KOMOS Achieves SRCC of 0.900 in Korean Speech MOS Prediction

·2026.11.02 12:50

Key point

The Korean speech quality assessment model KOMOS has been released as open source, recording a higher correlation than UTMOS.

The Korean speech quality assessment model KOMOS has been released as open source
Achieved higher correlation than UTMOS v1 and v2, recording an SRCC of 0.900 and LCC of 0.950 on 100 test utterances
Based on the Whisper-medium encoder and applies a Listener-conditioned prediction structure that reflects individual listener evaluation tendencies
Trained on 1,000 Korean utterances and 5,000 MOS scores, and can be used simply with Python

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.