In the wake of the past few years, virtual meetings have changed the way we communicate. However, we have generally embraced remote communication without considering how remote communication differs from the in-person or phone communication we’ve traditionally used. While turning on cameras and speaking face to face gives us more connection, and embracing a slower and more formal discussion style can make the meeting more enjoyable, the most essential communication still happens through audio.
As the audio quality we use gets better, our meetings can be more effective as our colleagues can spend their time thinking about our discussion instead of being distracted by poor audio quality. While our headphones might tune out our neighbour’s jackhammer, we don’t want to invite it into a virtual meeting!
To improve audio quality, we need to start by recording a good quality audio signal. It’s much easier to start with a good signal, instead of doing a lot of work to clean it up later. This can be done by getting a good quality microphone:
- Headset mounted microphones or earbuds tend to be the best out of the box as they keep you at a consistent distance from the microphone, and keep the microphone much closer than most background audio sources. However, most people don’t like to wear these microphones all day, and usually move them all over the place. Consequently, people with headset microphones are also most likely to be reminded that no-one can hear them because they moved their microphone.
- A high-quality desk microphone rated for podcasting allows you to setup your microphone and know it will just work all day long. When wired in, they work all day long without needing to be reconfigured or running out of batteries. When paired with digital audio software, these microphones will give you the best audio quality. While desk microphones do come with a marginally higher cost, there are good budgets options available. The higher-end options with shock mounts, tripods, and specialized condenser microphones are useful for audio and video recording, but don’t make a difference for online meetings.
- A good quality webcam will have a microphone built in. Depending on the quality of your webcam, both the video and audio quality vary widely. This is one area in tech where you definitely get what you pay for: better quality microphones will be acceptable if paired with software, but lower quality equipment has poor audio quality which is difficult to correct.
- You can use alternative hardware such as an old phone with an IP-Video app, or an SLR digital camera, but these generally have significant drawbacks better left avoided unless you have no other option, or are spending enough money to get a professional videography rig.
Once we have a microphone setup, we want to correct the audio signal to eliminate any static or background noise. A great example of what is possible with digital audio correction is @theGunrun’s demonstration.
To setup audio correction, we need:
- Audacity so we can record ourselves and hear what we sound like
- EqualizerAPO so we can EQ our incoming audio feeds
- Werman’s Noise Suppression VST Plugin to apply AI noise correction.
Note: If you have a qualifying graphics card, you may wish to use NVIDIA Broadcast as a more user friendly alternative to EqualizerAPI+werman. Make sure to test it thoroughly as the beta version predecessor (NVIDIA RTX Voice) had a bug where it would randomly turn the audio feed into maximum volume static with no warning.
To Configure EqualizerAPO, you need to:
- Run the installer
- Run the configurator and enable equalization on any mic or speaker you wish to correct
- Noise cancellation works best on your microphone. It’s mostly ineffective on correcting your speaker feed unless your colleagues’ microphones are exceptionally poor.
- Restart your machine
- Run the Configuration Editor and create a new profile
- Filter to just your microphone, so any other devices aren’t incidentally affected
- Add a preamp if needed. Sometimes EQ software will drop your volume; an adjustment of 0-10db can cleanup the signal.
- Include the example.txt to include a base frequency eq. Usually this makes people sound better by slightly amplifying midrange tones. You can safely leave this step out if you don’t like the adjustment.
- Import a VST Plugin and select \vst\rnnoise_stereo.dll. You want this to run on the best possible signal, so this should be the last item in the EQ sequence for best audio quality.
- Record and replay a test audio sample with Audacity. You can turn individual EQ effects on and off in real time while recording a test sample to see how effective the audio quality is. Ideally you want the audio sample to be flat, without a background hum, unless you are speaking, where the audio should be clear.
- Active noise cancellation works very well to remove background noise, but some loud sounds like kettles and box fans at maximum volume can confuse the AI. This is the audio equivalent of a virtual background hiding a person in a dark room. Always be sure to run a self test with Audacity after any major audio changes.
- Some Windows updates disable equalization. Re-running the configurator to select your hardware and restarting will re-enable audio correction.