UINT MMDeviceAudioSource::GetNextBuffer(float curVolume)
{
UINT captureSize = 0;
HRESULT err = mmCapture->GetNextPacketSize(&captureSize);
if(FAILED(err))
{
RUNONCE AppWarning(TEXT("MMDeviceAudioSource::GetBuffer: GetNextPacketSize failed"));
return NoAudioAvailable;
}
float *outputBuffer = NULL;
if(captureSize)
{
LPBYTE captureBuffer;
DWORD dwFlags = 0;
UINT numAudioFrames = 0;
UINT64 devPosition;
UINT64 qpcTimestamp;
err = mmCapture->GetBuffer(&captureBuffer, &numAudioFrames, &dwFlags, &devPosition, &qpcTimestamp);
if(FAILED(err))
{
RUNONCE AppWarning(TEXT("MMDeviceAudioSource::GetBuffer: GetBuffer failed"));
return NoAudioAvailable;
}
QWORD newTimestamp;
if(dwFlags & AUDCLNT_BUFFERFLAGS_TIMESTAMP_ERROR)
{
RUNONCE AppWarning(TEXT("MMDeviceAudioSource::GetBuffer: woa woa woa, getting timestamp errors from the audio subsystem. device = %s"), GetDeviceName().Array());
if(!bBrokenTimestamp)
newTimestamp = lastUsedTimestamp + numAudioFrames*1000/inputSamplesPerSec;
}
else
{
if(!bBrokenTimestamp)
newTimestamp = qpcTimestamp/10000;
/*UINT64 freq;
mmClock->GetFrequency(&freq);
Log(TEXT("position: %llu, numAudioFrames: %u, freq: %llu, newTimestamp: %llu, test: %llu"), devPosition, numAudioFrames, freq, newTimestamp, devPosition*8000/freq);*/
}
//have to do this crap to account for broken devices or device drivers. absolutely unbelievable.
if(!bFirstFrameReceived)
{
LARGE_INTEGER clockFreq;
QueryPerformanceFrequency(&clockFreq);
QWORD curTime = GetQPCTimeMS(clockFreq.QuadPart);
if(newTimestamp < (curTime-1000) || newTimestamp > (curTime+1000))
{
bBrokenTimestamp = true;
Log(TEXT("MMDeviceAudioSource::GetNextBuffer: Got bad audio timestamp offset %lld from device: '%s', timestamps for this device will be calculated. curTime: %llu, newTimestamp: %llu"), (LONGLONG)(newTimestamp - curTime), GetDeviceName().Array(), curTime, newTimestamp);
lastUsedTimestamp = newTimestamp = curTime;
}
else
lastUsedTimestamp = newTimestamp;
bFirstFrameReceived = true;
}
if(tempBuffer.Num() < numAudioFrames*2)
tempBuffer.SetSize(numAudioFrames*2);
outputBuffer = tempBuffer.Array();
float *tempOut = outputBuffer;
//------------------------------------------------------------
// channel upmix/downmix
if(inputChannels == 1)
{
UINT numFloats = numAudioFrames;
float *inputTemp = (float*)captureBuffer;
float *outputTemp = outputBuffer;
if(App->SSE2Available() && (UPARAM(inputTemp) & 0xF) == 0 && (UPARAM(outputTemp) & 0xF) == 0)
{
UINT alignedFloats = numFloats & 0xFFFFFFFC;
for(UINT i=0; i<alignedFloats; i += 4)
{
__m128 inVal = _mm_load_ps(inputTemp+i);
__m128 outVal1 = _mm_unpacklo_ps(inVal, inVal);
__m128 outVal2 = _mm_unpackhi_ps(inVal, inVal);
_mm_store_ps(outputTemp+(i*2), outVal1);
_mm_store_ps(outputTemp+(i*2)+4, outVal2);
}
numFloats -= alignedFloats;
inputTemp += alignedFloats;
outputTemp += alignedFloats*2;
}
while(numFloats--)
{
float inputVal = *inputTemp;
*(outputTemp++) = inputVal;
*(outputTemp++) = inputVal;
inputTemp++;
}
}
else if(inputChannels == 2) //straight up copy
{
if(App->SSE2Available())
SSECopy(outputBuffer, captureBuffer, numAudioFrames*2*sizeof(float));
else
mcpy(outputBuffer, captureBuffer, numAudioFrames*2*sizeof(float));
}
else
{
//todo: downmix optimization, also support for other speaker configurations than ones I can merely "think" of. ugh.
