What makes real-time AI translation difficult?
Hi everyone,


I'm interested in the challenges involved in real-time AI translation.


Text translation is already very useful, but live conversations seem much more complicated. People speak at different speeds, interrupt each other, use informal language, change topics, and sometimes mix multiple languages in the same conversation.


I'm curious about the technical and product challenges involved.


For those who have worked with speech recognition, voice AI, or multilingual systems:


What do you think is the biggest challenge in making real-time translation feel natural during an actual conversation?


I'd be interested to hear different perspective