Hy ASR 3.0 achieves low error rates in Mandarin, English, and Cantonese
November 13, 2026
Hy ASR 3.0 is an LLM-based speech recognition system that supports Mandarin, English, and 13 Chinese dialect groups. It uses external context like hotword lists and Q&A dialogues to improve accuracy and disambiguate similar-sounding keywords.
HOW THIS AFFECTS YOU
●
builderYou can improve transcription accuracy for multilingual apps by feeding task-specific context like hotword lists into the model.