Sorry, I didn't make myself clear.

============================
while(has_data){

    get_960_samples();

    av_audio_fifo_write(960);

    while(avio_audio_fifo_size() >= 1024){

      avio_fifo_read(1024);

      encode(1024); //encode in hear or other

    }
}

if(avio_audio_fifo_size() % 1024){

   fill_mute(1024 - avio_audio_fifo_size());  //Only fill the mute at the last!

   avio_fifo_read(1024);

   encode(); //encode in hear or other
}
============================


LeiHe
[email protected]<mailto:[email protected]>



在 2018年10月29日,下午12:52,강구철 <[email protected]<mailto:[email protected]>> 写道:

I already tried using av_fifo for make it size to 1024. but still encoding api 
output 960 plus blank 68 samples(total 1024).

//AVCodecContext initialization
AVCodecContext  *opus_context;
AVCodecContext  *aac_context;

        opus_context->channels = 1;
        opus_context->channel_layout = AV_CH_LAYOUT_MONO;
        opus_context->sample_rate    = 48000;//SRATE;
        opus_context->sample_fmt     = AV_SAMPLE_FMT_FLTP;//4BYTE
        opus_context->bit_rate       = 48000;//BITRATE;

        aac_context->channels = 1;
        aac_context->channel_layout = AV_CH_LAYOUT_MONO;
        aac_context->sample_rate    = 48000;//SRATE;
        aac_context->sample_fmt     = AV_SAMPLE_FMT_FLTP;
        //aac_context->sample_fmt   = AV_SAMPLE_FMT_S32;
        aac_context->bit_rate       = 48000;//BITRATE;//64000
                     //aac_context->strict_std_compliance = 
FF_COMPLIANCE_EXPERIMENTAL;

        fifo = av_audio_fifo_alloc(aac_context->sample_fmt, 
aac_context->channels, 1);

        opus_codec = avcodec_find_decoder( AV_CODEC_ID_OPUS );
        aac_codec  = avcodec_find_encoder( AV_CODEC_ID_AAC  );

//transcoder function for trans OPUS => ACC

tcode(AVPacket* src, AVPacket* dst){
           opus_context->frame_size=960;//TESTTEST
           decoded_frame->data[0] = (uint8_t*)av_malloc(4*1024);//TESTTEST
           ret = avcodec_decode_audio4(opus_context, decoded_frame,  
&data_present, src); // decoded_frame 960 sucess
}



    //WAVE Generate for 68 
samples------------------------------------------------------------
    unsigned char _cbuf[100];
    unsigned char *cbuf=NULL;
    cbuf = _cbuf;
    cbuf[0]=0xf2;
    cbuf[1]=0xdb;
    cbuf[2]=0x0;
    cbuf[3]=0x3f;
    decoded_frame->nb_samples = 1024;//change to 1024 for AAC encoding block

    for(int i=960; i<1024 ; i++){
               decoded_frame->data[0][i*4+0] = (uint8_t)cbuf[0]; //float
               decoded_frame->data[0][i*4+1] = (uint8_t)cbuf[1]; //float
               decoded_frame->data[0][i*4+2] = (uint8_t)cbuf[2]; //float
               decoded_frame->data[0][i*4+3] = (uint8_t)cbuf[3]; //float
               cbuf[2] += 2;        //make saw type wav using range is 0~96
               if(cbuf[2] >= 96)cbuf[2]=0;//mod reset to 0
    }//END OF Generate 68 sample of last WAVE(PCM Float 4byte LittleEndian 
type)---------------------------------------------------------

ret = avcodec_encode_audio2(aac_context, dst, decoded_frame, 
&data_present);//AAC ENCODING Success 1024 sampes compressed but contain last 
68 samples blank.

this mail attached sample result aac before muxing and next figure show you 
every 20msec data has 64 nomalized blank samples(marked red pen). thanks for 
any recommand or inform.

<abc1.m4a AAC recorded webrtc sample open with audacity>
<image001.png>

From: Libav-user [mailto:[email protected]] On Behalf Of He Lei
Sent: Friday, October 26, 2018 6:16 PM
To: This list is about using libavcodec, libavformat, libavutil, libavdevice 
and libavfilter.
Subject: Re: [Libav-user] OPUS transcoding to AAC but 960 sample increase to 
1024 with a nomalized blank data.

Using “audio_fifo” to cache samples, When the samples number in fifo is enough 
to 1024, and then encode it.
The last, If the number of samples  is less than 1024 in fifo, fill with mute


look at “doc/examples/transcode_aac.c”

LeiHec
[email protected]<mailto:[email protected]>




在 2018年10月26日,下午4:34,강구철 <[email protected]<mailto:[email protected]>> 写道:

Im transcode voice comming from WebRTC through by RTP with h264 video.
received sound unit is 20msec OPUS stereo 48000 2channel sample per second
its good decoded to PCM32 FLTP type and good play.
after decode I encode to AAC 48000 stereo frame nb_samples is 960. but encoding 
ffmpeg aac function
attach 64 samples every each decoded raw PCM samples. what should I do for it 
to improving final aac product quality ?

now I found AAC Context be able to control cypher block size 1024 to 960. some 
documents say aac encoder default block is 1024.

AACContext *ac = (AACContext*)aac_context->priv_data;
MPEG4AudioConfig *m4ac = &(ac->oc[0].m4ac);
m4ac->frame_length_short = 1;//1:960, 0:1024

is this right approching ? appriciate any kinds of oppinion of you guys!!


_______________________________________________
Libav-user mailing list
[email protected]<mailto:[email protected]>
http://ffmpeg.org/mailman/listinfo/libav-user

<abc1.m4a>_______________________________________________
Libav-user mailing list
[email protected]<mailto:[email protected]>
http://ffmpeg.org/mailman/listinfo/libav-user

_______________________________________________
Libav-user mailing list
[email protected]
http://ffmpeg.org/mailman/listinfo/libav-user

Reply via email to