Struggles of noise reduction in RTC — Part 1


Thus, noise reduction at the bare bones would be simply chopping off these unwanted frequencies and at the highest level would be just keeping the intended user’s voice in the stream and removing just about everything else.

The heterogeneity in the web doesn’t end being a problem — the fact that your application can be used by folks with a $500 laptop or an all maxed out MacBook Pro, doesn’t help!

Most applications now are left with two options — hope that your users don’t use a potato for video/voice calls OR throw some good deal of money at the problem, and perform noise reduction on their stream while the stream is in transit

The Expensive Route

While Google has all the resources, both monetary and in terms of engineering to throw at this problem, a startup at its early phases of development most likely won’t.

The Pragmatic Route

Up next...

My work straddles around webRTC, noise reduction, and everything to do with audio over the web