Record rtsp stream with ffmpeg in iOS - objective-c

Question

I've followed iFrameExtractor to successfully stream rtsp in my swift project. In this project, it also has recording function. It basically use avformat_write_header
, av_interleaved_write_frame and av_write_trailer to save the rtsp source into mp4 file.
When I used this project in my device, the rtsp streaming works fine, but recording function will always generate a blank mp4 file with no image and sound.
Could anyone tell me what step that I miss?
I'm using iPhone5 with iOS 9.1 and XCode 7.1.1.
The ffmpeg is 2.8.3 version and followed the compile instruction by CompilationGuide – FFmpeg
Following is the sample code in this project
The function that generate every frame:
-(BOOL)stepFrame {
// AVPacket packet;
int frameFinished=0;
static bool bFirstIFrame=false;
static int64_t vPTS=0, vDTS=0, vAudioPTS=0, vAudioDTS=0;
while(!frameFinished && av_read_frame(pFormatCtx, &packet)>=0) {
// Is this a packet from the video stream?
if(packet.stream_index==videoStream) {
// 20130525 albert.liao modified start
// Initialize a new format context for writing file
if(veVideoRecordState!=eH264RecIdle)
{
switch(veVideoRecordState)
{
case eH264RecInit:
{
if ( !pFormatCtx_Record )
{
int bFlag = 0;
//NSString *videoPath = [NSHomeDirectory() stringByAppendingPathComponent:#"Documents/test.mp4"];
NSString *videoPath = #"/Users/liaokuohsun/iFrameTest.mp4";
const char *file = [videoPath UTF8String];
pFormatCtx_Record = avformat_alloc_context();
bFlag = h264_file_create(file, pFormatCtx_Record, pCodecCtx, pAudioCodecCtx,/*fps*/0.0, packet.data, packet.size );
if(bFlag==true)
{
veVideoRecordState = eH264RecActive;
fprintf(stderr, "h264_file_create success\n");
}
else
{
veVideoRecordState = eH264RecIdle;
fprintf(stderr, "h264_file_create error\n");
}
}
}
//break;
case eH264RecActive:
{
if((bFirstIFrame==false) &&(packet.flags&AV_PKT_FLAG_KEY)==AV_PKT_FLAG_KEY)
{
bFirstIFrame=true;
vPTS = packet.pts ;
vDTS = packet.dts ;
#if 0
NSRunLoop *pRunLoop = [NSRunLoop currentRunLoop];
[pRunLoop addTimer:RecordingTimer forMode:NSDefaultRunLoopMode];
#else
[NSTimer scheduledTimerWithTimeInterval:5.0//2.0
target:self
selector:#selector(StopRecording:)
userInfo:nil
repeats:NO];
#endif
}
// Record audio when 1st i-Frame is obtained
if(bFirstIFrame==true)
{
if ( pFormatCtx_Record )
{
#if PTS_DTS_IS_CORRECT==1
packet.pts = packet.pts - vPTS;
packet.dts = packet.dts - vDTS;
#endif
h264_file_write_frame( pFormatCtx_Record, packet.stream_index, packet.data, packet.size, packet.dts, packet.pts);
}
else
{
NSLog(#"pFormatCtx_Record no exist");
}
}
}
break;
case eH264RecClose:
{
if ( pFormatCtx_Record )
{
h264_file_close(pFormatCtx_Record);
#if 0
// 20130607 Test
dispatch_async(dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0), ^(void)
{
ALAssetsLibrary *library = [[ALAssetsLibrary alloc]init];
NSString *filePathString = [NSHomeDirectory() stringByAppendingPathComponent:#"Documents/test.mp4"];
NSURL *filePathURL = [NSURL fileURLWithPath:filePathString isDirectory:NO];
