2 ms·
You are downvoted but I believe you are essentially correct. It is ridiculous to send voice data all over the globe to be interpreted and send back to IOT devic
by raxxorrax 7y ago
You are downvoted but I believe you are essentially correct. It is ridiculous to send voice data all over the globe to be interpreted and send back to IOT devices. Unnecessary and certainly not elegant, since mic and device are most often in very close proximity.
Good for learning to analyze voice on the other hand, but your point is still valid. We will get a compressed voice service that easily can be held in a small device at some point, that makes this ridiculous data exchange unnecessary.
- cameronbrown 7y agoAt I/O Google announced on-device transcription and NLP: https://www.youtube.com/watch?v=b9LQX9cLnZk https://www.youtube.com/watch?v=b9LQX9cLnZk
- mrguyorama 7y agoWindows has had no-internet-required voice recognition built in since XP. It is available for any programmer to use for any purpose through importing a DLL, and isn't too complicated to use. With an unconstrained grammar defined, it's something like 85% accurate, but depending on how much you constrain the grammar, can get incredibly accurate. Meanwhile my mother's flip phone in 2003ish also had on device voice control, and it worked pretty damn effectively