Ask HN: Why haven't we seen AI enhanced AirPods microphone yet?
Wouldn’t Apple be able to train a model that with AirPods mic recording vs high-quality mic recording to effectively upconvert the poor mic audio into higher quality audio.
I desperately want this feature to improve my audio on work calls all day long. Right now there is an unbelievable difference in quality between AirPods mics and sub 200 dollar wired mics.
Why not use a better mic?
:) I mean, you always by a faster car. My interest is in the ability to enhance the voice quality of the uber-portable AirPods using AI.
There's a lot of reasons. The largest reason though is that an Airpods mic consumes ~100x less power than the $200 XLR mic and therefore needs a 100x lower impedence microphone. The low-detail recordings are a feature, not a bug.
Can’t you do an AI power improvement of the input sound on your computer, from where you’re doing work video calls?
No? You can amplify the audio digitally (which also amplifies analog imperfections like noise/ground hum/wind) or you can try fancy denoising algorithms which are incredibly lossy and imperfect.
I don't know of any "download better audio" solutions to-wit, at least from my time handling live audio.
Exactly. This was my point. Televisions can upconvert from 720p to 4k. In the same sense, the machine learning model would fill in the waveform and mimic a high powered mic. It can do this at the connection point (iPhone / computer).
Televisions have considerably more temporal data to work with than an audio stream does. It's very easy to hack together interpolated images, not so easy to predict/denoise/upres time-series audio information.
Past a certain point it's probably easier/more efficient to use the Airpods as a speech-to-text mic and then infer a "high quality" text-to-speech version on your connected device.