opus - Oct 2012 - WAVE PCM to OPUS and back

Hello,

I have an p2p voice chat application using WAVE PCM (winmm). Now i am 
trying to add opus encoding to it to send it over the TCP/IP and then 
decoding it back to play - but without success (without opus it works ok)

Here is an example of my code. I get message from input device then 
encode with opus then decode it back to output wave header and play. 
Doing it i hear only noise in my headphones. So how can i use Opus in 
this way? Thanks in advance

WAVEFORMATEX waveform;
waveform.cbSize = 0;
waveform.wFormatTag = WAVE_FORMAT_PCM;
waveform.nChannels = 2;
waveform.wBitsPerSample = 8;
waveform.nSamplesPerSec = 48000;
waveform.nBlockAlign = waveform.nChannels * (waveform.wBitsPerSample >>
3);
waveform.nAvgBytesPerSec = waveform.nSamplesPerSec * waveform.nBlockAlign;

#define BUFSIZE 512

...

case  MM_WIM_DATA:
             {
                 WAVEHDR *Hdr = (WAVEHDR *) msg.lParam;
                  int  opusRes;
                 m_opusEncoder = opus_encoder_create(48000, 2,
OPUS_APPLICATION_AUDIO, &opusRes);
                 m_opusDecoder = opus_decoder_create(48000, 2, &opusRes);
  
                 unsigned  char  data[BUFSIZE];
                 opus_int32 dataLength = opus_encode(m_opusEncoder,
(short*)Hdr->lpData, 80, data, BUFSIZE);

                 dataLength = opus_decode(m_opusDecoder, (unsigned  char*)data,
dataLength, (short*)WaveHdrOut.lpData, BUFSIZE, 0);

                 waveOutWrite(hWaveOut, &WaveHdrOut,sizeof(WAVEHDR));

-- 
Best regards,

  Stepan Melnichuk

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.xiph.org/pipermail/opus/attachments/20121025/e65d2547/attachment.htm

On 12-10-25 6:41 AM, Stepan Melnichuk wrote:
>                 opus_int32 dataLength = opus_encode(m_opusEncoder,
(short*)Hdr->lpData, 80, data, BUFSIZE);
>                 dataLength = opus_decode(m_opusDecoder, (unsigned
char*)data, dataLength, (short*)WaveHdrOut.lpData, BUFSIZE, 0);
What does dataLength end up being here? It looks like you're only
encoding 80 samples, which should be an error. You need to pass in one
of the allowed packet durations' worth of data: e.g. 960 (20ms) would be
a better value.

Also, it looks like you're only encoding those 80 samples and then
leaking the encoder and decoder contexts. You need to loop over all
available audio data and code it, one packet duration at a time, using
the same encoder and decoder for each call.

Hope that helps,
 -r

On 12-10-25 8:56 PM, Stepan Melnichuk wrote:
> Thanks a lot! It was my mistake re sample. Now it works perfect.
Glad you got it working,
 -r