Full duplex vs half duplex - the spectrum of AI voice models [D]
It seems that there are two ways to build voice AI: Half-duplex: strict turn-taking. You speak, the other side waits until you’re done, one direction of speech at a time. ← This is how almost every voice assistant works today. Full-duplex: two channels, both sides can talk at any time - no more waiting for your “turn”. ← This is the way humans actually talk. In fact, there are three crucial things half-duplex voice models can't really do: Overlap - talking and listening at the same time without falling apar
评论
?
参与讨论