Quebec - Accurate word timestamps API for speech
Show HN: An API for word timestamps from speech, 75% within 20ms
Quebec is a forced aligner API that returns precise word-level timestamps from audio and transcripts. Built by the team behind Yarn, it achieves 77% of word boundaries within 20ms of human labels—far surpassing WhisperX, ElevenLabs Cue, and MFA. This accuracy enables clean transcript-based video editing without clipping adjacent words. Quebec combines MFA's acoustic model with a custom decoder and a lightweight neural boundary model, trained on LLM-generated labels. It's fast, scalable, and available now with no signup—free until Monday 8pm ET, then $0.32 per hour of audio.
Quebec is extremely close to as good as hand-labelling, and you can try it at quebec.yarn.so.