float *inputTemp = (float*)captureBuffer;
float *outputTemp = outputBuffer;
if(inputChannelMask == KSAUDIO_SPEAKER_QUAD)
{
UINT numFloats = numAudioFrames*4;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float rear = (inputTemp[2]+inputTemp[3])*surroundMix;
*(outputTemp++) = left - rear;
*(outputTemp++) = right + rear;
inputTemp += 4;
}
}
else if(inputChannelMask == KSAUDIO_SPEAKER_2POINT1)
{
UINT numFloats = numAudioFrames*3;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float lfe = inputTemp[2]*lowFreqMix;
*(outputTemp++) = left + lfe;
*(outputTemp++) = right + lfe;
inputTemp += 3;
}
}
else if(inputChannelMask == KSAUDIO_SPEAKER_4POINT1)
{
UINT numFloats = numAudioFrames*5;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float lfe = inputTemp[2]*lowFreqMix;
float rear = (inputTemp[3]+inputTemp[4])*surroundMix;
*(outputTemp++) = left + lfe - rear;
*(outputTemp++) = right + lfe + rear;
inputTemp += 5;
}
}
else if(inputChannelMask == KSAUDIO_SPEAKER_SURROUND)
{
UINT numFloats = numAudioFrames*4;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float center = inputTemp[2]*centerMix;
float rear = inputTemp[3]*(surroundMix*dbMinus3);
*(outputTemp++) = left + center - rear;
*(outputTemp++) = right + center + rear;
inputTemp += 4;
}
}
//don't think this will work for both
else if(inputChannelMask == KSAUDIO_SPEAKER_5POINT1)
{
UINT numFloats = numAudioFrames*6;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float center = inputTemp[2]*centerMix;
float lowFreq = inputTemp[3]*lowFreqMix;
float rear = (inputTemp[4]+inputTemp[5])*surroundMix;
*(outputTemp++) = left + center + lowFreq - rear;
*(outputTemp++) = right + center + lowFreq + rear;
inputTemp += 6;
}
}
//todo ------------------
//not sure if my 5.1/7.1 downmixes are correct
else if(inputChannelMask == KSAUDIO_SPEAKER_5POINT1_SURROUND)
{
UINT numFloats = numAudioFrames*6;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float center = inputTemp[2]*centerMix;
float lowFreq = inputTemp[3]*lowFreqMix;
float sideLeft = inputTemp[4]*dbMinus3;
float sideRight = inputTemp[5]*dbMinus3;
*(outputTemp++) = left + center + sideLeft + lowFreq;
*(outputTemp++) = right + center + sideRight + lowFreq;
inputTemp += 6;
}
}
else if(inputChannelMask == KSAUDIO_SPEAKER_7POINT1)
{
UINT numFloats = numAudioFrames*8;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float center = inputTemp[2]*(centerMix*dbMinus3);
float lowFreq = inputTemp[3]*lowFreqMix;
float rear = (inputTemp[4]+inputTemp[5])*surroundMix;
float centerLeft = inputTemp[6]*dbMinus6;
float centerRight = inputTemp[7]*dbMinus6;
*(outputTemp++) = left + centerLeft + center + lowFreq - rear;
*(outputTemp++) = right + centerRight + center + lowFreq + rear;
inputTemp += 8;
}
}
else if(inputChannelMask == KSAUDIO_SPEAKER_7POINT1_SURROUND)
{
UINT numFloats = numAudioFrames*8;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float center = inputTemp[2]*centerMix;
float lowFreq = inputTemp[3]*lowFreqMix;
float rear = (inputTemp[4]+inputTemp[5])*(surroundMix*dbMinus3);
float sideLeft = inputTemp[6]*dbMinus6;
float sideRight = inputTemp[7]*dbMinus6;
*(outputTemp++) = left + sideLeft + center + lowFreq - rear;
