Drop an audio file and the lyrics as plain text. One line per lyric line; [Verse]-style headers become sections.

Alignment runs on this machine and takes a few minutes for a full song.

Aligning…

Separating the vocal, aligning, then scoring. Leave this open.

Ready

Check the word timings

Two different alignment models will each time every word. Where they agree, the timing is right and you will never be shown it. Where they disagree, you decide by ear — you listen to two versions and pick the one that starts on the word.

Anything that could not possibly be right — a word lasting no time at all, a word starting where nobody is singing — is repaired without asking.

Takes about 15 seconds.

Listening through the track…

Re-timing every line with the second model.

Something is sung here that your lyrics file does not have. It was transcribed without being told the words, and what came back is a line you already wrote — so this is a repeat you left out, not a guess.

Adding writes the line into the project and every export. Your lyrics.txt is left alone — it is still missing this line. Inserting renumbers everything after it, so this one is not covered by ⌘Z.

Each button plays from the moment that version says the word begins. The right one starts on the word.

All checked