if(1)// ([library videoAtPathIsCompatibleWithSavedPhotosAlbum:filePathURL])
{
[library writeVideoAtPathToSavedPhotosAlbum:filePathURL completionBlock:^(NSURL *assetURL, NSError *error){
if (error) {
// TODO: error handling
NSLog(#"writeVideoAtPathToSavedPhotosAlbum error");
} else {
// TODO: success handling
NSLog(#"writeVideoAtPathToSavedPhotosAlbum success");
}
}];
}
[library release];
});
#endif
vPTS = 0;
vDTS = 0;
vAudioPTS = 0;
vAudioDTS = 0;
pFormatCtx_Record = NULL;
NSLog(#"h264_file_close() is finished");
}
else
{
NSLog(#"fc no exist");
}
bFirstIFrame = false;
veVideoRecordState = eH264RecIdle;
}
break;
default:
if ( pFormatCtx_Record )
{
h264_file_close(pFormatCtx_Record);
pFormatCtx_Record = NULL;
}
NSLog(#"[ERROR] unexpected veVideoRecordState!!");
veVideoRecordState = eH264RecIdle;
break;
}
}
// Decode video frame
avcodec_decode_video2(pCodecCtx, pFrame, &frameFinished, &packet);
}
else if(packet.stream_index==audioStream)
{
// 20131024 albert.liao modfied start
static int vPktCount=0;
BOOL bIsAACADTS = FALSE;
int ret = 0;
if(aPlayer.vAACType == eAAC_UNDEFINED)
{
tAACADTSHeaderInfo vxAACADTSHeaderInfo = {0};
bIsAACADTS = [AudioUtilities parseAACADTSHeader:(uint8_t *)packet.data ToHeader:&vxAACADTSHeaderInfo];
}
#synchronized(aPlayer)
{
if(aPlayer==nil)
{
aPlayer = [[AudioPlayer alloc]initAudio:nil withCodecCtx:(AVCodecContext *) pAudioCodecCtx];
NSLog(#"aPlayer initAudio");
if(bIsAACADTS)
{
aPlayer.vAACType = eAAC_ADTS;
//NSLog(#"is ADTS AAC");
}
}
else
{
if(vPktCount<5) // The voice is listened once image is rendered
{
vPktCount++;
}
else
{
if([aPlayer getStatus]!=eAudioRunning)
{
dispatch_async(dispatch_get_main_queue(), ^(void) {
#synchronized(aPlayer)
{
NSLog(#"aPlayer start play");
[aPlayer Play];
}
});
}
}
}
};
#synchronized(aPlayer)
{
int ret = 0;
ret = [aPlayer putAVPacket:&packet];
if(ret <= 0)
NSLog(#"Put Audio Packet Error!!");
}
// 20131024 albert.liao modfied end
if(bFirstIFrame==true)
{
switch(veVideoRecordState)
{
case eH264RecActive:
{
if ( pFormatCtx_Record )
{
h264_file_write_audio_frame(pFormatCtx_Record, pAudioCodecCtx, packet.stream_index, packet.data, packet.size, packet.dts, packet.pts);
}
else
{
NSLog(#"pFormatCtx_Record no exist");
}
}
}
}
}
else
{
//fprintf(stderr, "packet len=%d, Byte=%02X%02X%02X%02X%02X\n",\
packet.size, packet.data[0],packet.data[1],packet.data[2],packet.data[3], packet.data[4]);
}
// 20130525 albert.liao modified end
}
return frameFinished!=0;
}
avformat_write_header:
int h264_file_create(const char *pFilePath, AVFormatContext *fc, AVCodecContext *pCodecCtx,AVCodecContext *pAudioCodecCtx, double fps, void *p, int len )
{
int vRet=0;
AVOutputFormat *of=NULL;
AVStream *pst=NULL;
AVCodecContext *pcc=NULL, *pAudioOutputCodecContext=NULL;
avcodec_register_all();
av_register_all();
av_log_set_level(AV_LOG_VERBOSE);
if(!pFilePath)
{
fprintf(stderr, "FilePath no exist");
return -1;
}
if(!fc)
{
fprintf(stderr, "AVFormatContext no exist");
return -1;
}
