Launched this week
Gemini 3.5 Live Translate
Latest audio model for live speech-to-speech translation
397 followers
Latest audio model for live speech-to-speech translation
397 followers
Gemini 3.5 Live Translate brings near real-time, natural speech translation to Google AI Studio, Google Translate and Google Meet.
Degraded speech patterns, variable speaking rate, vocal fatigue, and noisy environments all stand in the way of finding invariant speech forms across naturally spoken signals. I wonder how new technologies will help us overcome these barriers to fully understanding speech perception and spoken word recognition. Perhaps tools like Gemini's Live Translate will help us identify additional cues that we rely on for language perception which have been eluding empirical investigators in scientific research(?). Perhaps not.
If Gemini 3.5 handles those edge cases well, it changes what's possible for multilingual meetings. How does the system handle speakers with heavy accents or non-native pronunciation patterns?
Live translation in Google Meet with zero setup is what really matters for schools. Parent–teacher calls across language barriers usually needed a third person in the room - now not anymore.
looks great, will it be in Google Meets for everybody? or...
Any benchmarks to share? Like what is the latency, accuracy across different languages, etc.?
Does the translated speech preserve tone and emphasis, or is it flattened into neutral delivery?
+1 to the accent question. loom's transcript misses quite a bit for non-native speakers - would be awesome if gemini tackles this!