fa-benchmark

Here is 1 public repository matching this topic...

olewave / fa-bench

Phone and word timestamp accuracy on human-labeled TIMIT and Buckeye, clean and under reverb, noise, music and babble. Two tracks, forced aligners given the transcript and ASR systems that find their own words. MFA, WhisperX, Charsiu, Qwen3, Parakeet-TDT, Olign and more. Monthly leaderboards, paper on arXiv.

benchmark timestamp stt asr forced-alignment boundary asr-benchmark asr-model stt-benchmark asr-services fa-benchmark

Updated Oct 12, 2026
Python

Add this topic to your repo

To associate your repository with the fa-benchmark topic, visit your repo's landing page and select "manage topics."

Learn more