At a glance
- Accepts
- MP3, WAV, M4A, AAC, OGG, Opus, FLAC, and the audio track of a video, up to 20 minutes
- Outputs
- An editable list of timed lyric lines, downloaded as a .lrc file (with title, artist and length tags) or copied as text
- Where it runs
- Entirely in your browser, on your device
- What leaves your device
- Nothing - the file never leaves your device
- Limits
- One timestamp per line, not per word; singing is harder than speech for Whisper, so fast rap, heavy distortion and layered vocals get more words wrong and every line should be checked; the language must be picked, not detected; downloads a vocal model (38 MB) and a speech model once
- Account
- Not required - there is no sign-up
- Price
- Free, with no watermark and no usage cap
- Made by
- Visrut
- Last updated
- 23 October 2026
Turn a song into a synced lyrics file. The tool takes the singer out of the music, writes down each line with the moment it is sung, and gives you an .lrc file for your music player or karaoke app. Every line stays editable, so you can fix a misheard word or nudge the timing while the song plays. It all runs in your browser, so the song never leaves your device.
How to use it
- Drop in a song and pick the language it is sung in.
- Press Write the lyrics. The first time, the AI models download and are saved for next time.
- Play the song and follow the highlighted line. Fix any words, use Earlier or Later if the timing is off, then download the .lrc file.
How it works
First, Spleeter, an open-source model from Deezer, separates the vocals from the instruments, the same step Remove Vocals uses. Then OpenAI's open-source Whisper model listens to the vocals alone and writes down the words, with the time each phrase starts. Finally, the tool removes what Whisper is known to add on songs, like music symbols, "[Music]", and phrases such as "Thanks for watching" over instrumental parts, and writes the lines out in the LRC format. The same speech model powers Transcribe Audio.
Making a karaoke track?
Pair the .lrc file with an instrumental from Remove Vocals, and if the song sits too high or low for you, change its key with Change Song Key & Tempo (the lyrics stay in time as long as the speed is unchanged).
Frequently asked questions
What is an LRC file?
An LRC file holds a song's lyrics with a timestamp in front of each line, like [01:02.46]Hello from the other side. Music players, karaoke apps and car stereos read it to show each line as it is sung. It is plain text, so you can open and edit it in any text editor. Name it the same as the song (Song.mp3 and Song.lrc) and most players pick it up automatically.
How accurate are the lyrics?
Good on clearly sung pop, rock and ballads, and weaker on fast rap, heavy distortion, mumbled or layered vocals, and words the model hasn't heard before (names, slang). Singing is harder for speech recognition than talking, which is why every line stays editable: listen through, fix the words that are wrong, and download. The Most accurate option uses the large Whisper model and gets noticeably more words right on singing, if your computer can run it.
Why does it take the vocals out first?
Speech recognition models were trained on speech, not songs. With drums and guitars underneath, they miss words and sometimes invent phrases over the instrumental parts. Separating the singer from the music first, with the same model as Remove Vocals, gives Whisper the voice on its own and makes a clear difference. For an a cappella recording or a spoken track, untick it to save time.
The lyrics are a bit early or late. How do I fix it?
Press play and use Earlier or Later until the highlighted line matches the singing; that shifts every line at once. To fix a single line, play the song and press the timer button on that line at the moment it starts, or type the time in. The timestamps are written into the file, so they work in every player.
Which languages work?
Thirty of the most widely sung languages, including English, Spanish, Portuguese, French, German, Italian, Hindi, Japanese, Korean and Chinese. Pick the language the song is sung in; the model doesn't detect it on its own, and a wrong choice gives poor or translated lyrics.
Is my song uploaded?
No. The vocals are separated and the words written down inside your browser tab. The only downloads are the AI models themselves, the first time, and they are saved for next time.