I want to record the system's audio output using WASAPI and then saving it to a .wav file.
So far I've followed these guides on WASAPI:
https://msdn.microsoft.com/en-us/library/windows/desktop/dd316551%28v=vs.85%29.aspx https://msdn.microsoft.com/en-us/library/windows/desktop/dd370800%28v=vs.85%29.aspx
I get the buffer data using
audioCaptureClient->GetBuffer(&data, &numFramesAvailable, &flags, NULL, NULL);
then, I'm processing this data, by just writing it at the end of the .wav file:
size_t dataSize = format.nChannels * (format.wBitsPerSample / 8) * numFramesAvailable;
fwrite(data, dataSize, 1, fp);
format
is a WAVEFORMATEX
received from audioClient->GetMixFormat(&format)
:
cbSize: 22
nAvgBytesPerSec: 352800
nBlockAlign: 8
nChannels: 2
nSamplesPerSec: 44100
wBitsPerSample: 32
wFormatTag: 65534 (WAVE_FORMAT_EXTENSIBLE)
Apparently the subtype of WAVE_FORMAT_EXTENSIBLE
is Float:
WAVEFORMATEXTENSIBLE *waveformatextensible = (WAVEFORMATEXTENSIBLE *)format;
if (IsEqualGUID(KSDATAFORMAT_SUBTYPE_IEEE_FLOAT, waveformatextensible->SubFormat)) { // true
Before writing all the captured data to the file I fill in the headers (following http://www.topherlee.com/software/pcm-tut-wavformat.html):
UINT32 sizePlaceholder = 0;
UINT32 fmtLength = 16;
// RIFF Header
fputs("RIFF", fp); // offset 0 (0x00)
fwrite(&sizePlaceholder, 4, 1, fp); // offset 4 (0x04)
fputs("WAVE", fp); // offset 8 (0x08)
// fmt-Section
fputs("fmt ", fp); // offset 12 (0x0C)
fwrite(&fmtLength, 4, 1, fp); // offset 16 (0x10)
fwrite(&format.wFormatTag, 2, 1, fp); // offset 20 (0x14)
fwrite(&format.nChannels, 2, 1, fp); // offset 22 (0x16)
fwrite(&format.nSamplesPerSec, 4, 1, fp); // offset 24 (0x18)
fwrite(&format.nAvgBytesPerSec, 4, 1, fp); // offset 28 (0x1C)
fwrite(&format.nBlockAlign, 2, 1, fp); // offset 32 (0x20)
fwrite(&format.wBitsPerSample, 2, 1, fp); // offset 34 (0x22)
// Data-Section
fputs("data", fp); // offset 36 (0x24)
fwrite(&sizePlaceholder, 4, 1, fp); // offset 40 (0x28)
After finishing writing 3 seconds of data I fill in the placeholders for the file size and the data section size using fwrite
.
The file is not being readable. I suspect it has to do with WAVE_FORMAT_EXTENSIBLE
, but I couldn't figure it out.
I tried overwriting several elements of format
like:
cbSize = 0;
wFormatTag = WAVE_FORMAT_IEEE_FLOAT;
Producing a readable .wav file, but playing as silence with some clicks in it (I tried recording a song).
wFormatTag = WAVE_FORMAT_PCM;
is producing all random noise.
So, finally after long hours of experimenting I found the solution.
There were multiple problems with the code.
WAVE_FORMAT_EXTENSIBLE
uses a file layout, being a bit different. See this great link for more details.fopen
, so the audio data got corrupted, because fwrite
detects newlines (\n
) in the data and adds a carriage return (\r
). I had to use fopen("foo.wav", "wb")
instead of fopen("foo.wav", "w")
.The second problem, was the decisive reason, because I already tried replacing the WAVE_FORMAT_EXTENSIBLE
-tag with the WAVE_FORMAT_IEEE_FLOAT
-tag, which should have worked, because the additional information are not needed for a .wav file to work.