Awesome-Korean-Speech-Recognition

Why we included this project

Choosing a Korean speech-to-text API usually means trusting vendor demos. This index takes a different approach: it runs each service, including OpenAI Whisper, Google, Naver Clova, ETRI, and AWS, against the same public AI-Hub test sets and publishes the Character Error Rate for every dataset. The repo also explains why CER suits Korean better than word-level metrics, since the language is agglutinative and word boundaries are fuzzy, and it documents the test data and normalization steps so the results can be reproduced. That makes it a solid reference for teams picking a Korean STT provider and for researchers who want a baseline to measure their own models against.