Glossary
Forced alignment
Automatically matching each word or phoneme in a transcript to its exact timestamp in the audio, producing the aligned text-and-speech pairs that ASR and TTS training depend on.
An aligner takes audio plus its known transcript and finds the boundaries of each word and phoneme; open tools such as the Montreal Forced Aligner are standard. The output is a timestamp per unit. Alignment fails where transcript and audio disagree, so it doubles as a quality check on transcription, and heavy crosstalk or noise degrades it.
Why it matters
Alignment is the step that turns an archive into training-ready audio–text pairs for ASR and TTS.