A song you can hear is good. A song whose words roll across the screen in time with the singer is something you can sing along to, clip for a Reel, turn into a lyric video, or use to learn how a line in Tamil or Punjabi is actually pronounced. That second thing is what 'synced lyrics' means: every line of the lyrics carries the exact second it starts, so a player can highlight it as the singer reaches it.
Most AI music tools hand you an audio file and a block of text and leave the timing to you. On Autunes the timing is part of making the song. This guide explains how that works, what makes the timing accurate or sloppy, and how to use the result outside Autunes.
How Autunes times the lyrics
There are two ways to find out when a word is sung. The slow way is to listen to the finished song and recognise the words, which is what karaoke apps and caption tools do. The fast way is to ask the model that sang the song, because it already knows: while it generates the vocal, it keeps track of which part of your lyrics it is on. Autunes uses the second way first and the first way as a safety net.
During the render, the engine records where each lyric line starts. That timing comes from inside the model, so it does not care how busy the mix is or how unusual the language is. When the engine's own timing is not confident enough, typically on very long songs, the finished audio goes through a speech-recognition pass, and the words it hears are matched back to the lyrics you wrote. The matching works across scripts, so a Hindi song written in Devanagari still lines up even when the recogniser hears it in Roman letters.
Either way you do nothing extra. There is no button to press for a new song and no extra credit cost: the timing arrives with the track.
Where you see them, and how to get them out
Open any of your songs and scroll to the lyrics. If the song is timed, a Karaoke / Plain switch sits above the text. In Karaoke mode the current line is highlighted as the song plays, and the view scrolls with the singer. It follows the player wherever you started playback, so skipping to the chorus jumps the lyrics too.
To use the timing somewhere else, you need it as a file. The standard format is LRC: plain text where each line starts with a timestamp in minutes, seconds and hundredths, like [00:12.40] followed by the words. Most lyric-video editors, karaoke players and music apps that show lyrics can read it. On Autunes the LRC file comes with the Vocal-to-MIDI export (Pro and Creator), because a vocal melody in a DAW is far more useful when you know which word sits on which note.
- New songs with vocals: timed automatically, on every plan.
- Song page: Karaoke / Plain switch above the lyrics, highlight follows playback.
- Older songs without timing: 'Sync lyrics' on the song page (Pro and Creator).
- LRC file: downloaded alongside the Vocal-to-MIDI export (Pro and Creator).
- Instrumentals: no lyrics, so nothing to time.
How to get accurate timing
Synced lyrics can only be as good as the match between what you wrote and what was sung. Nearly every badly timed song we look at has the same cause: the written lyrics and the sung lyrics drifted apart. These habits keep them together.
- One sung phrase per line. A line should be what the singer delivers in one breath. Two phrases on one line become one long highlight; a phrase split across two lines flickers.
- Use section tags. [Verse], [Chorus], [Bridge] help the model sing the structure you wrote instead of inventing its own, which keeps the lines in order.
- Write in the song's own script. Devanagari for Hindi, Gurmukhi for Punjabi, Tamil script for Tamil. The singer reads it more reliably, so what it sings matches the text.
- Pick the vocal language instead of leaving it on auto when your lyrics mix languages. A wrong language guess changes the words that come out.
- Do not pad a line with filler you do not want sung. If the text says 'oh oh oh' four times and the singer sings it twice, the lines after it start early.
- Very fast rap is the hardest case for any aligner. Keep rap lines short and let each bar be its own line.
What people do with synced lyrics
Lyric videos are the obvious one: import the LRC into your editor and the text appears on the beat without hand-placing a single caption. The same file makes Reels and Shorts captions quick, which matters because a large share of short-form video is watched on mute.
Singers use karaoke mode to learn their own AI song before performing it at a function or recording a cover of it, and people learning a language use it to follow a song line by line in its real script. Families play wedding and birthday songs in karaoke mode at the event itself, so everyone can join the chorus.
When a song has no karaoke timing
If the switch is missing, check the obvious first: instrumentals and songs whose lyrics are empty have nothing to time. For a vocal song, the two usual reasons are a very long track where both timing passes came back unsure, or lyrics that were heavily rewritten by the singer, for example a chant repeated far more often than written. Regenerating with cleaner lyrics fixes most of these; on Pro and Creator, 'Sync lyrics' re-runs the listening pass on the song as it is.