*(outputTemp++) = right + sideLeft + center + lowFreq + rear;
inputTemp += 8;
}
}
}
mmCapture->ReleaseBuffer(numAudioFrames);
//------------------------------------------------------------
// resample
if(bResample)
{
UINT frameAdjust = UINT((double(numAudioFrames) * resampleRatio) + 1.0);
UINT newFrameSize = frameAdjust*2;
if(tempResampleBuffer.Num() < newFrameSize)
tempResampleBuffer.SetSize(newFrameSize);
SRC_DATA data;
data.src_ratio = resampleRatio;
data.data_in = tempBuffer.Array();
data.input_frames = numAudioFrames;
data.data_out = tempResampleBuffer.Array();
data.output_frames = frameAdjust;
data.end_of_input = 0;
int err = src_process(resampler, &data);
if(err)
{
RUNONCE AppWarning(TEXT("Was unable to resample audio"));
return NoAudioAvailable;
}
if(data.input_frames_used != numAudioFrames)
{
RUNONCE AppWarning(TEXT("Failed to downsample buffer completely, which shouldn't actually happen because it should be using 10ms of samples"));
return NoAudioAvailable;
}
numAudioFrames = data.output_frames_gen;
}
//-----------------------------------------------------------------------------
// sort all audio frames into 10 millisecond increments (done because not all devices output in 10ms increments)
// NOTE: 0.457+ - instead of using the timestamps from windows, just compare and make sure it stays within a 100ms of their timestamps
float *newBuffer = (bResample) ? tempResampleBuffer.Array() : tempBuffer.Array();
if(storageBuffer.Num() == 0 && numAudioFrames == 441)
{
lastUsedTimestamp += 10;
if(!bBrokenTimestamp)
{
QWORD difVal = GetQWDif(newTimestamp, lastUsedTimestamp);
if(difVal > 70)
lastUsedTimestamp = newTimestamp;
}
if(lastUsedTimestamp > lastSentTimestamp)
{
QWORD adjustVal = (lastUsedTimestamp-lastSentTimestamp);
if(adjustVal < 10)
lastUsedTimestamp += 10-adjustVal;
AudioSegment &newSegment = *audioSegments.CreateNew();
newSegment.audioData.CopyArray(newBuffer, numAudioFrames*2);
newSegment.timestamp = lastUsedTimestamp;
MultiplyAudioBuffer(newSegment.audioData.Array(), numAudioFrames*2, curVolume);
lastSentTimestamp = lastUsedTimestamp;
}
}
else
{
UINT storedFrames = storageBuffer.Num();
storageBuffer.AppendArray(newBuffer, numAudioFrames*2);
if(storageBuffer.Num() >= (441*2))
{
lastUsedTimestamp += 10;
if(!bBrokenTimestamp)
{
QWORD difVal = GetQWDif(newTimestamp, lastUsedTimestamp);
if(difVal > 70)
lastUsedTimestamp = newTimestamp - (QWORD(storedFrames)/2*1000/44100);
}
//------------------------
// add new data
if(lastUsedTimestamp > lastSentTimestamp)
{
QWORD adjustVal = (lastUsedTimestamp-lastSentTimestamp);
if(adjustVal < 10)
lastUsedTimestamp += 10-adjustVal;
AudioSegment &newSegment = *audioSegments.CreateNew();
newSegment.audioData.CopyArray(storageBuffer.Array(), (441*2));
newSegment.timestamp = lastUsedTimestamp;
MultiplyAudioBuffer(newSegment.audioData.Array(), 441*2, curVolume);
storageBuffer.RemoveRange(0, (441*2));
}
//------------------------
// if still data pending (can happen)
while(storageBuffer.Num() >= (441*2))
{
lastUsedTimestamp += 10;
if(lastUsedTimestamp > lastSentTimestamp)
{
QWORD adjustVal = (lastUsedTimestamp-lastSentTimestamp);
if(adjustVal < 10)
lastUsedTimestamp += 10-adjustVal;
