I mean, this is basically just a wrapper around whisper, which has been around for a while now, so IMO this is nothing revolutionary. And the “small” model used is actually quite inaccurate, but the large model is not capable of real-time use on a typical system.
Just having read the README, even if it's a little faffy, the idea of being able to converse with someone and have it be translated back in near real-time is awesome!
This is the kind of application I was looking forward to seeing come out of open source AI - I think it'd still be a little too clunky to save you in an emergency, but probably less clunky for most people than learning a new language or having a human translator on retainer haha
That’s something we do need if we don’t want to stay behind the closed source empires. I hope they’ll get rid of that proprietary blob at some point. And please get a decent name when it’s more ready.
Add comment