Karaoke online from any song

Upload a track and a neural network splits it into a backing track and vocals right in your browser. Ultra removes the voice most cleanly and takes from a minute to a few minutes (faster with a graphics card); Fast finishes in seconds but leaves more of the voice. While it runs, paste the lyrics — after “Place on the voice” each line lands where it is sung. Studio recordings separate best.

Yes, you can shift the key up or down by up to 12 semitones (one full octave). This lets you match the song to your vocal range without affecting the tempo.

No. The tool uses pitch compensation so that changing tempo does not alter the musical pitch. You can slow down or speed up the song while keeping the original key.

Yes. Enable your microphone and choose from four reverb presets — Room, Hall, Cathedral, or Echo. Your voice is mixed with the karaoke track in real time.

Yes. Press Record while singing and your performance records together with the karaoke track. The recording is saved as a WebM file (M4A in Safari on iPhone and Mac); the backing track itself downloads as MP3, WAV or FLAC.

LRC is a standard karaoke lyrics format with timestamps. Upload or paste an LRC file, and the lyrics will highlight line by line in sync with playback.

A speech model listens to the separated vocal and finds where each word of your text is sung, so lines and words light up at their own moment. The model downloads once (188 to 305 MB — the exact size is named before the download) and then runs on your device. Heavily processed choruses can land a line off; pick that line in the tap editor and tap it once where it starts. A backing track has no voice to follow, so there the lines are timed by tap or loaded from an .lrc file.

The A-B loop lets you select a section of the song to repeat endlessly. Set point A and point B on the waveform, and the playback will loop between them — perfect for practicing difficult passages.

Yes — MP3, WAV, FLAC, OGG, AAC, M4A, AIFF, OPUS and WebM all go in, and an MP4 video works too: the audio track is pulled out for you. What matters more than the container is what is inside it. A studio recording separates into voice and backing cleanly; a live recording where the crowd sings along, or a phone video shot in the room, gives a much rougher result, because the crowd is in the same frequencies as the singer and there is nothing to separate them by. If you have a choice between a video rip and the released audio of the same song, take the audio.

Choose the Backing track mode before uploading. The tool skips vocal removal and takes your instrumental straight into the player, where you can add lyrics, change key and tempo, sing with your microphone, and record.