Alterego's 'Silent Speech' Could Be the Answer to Dictating Text in Public

That's someone that uses speech text for much more than just typing. This is amazing potential, but it has a very long way.

What kind of compute power will be needed for local inference? Its latency is pretty bad, which makes interaction tiresome. If you truly want to support the disabled, it needs to be open source at one point.

Even if it had 99.9% accuracy is not enough. One correction out of every 101 ish words.

Not insurmountable, but we will see.

This would honestly be really awesome. I barely type when I can voice dictate at this point, even though I'm a very fast and comfortable typist, especially with my ergonomic keyboard, but sometimes voice dictation just really is awkward or not an option, and this would completely solve that problem. Not to mention, it would solve the problem of background noise. Funnily enough, in all of the science fiction novels that I write, the main interface characters have with their computers — which I represent as augmented reality glasses similar to Meta's Orion Project — is subvocalization.