If you’ve been working from home for the last week alongside most of commercial humanity, then you’ve probably been exposed to the frustration of having a voice call catch-up with colleagues where at least one connection is, for want of a better word, wobbly. The good news is that Google has leveraged new machine learning technology to create a solution that may at least solve some of these concerns. The bad news? It only works on one of Google’s many messaging services, and it’s the one you’ve likely never used.
Rolling out now on Google Duo, WaveNetEQ is a new solution from Google’s DeepMind division which intends to solve jittery audio calls. WaveNetEQ, specifically, replaced audio jitter with artificial noise that’s designed to sound akin to human speech. The end benefit is that, when using the technology, jittery audio calls retain a semblance of greater quality and context, and sound far less stop and start (staccato) than they may usually over a bad connection.
There are limits to how well WaveNetEQ can function, however. Typically, existing NetEQ technology – which already ‘fills’ in broken soundwaves – functions by filling about 60ms gaps between spoken words. WaveNetEQ, on the other hand, can roughly perform double this load – up to 120ms – with arguably better results.
Take a listen for yourself, below – the first example is traditional NetEQ, while the latter is WaveNetEQ.
Whether Google plans to unveil the technology to other partner services or entirely third-party messaging apps remains unclear – however, the company has confirmed that the system will arrive on other handsets beyond its Pixel range in the near future. The revelation is well-timed, given the number of workplaces transitioning to remote offices at present.
What are your thoughts? Would WaveNetEQ make your work or personal life easier? Be sure to let us know your opinion in the comments below!