fprintf(stderr, "file=%s\n",pFilePath);
// Create container
of = av_guess_format( 0, pFilePath, 0 );
fc->oformat = of;
strcpy( fc->filename, pFilePath );
// Add video stream
pst = avformat_new_stream( fc, 0 );
vVideoStreamIdx = pst->index;
fprintf(stderr,"Video Stream:%d",vVideoStreamIdx);
pcc = pst->codec;
avcodec_get_context_defaults3( pcc, AVMEDIA_TYPE_VIDEO );
// Save the stream as origin setting without convert
pcc->codec_type = pCodecCtx->codec_type;
pcc->codec_id = pCodecCtx->codec_id;
pcc->bit_rate = pCodecCtx->bit_rate;
pcc->width = pCodecCtx->width;
pcc->height = pCodecCtx->height;
if(fps==0)
{
double fps=0.0;
AVRational pTimeBase;
pTimeBase.num = pCodecCtx->time_base.num;
pTimeBase.den = pCodecCtx->time_base.den;
fps = 1.0/ av_q2d(pCodecCtx->time_base)/ FFMAX(pCodecCtx->ticks_per_frame, 1);
fprintf(stderr,"fps_method(tbc): 1/av_q2d()=%g",fps);
pcc->time_base.num = 1;
pcc->time_base.den = fps;
}
else
{
pcc->time_base.num = 1;
pcc->time_base.den = fps;
}
// reference ffmpeg\libavformat\utils.c
// For SPS and PPS in avcC container
pcc->extradata = malloc(sizeof(uint8_t)*pCodecCtx->extradata_size);
memcpy(pcc->extradata, pCodecCtx->extradata, pCodecCtx->extradata_size);
pcc->extradata_size = pCodecCtx->extradata_size;
// For Audio stream
if(pAudioCodecCtx)
{
AVCodec *pAudioCodec=NULL;
AVStream *pst2=NULL;
pAudioCodec = avcodec_find_encoder(AV_CODEC_ID_AAC);
// Add audio stream
pst2 = avformat_new_stream( fc, pAudioCodec );
vAudioStreamIdx = pst2->index;
pAudioOutputCodecContext = pst2->codec;
avcodec_get_context_defaults3( pAudioOutputCodecContext, pAudioCodec );
fprintf(stderr,"Audio Stream:%d",vAudioStreamIdx);
fprintf(stderr,"pAudioCodecCtx->bits_per_coded_sample=%d",pAudioCodecCtx->bits_per_coded_sample);
pAudioOutputCodecContext->codec_type = AVMEDIA_TYPE_AUDIO;
pAudioOutputCodecContext->codec_id = AV_CODEC_ID_AAC;
// Copy the codec attributes
pAudioOutputCodecContext->channels = pAudioCodecCtx->channels;
pAudioOutputCodecContext->channel_layout = pAudioCodecCtx->channel_layout;
pAudioOutputCodecContext->sample_rate = pAudioCodecCtx->sample_rate;
pAudioOutputCodecContext->bit_rate = 12000;//pAudioCodecCtx->sample_rate * pAudioCodecCtx->bits_per_coded_sample;
pAudioOutputCodecContext->bits_per_coded_sample = pAudioCodecCtx->bits_per_coded_sample;
pAudioOutputCodecContext->profile = pAudioCodecCtx->profile;
//FF_PROFILE_AAC_LOW;
// pAudioCodecCtx->bit_rate;
// AV_SAMPLE_FMT_U8P, AV_SAMPLE_FMT_S16P
//pAudioOutputCodecContext->sample_fmt = AV_SAMPLE_FMT_FLTP;//pAudioCodecCtx->sample_fmt;
pAudioOutputCodecContext->sample_fmt = pAudioCodecCtx->sample_fmt;
//pAudioOutputCodecContext->sample_fmt = AV_SAMPLE_FMT_U8;
pAudioOutputCodecContext->sample_aspect_ratio = pAudioCodecCtx->sample_aspect_ratio;
pAudioOutputCodecContext->time_base.num = pAudioCodecCtx->time_base.num;
pAudioOutputCodecContext->time_base.den = pAudioCodecCtx->time_base.den;
pAudioOutputCodecContext->ticks_per_frame = pAudioCodecCtx->ticks_per_frame;
pAudioOutputCodecContext->frame_size = 1024;
