Navi uses the same tech, only the in-call app isn’t a movieit’s a real-time translation engine.
Then, other participants can also join the action and see live subtitles for the current speaker.
If somebody is monologuing, their speech bubble grows and sticks around a little longer.
Jasmin Merdan / Getty Images
For the deaf, this could mean the difference between calling people or not.
And for anyone, it means you might have useful conversations between people who don’t share a language.
Universal Text
The internet is built on text, and thats great.
Its small and easy to create, read, and translate.
Its also simple to turn into synthesized speech.
The result is that anyone from anywhere can participate in any conversation.
Navi
But the spoken word is much harder to process.
Introduced in iOS 15, it offers real-time audio translations.
If we still went on foreign vacations, it would be perfect.
Now we use video more and more for work and to stay in touch with friends and family.
No matter how we work in the future, the barrier to video calls has been thoroughly smashed.
Its now a common tool, but it lacks a lot of the finesse of written communication tools.
Something like Navi, which offers real-time subtitles and translation, could be significant.
Navi
In Action
I tested Navi with app developer, author, and hearing-aid userGraham Bower.
It’s pretty good but not yet ready for critical tasks.
Some of the transcriptions were comically bad and too vulgar to relate.
As our conversation went on, though, it got a lot better at accurately recognizing his speech.
That makes sense because the iOS dictation engine adapts to your voice over time.
The translation also worked, although the quality of its translations depends on the accuracy of the input.
“I can see this working in AR glasses,” said Bower during our conversation.
“Some people, even with normal hearing, prefer subtitles in movies.
This would be like subtitles for real life.”
While an impressive tech demo, Navi isn’t there yet.
For reliable business use, Apple’s initial speech recognition will have to get a lot more accurate.
But speed-wise, it’s fine, and the translations are as good as any.
But we’re on the path now, and this kind of thing will only get better.