AudioSegment &newSegment = *audioSegments.CreateNew();
newSegment.audioData.CopyArray(storageBuffer.Array(), (441*2));
storageBuffer.RemoveRange(0, (441*2));
MultiplyAudioBuffer(newSegment.audioData.Array(), 441*2, curVolume);
newSegment.timestamp = lastUsedTimestamp;
lastSentTimestamp = lastUsedTimestamp;
}
}
}
}
//-----------------------------------------------------------------------------
return ContinueAudioRequest;
}
return NoAudioAvailable;
}
UINT AudioSource::QueryAudio(float curVolume)
{
LPVOID buffer;
UINT numAudioFrames;
QWORD newTimestamp;
if(GetNextBuffer((void**)&buffer, &numAudioFrames, &newTimestamp))
{
//------------------------------------------------------------
// convert to float
float *captureBuffer;
if(!bFloat)
{
UINT totalSamples = numAudioFrames*inputChannels;
if(convertBuffer.Num() < totalSamples)
convertBuffer.SetSize(totalSamples);
if(inputBitsPerSample == 8)
{
float *tempConvert = convertBuffer.Array();
char *tempSByte = (char*)buffer;
while(totalSamples--)
{
*(tempConvert++) = float(*(tempSByte++))/127.0f;
}
}
else if(inputBitsPerSample == 16)
{
float *tempConvert = convertBuffer.Array();
short *tempShort = (short*)buffer;
while(totalSamples--)
{
*(tempConvert++) = float(*(tempShort++))/32767.0f;
}
}
else if(inputBitsPerSample == 24)
{
float *tempConvert = convertBuffer.Array();
BYTE *tempTriple = (BYTE*)buffer;
TripleToLong valOut;
while(totalSamples--)
{
TripleToLong &valIn = (TripleToLong&)tempTriple;
valOut.wVal = valIn.wVal;
valOut.tripleVal = valIn.tripleVal;
if(valOut.tripleVal > 0x7F)
valOut.lastByte = 0xFF;
*(tempConvert++) = float(double(valOut.val)/8388607.0);
tempTriple += 3;
}
}
else if(inputBitsPerSample == 32)
{
float *tempConvert = convertBuffer.Array();
long *tempShort = (long*)buffer;
while(totalSamples--)
{
*(tempConvert++) = float(double(*(tempShort++))/2147483647.0);
}
}
captureBuffer = convertBuffer.Array();
}
else
captureBuffer = (float*)buffer;
//------------------------------------------------------------
// channel upmix/downmix
if(tempBuffer.Num() < numAudioFrames*2)
tempBuffer.SetSize(numAudioFrames*2);
float *dataOutputBuffer = tempBuffer.Array();
float *tempOut = dataOutputBuffer;
if(inputChannels == 1)
{
UINT numFloats = numAudioFrames;
float *inputTemp = (float*)captureBuffer;
float *outputTemp = dataOutputBuffer;
if((UPARAM(inputTemp) & 0xF) == 0 && (UPARAM(outputTemp) & 0xF) == 0)
{
UINT alignedFloats = numFloats & 0xFFFFFFFC;
for(UINT i=0; i<alignedFloats; i += 4)
{
__m128 inVal = _mm_load_ps(inputTemp+i);
__m128 outVal1 = _mm_unpacklo_ps(inVal, inVal);
__m128 outVal2 = _mm_unpackhi_ps(inVal, inVal);
_mm_store_ps(outputTemp+(i*2), outVal1);
_mm_store_ps(outputTemp+(i*2)+4, outVal2);
}
numFloats -= alignedFloats;
inputTemp += alignedFloats;
outputTemp += alignedFloats*2;
}
while(numFloats--)
{
float inputVal = *inputTemp;
*(outputTemp++) = inputVal;
*(outputTemp++) = inputVal;
inputTemp++;
}
}
else if(inputChannels == 2) //straight up copy
{
SSECopy(dataOutputBuffer, captureBuffer, numAudioFrames*2*sizeof(float));
}
else
{
//todo: downmix optimization, also support for other speaker configurations than ones I can merely "think" of. ugh.
