How to sync lyrics to music

Start with auto-sync every time, even when you intend to hand-time the whole thing. Auto-sync isolates the vocal from the mix and places every line you pasted against what it hears, so the words are already in roughly the right place and in the right order. Hand-timing from zero means placing every line twice: once to get it near, once to get it right.

The exception is a track auto-sync struggles with, and the pattern is predictable. Heavily processed or layered vocals give it less to isolate; long instrumental intros throw off the range if you have not marked where the singing starts; spoken-word and rap sections with dense syllables tend to land early. Roughly one song in five needs real manual work. When you hit one, use Tap to sync: play the track and tap once per line as you hear it, which produces better timings in one pass than dragging ever will.

When timing is wrong, it is wrong in one of two ways, and they have completely different fixes. Telling them apart takes ten seconds: check a line near the start and a line near the end.

If both are out by about the same amount, you have a constant offset. Every line is early or late by the same margin, usually because the audio has silence at the head that the analysis counted, or because you set the sung range a beat off. The fix is one action: select every lyric line and drag them together. Relative timing is untouched.

If the first line is close and the last line is badly out, the error is accumulating, and dragging everything will fix one end while breaking the other. This means the spacing itself is wrong — commonly a trimmed or extended intro, a tempo change mid-song, or a version of the track that differs from the one the lyrics were timed against. Do not chase it line by line from the start. Re-run auto-sync with the sung range set correctly, or split the song at the section that changes and sync each half against its own range.

One more effect that reads as bad timing but is not: text needs a moment to be read before it can feel simultaneous. Lines that are technically perfect feel a fraction late. Pull them 100–200 ms ahead of the vocal and they will land right, which is why finished lyric videos almost always sit slightly early on paper.

LRC is the plain-text format karaoke tools and music players use for timed lyrics, and it is worth understanding because it is the interchange format between everything. It is readable enough to fix in any text editor. Each line begins with a timestamp in square brackets, in minutes, seconds, and hundredths, followed by the lyric.

That is standard, line-level LRC: it tells a player when each line begins, and the line stays up until the next timestamp. It is all you need for classic lyric videos where whole lines appear together.

For karaoke highlighting, where each word lights up as it is sung, you need enhanced LRC (sometimes called the A2 extension). It keeps the line timestamp and adds an inline timestamp in angle brackets before each word. Motion Text reads both forms, and a tag covering several words simply times them as a group.

Two practical notes. Motion Text ignores LRC metadata tags — [ti:], [ar:], [al:] and, importantly, [offset:] — and reads only the timestamps themselves, so a file that relies on an offset tag needs it applied to the timestamps first or corrected with one drag after import. And because LRC marks only where lines start, the final line has no following timestamp to end it; it is held for three seconds on import, which is usually the one cue worth trimming by hand.

Timings you make in Motion Text are not trapped in Motion Text, and timings you made elsewhere are not trapped either. Anything that exports line-level or enhanced LRC — desktop karaoke tools, the timing editors built into music players, the lyric sync features in streaming apps that allow export — imports directly. So do SRT and VTT subtitle files, which is the usual route if your lyrics started life as captions on a video.

Going the other way, the practical export is the video itself: a rendered MP4 with the words burned in, which is what every platform wants anyway. If you need the timings as data for a karaoke player or another editor, hand-editing an LRC file against the timings you can read off the timeline is a five-minute job for a three-minute song, and the format above is all you need to know to write one.