2

I'm writing a program in C++, on Windows. I need to support Windows Vista+.

I want to record both the microphone and speaker simultaneously.

I'm using the WASAPI and can record the microphone and speaker separately, but I would like to have just one stream supplying me the input from both streams (for example, for recording a client play the guitar along with the music he hears on his headphones), instead of merging the two buffers together somehow (which I guess will lead me to timing issues).

Is there a way to do this?

The Paramagnetic Croissant
  • 9,223
  • 3
  • 25
  • 38
Erez Eshkol
  • 159
  • 4
  • If you're dealing with realtime audio, you need to make sure your capture thread(s?) has appropriate priority - and if you have it and call the APIs properly, the timing shouldn't be a problem. – o11c May 21 '16 at 00:04

1 Answers1

0

I'm actually working on a library which can do exactly that, merge streams from multiple devices. You might want to give it a try: see xt-audio.com. If you're implementing this yourself, here's some things to consider:

  • If you're capturing the speakers through a WASAPI loopback interface you're operating in shared mode, in this case latency might be unacceptable for live performance. If possible stick to exclusive mode and use a loopback cable or hardware loopback device if you have one (e.g. the old fashioned "stereo mix" devices etc).
  • If you're merging buffers then yes, you're going to have timing issues. This is generally unavoidable when syncing independent devices. Pops/clicks can largely be avoided using a secondary intermediate buffer which introduces additional latency, but eventually you're going to have to pad/drop some samples to keep streams in sync.

  • Do NOT use separate threads for each independent stream. This will increase context switches and thereby increase the minimum achievable latency. Instead, designate one device as the master device, wait for that device's event to be raised, then read input from all devices whether they are "ready" or not (this is were dropping/padding comes into play).

  • In general you can get really decent performance from WASAPI exclusive mode, even running multiple streams together. But for something as critical as live performance you might want to consider a pro audio interface with ASIO drivers where everything just ticks off the same clock, or synchronization is at least handled at the driver level.

Sjoerd van Kreel
  • 1,000
  • 6
  • 19