fprintf(stderr,"profile:%d, sample_rate:%d, channles:%d", pAudioOutputCodecContext->profile, pAudioOutputCodecContext->sample_rate, pAudioOutputCodecContext->channels);
AVDictionary *opts = NULL;
av_dict_set(&opts, "strict", "experimental", 0);
if (avcodec_open2(pAudioOutputCodecContext, pAudioCodec, &opts) < 0) {
fprintf(stderr, "\ncould not open codec\n");
}
av_dict_free(&opts);
#if 0
// For Audio, this part is no need
if(pAudioCodecCtx->extradata_size!=0)
{
NSLog(#"extradata_size !=0");
pAudioOutputCodecContext->extradata = malloc(sizeof(uint8_t)*pAudioCodecCtx->extradata_size);
memcpy(pAudioOutputCodecContext->extradata, pAudioCodecCtx->extradata, pAudioCodecCtx->extradata_size);
pAudioOutputCodecContext->extradata_size = pAudioCodecCtx->extradata_size;
}
else
{
// For WMA test only
pAudioOutputCodecContext->extradata_size = 0;
NSLog(#"extradata_size ==0");
}
#endif
}
if(fc->oformat->flags & AVFMT_GLOBALHEADER)
{
pcc->flags |= CODEC_FLAG_GLOBAL_HEADER;
pAudioOutputCodecContext->flags |= CODEC_FLAG_GLOBAL_HEADER;
}
if ( !( fc->oformat->flags & AVFMT_NOFILE ) )
{
vRet = avio_open( &fc->pb, fc->filename, AVIO_FLAG_WRITE );
if(vRet!=0)
{
fprintf(stderr,"avio_open(%s) error", fc->filename);
}
}
// dump format in console
av_dump_format(fc, 0, pFilePath, 1);
vRet = avformat_write_header( fc, NULL );
if(vRet==0)
return 1;
else
return 0;
}
av_interleaved_write_frame:
void h264_file_write_frame(AVFormatContext *fc, int vStreamIdx, const void* p, int len, int64_t dts, int64_t pts )
{
AVStream *pst = NULL;
AVPacket pkt;
if ( 0 > vVideoStreamIdx )
return;
// may be audio or video
pst = fc->streams[ vStreamIdx ];
// Init packet
av_init_packet( &pkt );
if(vStreamIdx ==vVideoStreamIdx)
{
pkt.flags |= ( 0 >= getVopType( p, len ) ) ? AV_PKT_FLAG_KEY : 0;
//pkt.flags |= AV_PKT_FLAG_KEY;
pkt.stream_index = pst->index;
pkt.data = (uint8_t*)p;
pkt.size = len;
pkt.dts = AV_NOPTS_VALUE;
pkt.pts = AV_NOPTS_VALUE;
// TODO: mark or unmark the log
//fprintf(stderr, "dts=%lld, pts=%lld\n",dts,pts);
// av_write_frame( fc, &pkt );
}
av_interleaved_write_frame( fc, &pkt );
}
av_write_trailer:
void h264_file_close(AVFormatContext *fc)
{
if ( !fc )
return;
av_write_trailer( fc );
if ( fc->oformat && !( fc->oformat->flags & AVFMT_NOFILE ) && fc->pb )
avio_close( fc->pb );
av_free( fc );
}
Thanks.

Answer 1

It looks like you're using the same AVFormatContext for both the input and output?
In the line
pst = fc->streams[ vStreamIdx ];
You're assigning the AVStream* from your AVFormatContext connected with your input (RTSP stream). But then later on you're trying to write the packet back to the same context av_interleaved_write_frame( fc, &pkt );. I kind of think of a context as a file which has helped me navagate this type of thing better. I do something identicial to what you're doing (not iOS though) where I use a separate AVFormatContext for each of the input (RTSP stream) and output (mp4 file). If I'm correct, I think what you just need to do is initialize an AVFormatContext and properly.
