Show HN: An API for word timestamps from speech, 75% within 20ms
Summary
Show HN describes Quebec, an API and forced aligner for English speech that achieves 75% of word boundaries within 20 ms of human labels. It uses a fence-based decoder, a boundary model, and training data, and claims to outperform several baselines like MFA, ElevenLabs Cue, and WhisperX, with live docs and a pricing model.