA_TYPE_MIXEDARRAY);
-    put_be32(pb, 5*flv->hasVideo + 4*flv->hasAudio + 2); // +2 for duration and file size
+    put_be32(pb, 5*!!video_enc + 4*!!audio_enc + 2); // +2 for duration and file size
 
     put_amf_string(pb, "duration");
     flv->duration_offset= url_ftell(pb);
     put_amf_double(pb, 0); // delayed write
 
-    if(flv->hasVideo){
+    if(video_enc){
         put_amf_string(pb, "width");
-        put_amf_double(pb, width);
+        put_amf_double(pb, video_enc->width);
 
         put_amf_string(pb, "height");
-        put_amf_double(pb, height);
+        put_amf_double(pb, video_enc->height);
 
         put_amf_string(pb, "videodatarate");
-        put_amf_double(pb, s->bit_rate / 1024.0);
+        put_amf_double(pb, video_enc->bit_rate / 1024.0);
 
         put_amf_string(pb, "framerate");
         put_amf_double(pb, framerate);
 
         put_amf_string(pb, "videocodecid");
-        put_amf_double(pb, videocodecid);
+        put_amf_double(pb, video_enc->codec_tag);
     }
 
-    if(flv->hasAudio){
+    if(audio_enc){
+        put_amf_string(pb, "audiodatarate");
+        put_amf_double(pb, audio_enc->bit_rate / 1024.0);
+
         put_amf_string(pb, "audiosamplerate");
-        put_amf_double(pb, samplerate);
+        put_amf_double(pb, audio_enc->sample_rate);
 
         put_amf_string(pb, "audiosamplesize");
-        put_amf_double(pb, samplesize);
+        put_amf_double(pb, audio_enc->codec_id == CODEC_ID_PCM_S8 ? 8 : 16);
 
         put_amf_string(pb, "stereo");
-        put_amf_bool(pb, (channels == 2));
+        put_amf_bool(pb, audio_enc->channels == 2);
 
         put_amf_string(pb, "audiocodecid");
-        put_amf_double(pb, audiocodecid);
+        put_amf_double(pb, audio_enc->codec_tag);
     }
 
     put_amf_string(pb, "filesize");
@@ -247,6 +254,35 @@ static int flv_write_header(AVFormatContext *s)
     url_fseek(pb, data_size + 10 - 3, SEEK_CUR);
     put_be32(pb, data_size + 11);
 
+    for (i = 0; i < s->nb_streams; i++) {
+        AVCodecContext *enc = s->streams[i]->codec;
+        if (enc->codec_id == CODEC_ID_AAC || enc->codec_id == CODEC_ID_H264) {
+            int64_t pos;
+            put_byte(pb, enc->codec_type == CODEC_TYPE_VIDEO ?
+                     FLV_TAG_TYPE_VIDEO : FLV_TAG_TYPE_AUDIO);
+            put_be24(pb, 0); // size patched later
+            put_be24(pb, 0); // ts
+            put_byte(pb, 0); // ts ext
+            put_be24(pb, 0); // streamid
+            pos = url_ftell(pb);
+            if (enc->codec_id == CODEC_ID_AAC) {
+                put_byte(pb, get_audio_flags(enc));
+                put_byte(pb, 0); // AAC sequence header
+                put_buffer(pb, enc->extradata, enc->extradata_size);
+            } else {
+                put_byte(pb, enc->codec_tag | FLV_FRAME_KEY); // flags
+                put_byte(pb, 0); // AVC sequence header
+                put_be24(pb, 0); // composition time
+                ff_isom_write_avcc(pb, enc->extradata, enc->extradata_size);
+            }
+            data_size = url_ftell(pb) - pos;
+            url_fseek(pb, -data_size - 10, SEEK_CUR);
+            put_be24(pb, data_size);
+            url_fseek(pb, data_size + 10 - 3, SEEK_CUR);
+            put_be32(pb, data_size + 11); // previous tag size
+        }
+    }
+
     return 0;
 }
 
@@ -254,7 +290,7 @@ static int flv_write_trailer(AVFormatContext *s)
 {
     int64_t file_size;
 
-    ByteIOContext *pb = &s->pb;
+    ByteIOContext *pb = s->pb;
     FLVContext *flv = s->priv_data;
 
     file_size = url_ftell(pb);
@@ -271,18 +307,28 @@ static int flv_write_trailer(AVFormatContext *s)
 
 static int flv_write_packet(AVFormatContext *s, AVPacket *pkt)
 {
-    ByteIOContext *pb = &s->pb;
+    ByteIOContext *pb = s->pb;
     AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
     FLVContext *flv = s->priv_data;
+    unsigned ts;
     int size= pkt->size;
-    int flags;
+    uint8_t *data= NULL;
+    int flags, flags_size;
 