The following code (without error checking everything) is what I do to take an AVFormatContext * output_format_context = NULL and the AVFormatContext * input_format_context that I had associated with the RTSP stream and write from one to the other. This is after I have fetched a packet, etc., which in your case it looks like you're populating (I just take the packet from av_read_frame and re-package it.
This is code that could be in your write frame function (but it also does include the writing of the header).
AVFormatContext * output_format_context;
AVStream * in_stream_2;
AVStream * out_stream_2;
// Allocate the context with the output file
avformat_alloc_output_context2(&output_format_context, NULL, NULL, out_filename.c_str());
// Point to AVOutputFormat * output_format for manipulation
output_format = output_format_context->oformat;
// Loop through all streams
for (i = 0; i < input_format_context->nb_streams; i++) {
// Create a pointer to the input stream that was allocated earlier in the code
AVStream *in_stream = input_format_context->streams[i];
// Create a pointer to a new stream that will be part of the output
AVStream *out_stream = avformat_new_stream(output_format_context, in_stream->codec->codec);
// Set time_base of the new output stream to equal the input stream one since I'm not changing anything (can avoid but get a deprecation warning)
out_stream->time_base = in_stream->time_base;
// This is the non-deprecated way of copying all the parameters from the input stream into the output stream since everything stays the same
avcodec_parameters_from_context(out_stream->codecpar, in_stream->codec);
// I don't remember what this is for :)
out_stream->codec->codec_tag = 0;
// This just sets a flag from the format context to the stream relating to the header
if (output_format_context->oformat->flags & AVFMT_GLOBALHEADER)
out_stream->codec->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
}
// Check NOFILE flag and open the output file context (previously the output file was associated with the format context, now it is actually opened.
if (!(output_format->flags & AVFMT_NOFILE))
avio_open(&output_format_context->pb, out_filename.c_str(), AVIO_FLAG_WRITE);
// Write the header (not sure if this is always needed but h264 I believe it is.
avformat_write_header(output_format_context,NULL);
// Re-getting the appropriate stream that was populated above (this should allow for both audio/video)
in_stream_2 = input_format_context->streams[packet.stream_index];
out_stream_2 = output_format_context->streams[packet.stream_index];
// Rescaling pts and dts, duration and pos - you would do as you need in your code.
packet.pts = av_rescale_q_rnd(packet.pts, in_stream_2->time_base, out_stream_2->time_base, (AVRounding) (AV_ROUND_NEAR_INF | AV_ROUND_PASS_MINMAX));
packet.dts = av_rescale_q_rnd(packet.dts, in_stream_2->time_base, out_stream_2->time_base, (AVRounding) (AV_ROUND_NEAR_INF | AV_ROUND_PASS_MINMAX));
packet.duration = av_rescale_q(packet.duration, in_stream_2->time_base, out_stream_2->time_base);
packet.pos = -1;
// The first packet of my stream always gives me negative dts/pts so this just protects that first one for my purposes. You probably don't need.
if (packet.dts < 0) packet.dts = 0;
if (packet.pts < 0) packet.pts = 0;
// Finally write the frame
av_interleaved_write_frame(output_format_context, &packet);
// ....
// Write header, close/cleanup... etc
// ....
This code is fairly bare bones and doesn't include the setup (which it sounds like you're doing correctly anyway). I would also imagine this code could be cleaned up and tweaked for your purposes, but this works for me to re-write the RTSP stream into a file (in my case many files but code not shown).
The code is C code, so you might need to do minor tweaks for making it Swift compatible (for some of the library function calls maybe). I think overall it should be compatible though.
Hopefully this helps point you to the right direction. This was cobbled together thanks to several sample code sources (I don't remember where), along with warning prompts from the libraries themselves.