karanow vs Moises, KaraFun and Youka

Three kinds of tool

File-based separators take a song you already have as a file, or a link, spend as long as they need on it, and hand back stems — vocals, instrumental, sometimes drums and bass on their own. Moises, LALAL.AI and SongSplit AI are this kind. Youka is a variant that produces a karaoke video rather than a stem.

Catalogue karaoke services license songs one at a time and let you sing the ones they have. KaraFun is this kind. The song you want is either in the catalogue or it is not.

Live tools work on the sound as it plays. Apple Music Sing is this kind, inside the Apple Music app on iPhone, iPad and Apple TV. karanow is this kind, on the Mac, for any application.

Moises

Moises separates uploaded files. Premium is US$3.99 a month or US$35.88 a year. You upload a track, wait, and get stems you can keep, along with tools built around them. It does not work live on Spotify or YouTube: the song has to be a file you give it. If you want a clean instrumental to keep, load into a DAW or hand to a band, this is the shape of tool you want, and karanow is not.

KaraFun

KaraFun is a subscription to a licensed catalogue of about 53,000 songs, produced as karaoke versions. The upside is that each track was made to be sung to. The limit is the catalogue: it is large, and it still will not have the B-side, the live version or the song by the band with four listeners, because every song in it had to be licensed individually.

Youka

Youka makes karaoke videos. It is credit-based, from US$9.99 a month for roughly 20 karaoke tracks, and takes either a file or a YouTube URL. The result is a video you export and keep. It is the right tool when the point is to produce something — a video for a party, a file for a screen — rather than to sing right now.

Apple Music Sing, LALAL.AI, SongSplit AI

Apple Music Sing is included with Apple Music and turns the vocal down live on a supported song, but it is on iPhone, iPad and Apple TV, not on the Mac. LALAL.AI and SongSplit AI are file-based separators: upload, wait, download stems. Like Moises they give you a file and do not work on what is playing.

Where karanow sits

karanow works live, on the Mac, on the output stream — the sound already on its way to your speakers. You play a song in Apple Music, Spotify, a browser tab or a video player, press one button, and about two seconds later the lead vocal is gone while the band keeps going. The original singer can be blended back with a slider. There is no upload, no download and no catalogue: whatever your Mac can play, you can sing to.

It produces no file. The backing track exists while it is playing and nowhere else; nothing is exported or saved. And because it has about a second on a moving window rather than minutes on a whole track, its separation is not as clean as the offline tools above — traces of the vocal remain on dense mixes. That is the trade: they give you quality and a file; karanow gives you nothing to prepare and nothing to wait for.

With Apple Music and Spotify it also draws the notes to aim at, read from the original singing on the recording you are playing, with your own voice drawn over them from the microphone. That part is marked beta in the app. It runs on Apple Silicon Macs with macOS 14.2 or later, processes audio in memory on your machine, and never uploads it.

Which one to pick

You want a stem to keep: Moises, LALAL.AI or SongSplit AI. You want a karaoke video to share: Youka. You want a big catalogue of songs made for singing and do not mind the song you want sometimes being missing: KaraFun. You have an iPhone, iPad or Apple TV and Apple Music: Apple Music Sing is already there. You are on a Mac, the song is already playing, and you want to sing to it now: karanow.

Wanting two of these is normal — a file for later and something to practise against now — and using two tools is the honest answer, not a failure of either.