GuideMost advice about transcription accuracy is generic. This guide ranks the five levers by measured effect: microphone distance, overlapping speakers, noise and silence, language and vocabulary, and the model variant — with the published numbers that show how much each one moves your error rate.
AIA reference page of speech-to-text statistics that all trace to a primary source: benchmark WERs by audio condition, model-by-model error rates, the human baseline, bias and hallucination findings, and dictation speed — plus the widely quoted numbers we could not verify and left out.
AIWord error rate counts substitutions, deletions, and insertions against the words actually spoken. What the formula includes, what it quietly ignores, why the same model scores differently in two papers, and how to measure your own WER in ten minutes.
AIApple's on-device SpeechAnalyzer beat Whisper in a widely shared benchmark — but that test ran Whisper Small on English audiobooks. Here is what it measured, what it did not, and which engine fits which recording.
AIWhat Whisper accuracy actually looks like: roughly 2–3% word error rate on clean read speech, about 7% across mixed English benchmarks, around 16% in meeting rooms — and what moves those numbers most.
