Real-time Speech Translation (interpreter mode)
Speech translation on the robot itself: to understand a sentence, or for two people who speak different languages.
How it works
Your sentence is recognized on the robot, translated by the on-device model and spoken in the target language's voice.
Video of translating a conversation between two people, with bilingual subtitles
Three benefits
Made for everyday conversation
Languages are recognized separately
At home / In the cloud
- At home
- Speech recognition, translation and speech output.
- In the cloud
- When no on-device translation exists for a language pair.
Today's limits
- Interpreter mode between two people has not been built yet.
- The translation service is installed on the robot, but a full spoken translation has not been tested yet.
- Translation quality is not the same for every language pair.
Capabilities in this pillar
On-device Speech Translation
Built, not yet connected or tested
- Today
- The translation service is installed and running on the robot, but a full spoken translation has not yet been tested with you.
- Where
- On the robot; a text sample on this site
- Blocker
- No real spoken translation test has been recorded.
- Next step
- Test a spoken translation on the robot.
Try the sample: Translation sample on this site
Evidence
- Tester or source
- Internal evaluation record
- Date
- Scope
- Service health response only
Interpreter mode
Future plan
- Today
- Not usable yet.
- Where
- —
- Blocker
- Two-person interpreter mode has not been built yet.
- Next step
- Build and test interpreter mode.
Evidence
- Tester or source
- Internal evaluation record
- Date
- Scope
- Work-in-progress inventory
FAQ
Which languages?
Persian and English come first; speech recognition for 25 other languages is also on the robot.