Making an instrumental, sampling one drum loop, or practising a part without the original vocal used to mean phase tricks and disappointing results. LALAL.AI separates a finished mix into stems with a neural network instead. This review covers what the separation is good for, how the free tier is metered, and where the output still shows its seams.
What this tool is useful for
LALAL.AI takes an uploaded track and returns stems: vocals and instrumental at minimum, with drums, bass, piano, guitar, synth and other parts available as separate targets. Typical jobs are karaoke or backing tracks, remix and mashup work, isolating a sample to re-time it, pulling a vocal for a podcast or video bed, and giving students a version of a piece with one part muted. It accepts common audio containers, supports batch uploads on paid plans, and offers a desktop app that processes locally plus a VST plugin for use inside a DAW.
A practical workflow
1. Upload and choose stems
Start with the free tier and one short excerpt rather than a full album: the same three minutes that reveal artefacts will also reveal whether the model suits your genre. Select only the stems you need — voices and instrumental for karaoke, the full set when you plan to rebuild the mix.
2. Audition before you commit
Listen for bleed on held notes and cymbals, and for the hollow, watery texture that separation models leave when two sources overlap in the same frequency range. Compare a separated stem against the original at the same level, ideally on headphones.
3. Finish in your DAW
Separated stems are raw material, not a finished mix: gate, EQ and re-balance them in a DAW, or clean noise and level a voice track with tools such as Adobe Audition before publishing. If your source is a noisy live recording, expect to spend most of the time on repair rather than separation.
Plans and queue limits
Starter is free and always available: about 10 minutes of processing in the slower relaxed queue, with a 200 MB upload limit per file. Lite at roughly $7.50 per month (about $90 billed annually) adds a monthly fast-queue allowance of around 90 minutes, 2 GB uploads, batch processing, and one saved voice-pack slot. Pro at about $15 per month ($180 annually) raises the fast queue to roughly 250 minutes with unlimited relaxed processing, three voice-pack slots, and unlocks the VST plugin, local processing in the desktop app and API access. The free tier is genuinely useful for testing; heavy production work is metered.
Who should consider it?
DJs and remixers, karaoke and cover-track creators, video editors who need an instrumental bed, and music teachers preparing practice material. It also fits podcast work when a musical intro has to be removed or a guest’s track has to be isolated, though dedicated noise and speech tools handle that side better.
Limitations to check
Separation quality depends on the source: dense, heavily-compressed mixes with overlapping instruments separate less cleanly than sparse arrangements, and live recordings carry bleed that no model fully removes. The fast queue is metered in minutes, so plan uploads. You must own the rights to anything you upload — separating a commercial track for release is a licensing question, not just a technical one. There is no automatic transcription of a separated part into MIDI, and the free queue is slower by design.
How to compare alternatives
The closest everyday alternative is Moises, which bundles separation with practice tools such as tempo and key changes, chord detection and a mobile-first interface. For simple centre-channel vocal removal, Audacity does an acceptable rough job for free, though the result is a compromise rather than a clean instrumental. If you are generating music rather than taking it apart, Suno and Udio are the relevant reviews. For podcast-focused cleanup, compare the picks in our podcast editing software roundup.
Frequently asked questions
Is there a free version? Yes — the Starter tier processes roughly 10 minutes in a slower queue with a 200 MB per-file limit. Can it remove vocals completely? It gets close on studio recordings, but quiet artefacts remain where the voice shares frequencies with instruments. Does it work offline? Local processing is available in the desktop app on the Pro plan. Can I use the stems commercially? Only if you hold the rights to the source recording; the tool does not grant them.
Verdict
LALAL.AI is one of the cleanest ways to pull a usable instrumental or vocal stem out of a finished mix, and the free tier is enough to judge quality on your own material. Test one representative track first, keep a DAW in the loop for finishing, and pick a paid tier only when your monthly minutes justify it.
Official product information: https://www.lalal.ai/. Pricing, queue limits and plan features change; confirm current details before purchasing.
Unlock the Telegram group to access all tools on this site
A one-time payment of US$5 gives you access to the Telegram group and all tools on this site. No subscription.
Enable JavaScript to securely check payment and reveal the invite link.
Share this article
[…] hostovanjem, Podcastle uklanja premeštanje fajlova. Za razdvajanje vokala iz gotovih mixeva, LALAL.AI i Ultimate Vocal Remover rade ono što nijedan editor ne može. Audacity-eva niša je upravo ono […]