float *inputTemp = (float*)captureBuffer;
float *outputTemp = dataOutputBuffer;
if(inputChannelMask == KSAUDIO_SPEAKER_QUAD)
{
UINT numFloats = numAudioFrames*4;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float rearLeft = inputTemp[2]*surroundMix4;
float rearRight = inputTemp[3]*surroundMix4;
// When in doubt, use only left and right .... and rear left and rear right :)
// Same idea as with 5.1 downmix
*(outputTemp++) = (left + rearLeft) * attn4dotX;
*(outputTemp++) = (right + rearRight) * attn4dotX;
inputTemp += 4;
}
}
else if(inputChannelMask == KSAUDIO_SPEAKER_2POINT1)
{
UINT numFloats = numAudioFrames*3;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
// Drop LFE since we don't need it
//float lfe = inputTemp[2]*lowFreqMix;
*(outputTemp++) = left;
*(outputTemp++) = right;
inputTemp += 3;
}
}
else if(inputChannelMask == KSAUDIO_SPEAKER_4POINT1)
{
UINT numFloats = numAudioFrames*5;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
// Skip LFE , we don't really need it.
//float lfe = inputTemp[2];
float rearLeft = inputTemp[3]*surroundMix4;
float rearRight = inputTemp[4]*surroundMix4;
// Same idea as with 5.1 downmix
*(outputTemp++) = (left + rearLeft) * attn4dotX;
*(outputTemp++) = (right + rearRight) * attn4dotX;
inputTemp += 5;
}
}
else if(inputChannelMask == KSAUDIO_SPEAKER_SURROUND)
{
UINT numFloats = numAudioFrames*4;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float frontCenter = inputTemp[2];
float rearCenter = inputTemp[3];
// When in doubt, use only left and right :) Seriously.
// THIS NEEDS TO BE PROPERLY IMPLEMENTED!
*(outputTemp++) = left;
*(outputTemp++) = right;
inputTemp += 4;
}
}
// Both speakers configs share the same format, the difference is in rear speakers position
// See: http://msdn.microsoft.com/en-us/library/windows/hardware/ff537083(v=vs.85).aspx
// Probably for KSAUDIO_SPEAKER_5POINT1_SURROUND we will need a different coefficient for rear left/right
else if(inputChannelMask == KSAUDIO_SPEAKER_5POINT1 || inputChannelMask == KSAUDIO_SPEAKER_5POINT1_SURROUND)
{
UINT numFloats = numAudioFrames*6;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float center = inputTemp[2]*centerMix;
//We don't need LFE channel so skip it (see below)
//float lowFreq = inputTemp[3]*lowFreqMix;
float rearLeft = inputTemp[4]*surroundMix;
float rearRight = inputTemp[5]*surroundMix;
// According to ITU-R BS.775-1 recommendation, the downmix from a 3/2 source to stereo
// is the following:
// L = FL + k0*C + k1*RL
// R = FR + k0*C + k1*RR
// FL = front left
// FR = front right
// C = center
// RL = rear left
// RR = rear right
// k0 = centerMix = dbMinus3 = 0.7071067811865476 [for k0 we can use dbMinus6 = 0.5 too, probably it's better]
// k1 = surroundMix = dbMinus3 = 0.7071067811865476
// The output (L,R) can be out of (-1,1) domain so we attenuate it [ attn5dot1 = 1/(1 + centerMix + surroundMix) ]
// Note: this method of downmixing is far from "perfect" (pretty sure it's not the correct way) but the resulting downmix is "okayish", at least no more bleeding ears.
