Agora launches speech-to-text for 60+ languages
Agora has announced a multilingual speech-to-text solution that provides native-level accuracy across more than 60 languages. This low-latency Automatic Speech Recognition (ASR) technology aims to reduce 'hallucinations' and is designed for real-world applications beyond English-first models.
Key Takeaways
- Agora says its ASR reaches native-level accuracy in more than 60 languages.
- The speech-to-text system is built for low-latency use, a key requirement for live video applications.
- Agora says the model is designed to produce fewer hallucinations than English-first ASR systems.
Why It Matters
Agora is targeting one of the biggest friction points in global video workflows: speech-to-text that works reliably outside English and does it with low latency. The product positions Agora in the multilingual ASR stack for real-world applications, where hallucinations and delay can break captions, transcription, and moderation. The clearest signal to watch next is whether Agora expands details on language coverage or deployment specifics beyond the current 60+ language claim.
Read full article at prod.agora.io