 //    av_log(s, AV_LOG_DEBUG, "type:%d pts: %"PRId64" size:%d\n", enc->codec_type, timestamp, size);
 
+    if(enc->codec_id == CODEC_ID_VP6 || enc->codec_id == CODEC_ID_VP6F ||
+       enc->codec_id == CODEC_ID_AAC)
+        flags_size= 2;
+    else if(enc->codec_id == CODEC_ID_H264)
+        flags_size= 5;
+    else
+        flags_size= 1;
+
     if (enc->codec_type == CODEC_TYPE_VIDEO) {
         put_byte(pb, FLV_TAG_TYPE_VIDEO);
 
-        flags = codec_get_tag(flv_video_codec_ids, enc->codec_id);
+        flags = enc->codec_tag;
         if(flags == 0) {
             av_log(enc, AV_LOG_ERROR, "video codec %X not compatible with flv\n",enc->codec_id);
             return -1;
@@ -298,39 +344,60 @@ static int flv_write_packet(AVFormatContext *s, AVPacket *pkt)
         put_byte(pb, FLV_TAG_TYPE_AUDIO);
     }
 
-    if ((enc->codec_id == CODEC_ID_VP6) || (enc->codec_id == CODEC_ID_VP6F))
-        put_be24(pb,size+2); // include the extra byte needed for VP6 in flv and flags
-    else
-        put_be24(pb,size+1); // include flags
-    put_be24(pb,pkt->pts);
-    put_be32(pb,flv->reserved);
+    if (enc->codec_id == CODEC_ID_H264) {
+        /* check if extradata looks like mp4 formated */
+        if (enc->extradata_size > 0 && *(uint8_t*)enc->extradata != 1) {
+            if (ff_avc_parse_nal_units_buf(pkt->data, &data, &size) < 0)
+                return -1;
+        }
+        if (!flv->delay && pkt->dts < 0)
+            flv->delay = -pkt->dts;
+    }
+
+    ts = pkt->dts + flv->delay; // add delay to force positive dts
+    put_be24(pb,size + flags_size);
+    put_be24(pb,ts);
+    put_byte(pb,(ts >> 24) & 0x7F); // timestamps are 32bits _signed_
+    put_be24(pb,flv->reserved);
     put_byte(pb,flags);
     if (enc->codec_id == CODEC_ID_VP6)
         put_byte(pb,0);
     if (enc->codec_id == CODEC_ID_VP6F)
         put_byte(pb, enc->extradata_size ? enc->extradata[0] : 0);
-    put_buffer(pb, pkt->data, size);
-    put_be32(pb,size+1+11); // previous tag size
-    flv->duration = pkt->pts + pkt->duration;
+    else if (enc->codec_id == CODEC_ID_AAC)
+        put_byte(pb,1); // AAC raw
+    else if (enc->codec_id == CODEC_ID_H264) {
+        put_byte(pb,1); // AVC NALU
+        put_be24(pb,pkt->pts - pkt->dts);
+    }
+
+    put_buffer(pb, data ? data : pkt->data, size);
+
+    put_be32(pb,size+flags_size+11); // previous tag size
+    flv->duration = FFMAX(flv->duration, pkt->pts + flv->delay + pkt->duration);
 
     put_flush_packet(pb);
+
+    av_free(data);
+
     return 0;
 }
 
 AVOutputFormat flv_muxer = {
     "flv",
-    "flv format",
+    NULL_IF_CONFIG_SMALL("FLV format"),
     "video/x-flv",
     "flv",
     sizeof(FLVContext),
-#ifdef CONFIG_LIBMP3LAME
+#if CONFIG_LIBMP3LAME
     CODEC_ID_MP3,
 #else // CONFIG_LIBMP3LAME
-    CODEC_ID_NONE,
+    CODEC_ID_ADPCM_SWF,
 #endif // CONFIG_LIBMP3LAME
     CODEC_ID_FLV1,
     flv_write_header,
     flv_write_packet,
     flv_write_trailer,
-    .codec_tag= {flv_video_codec_ids, flv_audio_codec_ids},
+    .codec_tag= (const AVCodecTag* const []){flv_video_codec_ids, flv_audio_codec_ids, 0},
+    .flags= AVFMT_GLOBALHEADER,
 };