// (maybe have a look at http://forum.doom9.org/archive/index.php/t-148228.html too [ 5.1 -> stereo ] the approach seems almost the same [but different coefficients])
// http://acousticsfreq.com/blog/wp-content/uploads/2012/01/ITU-R-BS775-1.pdf
// http://ir.lib.nctu.edu.tw/bitstream/987654321/22934/1/030104001.pdf
*(outputTemp++) = (left + center + rearLeft) * attn5dot1;
*(outputTemp++) = (right + center + rearRight) * attn5dot1;
inputTemp += 6;
}
}
// According to http://msdn.microsoft.com/en-us/library/windows/hardware/ff537083(v=vs.85).aspx
// KSAUDIO_SPEAKER_7POINT1 is obsolete and no longer supported in Windows Vista and later versions of Windows
// Not sure what to do about it, meh , drop front left of center/front right of center -> 5.1 -> stereo;
else if(inputChannelMask == KSAUDIO_SPEAKER_7POINT1)
{
UINT numFloats = numAudioFrames*8;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float center = inputTemp[2] * centerMix;
// Drop LFE since we don't need it
//float lowFreq = inputTemp[3]*lowFreqMix;
float rearLeft = inputTemp[4] * surroundMix;
float rearRight = inputTemp[5] * surroundMix;
// Drop SPEAKER_FRONT_LEFT_OF_CENTER , SPEAKER_FRONT_RIGHT_OF_CENTER
//float centerLeft = inputTemp[6];
//float centerRight = inputTemp[7];
// Downmix from 5.1 to stereo
*(outputTemp++) = (left + center + rearLeft) * attn5dot1;
*(outputTemp++) = (right + center + rearRight) * attn5dot1;
inputTemp += 8;
}
}
// Downmix to 5.1 (easy stuff) then downmix to stereo as done in KSAUDIO_SPEAKER_5POINT1
else if(inputChannelMask == KSAUDIO_SPEAKER_7POINT1_SURROUND)
{
UINT numFloats = numAudioFrames*8;
float *endTemp = inputTemp+numFloats;
while(inputTemp < endTemp)
{
float left = inputTemp[0];
float right = inputTemp[1];
float center = inputTemp[2] * centerMix;
// Skip LFE we don't need it
//float lowFreq = inputTemp[3]*lowFreqMix;
float rearLeft = inputTemp[4];
float rearRight = inputTemp[5];
float sideLeft = inputTemp[6];
float sideRight = inputTemp[7];
// combine the rear/side channels first , baaam! 5.1
rearLeft = (rearLeft + sideLeft) * 0.5f;
rearRight = (rearRight + sideRight) * 0.5f;
// downmix to stereo as in 5.1 case
*(outputTemp++) = (left + center + rearLeft * surroundMix) * attn5dot1;
*(outputTemp++) = (right + center + rearRight * surroundMix) * attn5dot1;
inputTemp += 8;
}
}
}
ReleaseBuffer();
//------------------------------------------------------------
// resample
if(bResample)
{
UINT frameAdjust = UINT((double(numAudioFrames) * resampleRatio) + 1.0);
UINT newFrameSize = frameAdjust*2;
if(tempResampleBuffer.Num() < newFrameSize)
tempResampleBuffer.SetSize(newFrameSize);
SRC_DATA data;
data.src_ratio = resampleRatio;
data.data_in = tempBuffer.Array();
data.input_frames = numAudioFrames;
data.data_out = tempResampleBuffer.Array();
data.output_frames = frameAdjust;
data.end_of_input = 0;
int err = src_process((SRC_STATE*)resampler, &data);
if(err)
{
RUNONCE AppWarning(TEXT("AudioSource::QueryAudio: Was unable to resample audio for device '%s'"), GetDeviceName());
return NoAudioAvailable;
}
if(data.input_frames_used != numAudioFrames)
{
RUNONCE AppWarning(TEXT("AudioSource::QueryAudio: Failed to downsample buffer completely, which shouldn't actually happen because it should be using 10ms of samples"));
return NoAudioAvailable;
}
numAudioFrames = data.output_frames_gen;
}
//-----------------------------------------------------------------------------
// sort all audio frames into 10 millisecond increments (done because not all devices output in 10ms increments)
