Being able to validate that audio configuration is correctly set up is key to ensuring your users have a successful interaction every single time. Below you'll find some key things to confirm should before setting users live:
Common Mic/Audio Configurations to Check
Configure your app to pull audio from the correct input device
Depending on how you have your application configured, sometimes if you don’t explicitly pick the device ID, the system may default to an internal mic or a silent channel.
Common problems:
Your app opens “default input device” → OS chooses the wrong one.
The device being used exposes multiple channels; your app might be reading from a channel with no active beam.
Potential Solution:
Enumerate devices → select the necessary device by exact device ID → log number of channels → read from the active one(s).
Confirm Sample Rate and Format Requirements
Your capture code should match what the device provides. For example, one device some organizations use has the following specs:
48 kHz
24-bit or 32-bit float
Multiple channels
If your app tries to open at 44.1 kHz, mono, or 16-bit, the OS often returns silence, not an error.
Note: Other apps like Zoom/Teams automatically negotiate correct formats. Your app must request formats the device supports or do resampling.
Ensure other apps aren't going to steal the uudio
If Zoom/Teams/Avaya/NICE/etc. is running in the background, they may take:
Exclusive WASAPI access
DSP control session
Driver lock
Your app then receives zero frames.
Other apps based on WebRTC usually release shared access; low-level native audio apps often don’t.
Test by quitting all apps that might use the mic → restart your service.
Confirm your app’s audio buffer size / callback timing is compatible
If the mic is supplying frames at 10 ms frames and your app expects 20 ms (or vice versa), you can get:
Stutters
Dropped frames
All-zero buffers
Most conferencing tools adapt automatically; your app may not.
Match the buffer size to what the audio device wants (usually negotiable via WASAPI or CoreAudio).
When in Doubt, Ask Us
We at Corti are here to help! If you're running tests and something doesn't seem like it's behaving quite right, reach out to us at [email protected]. We can extract the audio from your interaction, just send us over the interaction ID. That can help us to start looking at if we're receiving good audio and inform future troubleshooting.