← Guides

How to put timed lyrics on screen

Stripping the vocals is the easy half. This is the other one.

Lyrics on screen ≠ timed lyrics

Putting the words of a song on a screen is easy — any video editor does it. What actually works for singing is different: the right syllable highlighting at the right moment. Without that, anyone who doesn't know the song by heart can't follow.

The formats

By hand

Karaoke Builder Studio (Windows, buy once) is the classic for CDG. You play the song and mark as it goes. Budget about an hour per song once you have the hang of it.

Automatically

A speech recognition system listens to the isolated vocal and produces the timings. It's fast, and it fails predictably:

On a studio recording with clear diction, the result is usually good enough that you're only nudging details.

One thing that alarms people the first time they generate a file: long stretches with no lyrics at all are normal. Intros, solos and bridges are genuinely instrumental — compared against official synced lyrics, 25% to 40% of a song having no singing is common.

Frequently asked questions

What's the difference between LRC, CDG and UltraStar?

LRC syncs per line and is plain text. CDG is the graphical format used by karaoke machines, an mp3 plus cdg pair. UltraStar marks every syllable with time, duration and pitch — it's the only one that allows pitch scoring.

How do I sync lyrics to music automatically?

A speech recognition system listens to the isolated vocal and generates the timings. It works well on studio recordings with clear diction and gets a lot wrong on screamed, distorted or live vocals.

Is it normal for the lyrics to disappear during a song?

Yes. Intros, solos and bridges are instrumental. Compared against official synced lyrics, it's common for 25% to 40% of a song's duration to have no singing at all.

StarSinger does those three steps for you. You upload a file you already have, it separates the vocals, transcribes the lyrics with per-syllable timing and gives you a karaoke that plays in the browser, with pitch scoring. Free right now.

The caveats, because they're real: the transcription gets words wrong on screamed or heavily distorted vocals, and processing takes well over the length of the song.

Try it free