Forced alignment for accurate SRT from known lyrics
Generated an SRT subtitle file from an audio track when the exact lyrics were already known
stable-ts has a model.align() function that takes audio + known text and only solves for timestamps — far more accurate than transcribe-then-match when the lyrics are romanized non-English (Hinglish) where Whisper would otherwise output Devanagari. Pass language=en for Latin-script lyrics, original_split=True to honor line breaks as SRT segment boundaries, and strip bracket annotations like [Intro]/[Chorus]/[bright guitar] from the source text since they are not sung.
When a user has both audio and a known transcript, reach for forced alignment (stable-ts model.align) immediately instead of running plain Whisper transcription and post-processing the words.