Breaking01 / AI
Speech recognition model tags breaths in transcripts
Xing-Audio-NVVASR led on mean composite score in benchmarks with human annotations, according to results published by the researchers.

Researchers announced Xing-Audio-NVVASR on 20 Nov 2026 in a paper that described how the model transcribes speech and tags six types of non-verbal vocalisation.
The paper showed how the model inserts tags for sounds such as breaths directly into the transcribed sentences.
The paper reported that the researchers tested the model on four evaluation sets with human annotations.
In results published by the researchers, Xing-Audio-NVVASR ranked first with a mean composite score of 64.46. Its score was 6.03 points higher than that of StepAudio 3 Chat Preview.
The researchers gave no release date for the model weights or inference code.