// NOTE: 0.457+ - instead of using the timestamps from windows, just compare and make sure it stays within a 100ms of their timestamps
if(!bFirstBaseFrameReceived)
{
lastUsedTimestamp = newTimestamp;
bFirstBaseFrameReceived = true;
}
float *newBuffer = (bResample) ? tempResampleBuffer.Array() : tempBuffer.Array();
if (bSmoothTimestamps) {
lastUsedTimestamp += 10;
QWORD difVal = GetQWDif(newTimestamp, lastUsedTimestamp);
if(difVal > 70)
{
//OSDebugOut(TEXT("----------------------------1\r\nlastUsedTimestamp before: %llu - device: %s\r\n"), lastUsedTimestamp, GetDeviceName());
lastUsedTimestamp = newTimestamp;
//OSDebugOut(TEXT("lastUsedTimestamp after: %llu\r\n"), lastUsedTimestamp);
}
if(lastUsedTimestamp > lastSentTimestamp)
{
QWORD adjustVal = (lastUsedTimestamp-lastSentTimestamp);
if(adjustVal < 10)
lastUsedTimestamp += 10-adjustVal;
AudioSegment *newSegment = new AudioSegment(newBuffer, numAudioFrames*2, lastUsedTimestamp);
AddAudioSegment(newSegment, curVolume*sourceVolume);
lastSentTimestamp = lastUsedTimestamp;
}
} else {
// OSDebugOut(TEXT("newTimestamp: %llu\r\n"), newTimestamp);
AudioSegment *newSegment = new AudioSegment(newBuffer, numAudioFrames*2, newTimestamp);
AddAudioSegment(newSegment, curVolume*sourceVolume);
}
//-----------------------------------------------------------------------------
return AudioAvailable;
}
return NoAudioAvailable;
}
QWORD MMDeviceAudioSource::GetTimestamp(QWORD qpcTimestamp)
{
QWORD newTimestamp;
if(bIsMic)
{
newTimestamp = (bUseQPC) ? qpcTimestamp : App->GetAudioTime();
newTimestamp += GetTimeOffset();
//Log(TEXT("got some mic audio, timestamp: %llu"), newTimestamp);
return newTimestamp;
}
else
{
//we're doing all these checks because device timestamps are only reliable "sometimes"
if(!bFirstFrameReceived)
{
QWORD curTime = GetQPCTimeMS();
newTimestamp = qpcTimestamp;
curVideoTime = lastVideoTime = App->GetVideoTime();
if(bUseVideoTime || newTimestamp < (curTime-App->bufferingTime) || newTimestamp > (curTime+2000))
{
if(!bUseVideoTime)
Log(TEXT("Bad timestamp detected, syncing audio to video time"));
else
Log(TEXT("Syncing audio to video time (WARNING: you should not be doing this if you are just having webcam desync, that's a separate issue)"));
SetTimeOffset(GetTimeOffset()-int(lastVideoTime-App->GetSceneTimestamp()));
bUseVideoTime = true;
newTimestamp = lastVideoTime+GetTimeOffset();
}
lastUsedTimestamp = newTimestamp;
bFirstFrameReceived = true;
}
else
{
QWORD newVideoTime = App->GetVideoTime();
if(newVideoTime != lastVideoTime)
curVideoTime = lastVideoTime = newVideoTime;
else
curVideoTime += 10;
newTimestamp = (bUseVideoTime) ? curVideoTime : qpcTimestamp;
newTimestamp += GetTimeOffset();
lastUsedTimestamp += 10;
}
//------------------------------------------------------
// timestamp smoothing
QWORD difVal = GetQWDif(newTimestamp, lastUsedTimestamp);
//Log(TEXT("qpc timestamp: %llu, lastUsed: %llu, dif: %llu"), newTimestamp, lastUsedTimestamp, difVal);
if (difVal > 70) {
/*QWORD curTimeMS = App->GetVideoTime()-App->GetSceneTimestamp();
UINT curTimeTotalSec = (UINT)(curTimeMS/1000);
UINT curTimeTotalMin = curTimeTotalSec/60;
UINT curTimeHr = curTimeTotalMin/60;
UINT curTimeMin = curTimeTotalMin-(curTimeHr*60);
UINT curTimeSec = curTimeTotalSec-(curTimeTotalMin*60);
Log(TEXT("A timestamp adjustment was encountered for device %s, approximate stream time is: %u:%u:%u, prev value: %llu, new value: %llu"), GetDeviceName(), curTimeHr, curTimeMin, curTimeSec, lastUsedTimestamp, newTimestamp);*/
lastUsedTimestamp = newTimestamp;
}
//Log(TEXT("got some desktop audio, timestamp: %llu"), lastUsedTimestamp);
return lastUsedTimestamp;
}
}
