LALAL.ai

LALAL.ai has one premise: you have a finished mix and you need the parts back. A vocal without the band. A drum track without the vocal. A clean dialogue line out of a video where music is playing underneath. If that describes a problem you actually have, this is a specialist tool that does the job properly. If you are looking for something that makes music, this is the opposite tool, and the distinction is worth settling before you read the pricing.

What it can pull apart

The separation list is granular. Vocals and instrumental are the obvious pair, but the engine also isolates drums, bass, acoustic guitar, electric guitar, piano, synthesizer, strings and wind instruments as individual stems. A Lead and Back Splitter separates a lead vocal from its backing vocals, which is a genuinely difficult problem and the kind of thing that decides whether a karaoke track or a remix is usable.

The engine behind this is called Andromeda and is described as a sixth-generation model built over six years of work on the problem. What matters practically is not the version number but that separation quality is the entire product here rather than one feature in a larger suite, and the roadmap has been pointed at it for a long time.

What it accepts

Audio input covers MP3, OGG, WAV, FLAC, AIFF, AAC and M4A. Video input covers AVI, MP4, MKV, MOV and M4V, which is more useful than it first appears: a podcast recorded over a music bed, an interview shot in a noisy room, or a lecture recording with a laptop fan in the background can all be fed in directly rather than exported to audio first.

The tools around the separation engine

Several adjacent features exist for cleaning rather than splitting. Voice Cleaner removes background noise, vocal plosives and microphone rumble. The Echo and Reverb Remover reduces room acoustics in a recording made somewhere unsuitable. Voice Changer alters accent and tonality across audio and video. Voice Cloner builds a personalised AI voice model. Together these turn the product from a single-purpose separator into a small audio repair kit, which is the difference between using it once and keeping the subscription.

Where it runs

There is a browser interface, desktop applications for Windows, macOS and Linux, mobile apps for iOS and Android, a VST plugin for working inside a DAW and an API for developers. The VST plugin and API are Pro-plan features. For a producer, the plugin is the difference between a workflow and a chore, because separation happens in the session rather than in a browser tab with files moving back and forth.

How the pricing actually works, checked August 2026

The structure is unusual and worth understanding properly, because it is easy to misread. Processing happens in one of two queues. The Relaxed Queue is standard priority and is unlimited on every paid plan. The Fast Queue is priority processing and is the metered resource. Minutes are deducted according to the duration of the file and the number of stem types selected, and Fast Queue minutes do not roll over.

Starter is free and always free, giving 10 minutes in the Relaxed Queue with a 200 MB file limit, and no downloads, batch processing, VST plugin or API access. Lite is 7.50 dollars per month or 90 dollars per year, with unlimited Relaxed Queue processing, 90 Fast Queue minutes per month, a 2 GB file limit, result downloads and batch processing, but no VST plugin or API. Pro is 15 dollars per month or 180 dollars per year, with unlimited Relaxed Queue processing, 250 Fast Queue minutes per month, a 2 GB file limit, downloads, batch processing, the VST plugin, API access and early access to new features. One-time top-ups add Fast Queue minutes without a subscription: Master at 750 minutes for 50 dollars, Premium at 3,000 minutes for 190 dollars and Enterprise at 5,000 minutes for 300 dollars.

The practical reading is that on any paid plan you are never actually blocked, only slowed. Fast Queue minutes buy impatience. That makes Lite a reasonable plan for a great deal more work than its 90 minutes suggests, as long as you are willing to wait during busy periods.

Who this fits

Producers and remixers are the obvious audience, along with DJs building edits, musicians learning parts by ear, and karaoke and cover-song creators. Less obviously, it fits video and podcast editors who inherit recordings with problems baked in, and content teams who need to reuse footage where the licensed music has to come out. The free tier will tell you within ten minutes whether the quality holds up on your specific material, which is the only test worth running, since separation quality varies with the density of the mix rather than with any advertised number.

It is the wrong purchase if you need music rather than the parts of music. A separator does not compose. Match the tool to whether your audio already exists, and this becomes a straightforward decision at a low monthly cost.

The case for LALAL.ai

LALAL.ai is a specialist, and that is the argument for it. Separation quality is the product rather than a feature inside a larger suite, and the stem list reflects it: vocals, instrumental, drums, bass, acoustic and electric guitar, piano, synthesizer, strings, wind instruments, and lead vocals split from backing vocals. It takes video files as readily as audio, so a dialogue track can be recovered from a recording with music underneath it, and the surrounding repair tools cover noise, plosives, mic rumble, echo and reverb.

Reading the plans correctly

The pricing is easy to misjudge. Every paid plan includes unlimited processing in the Relaxed Queue; the metered Fast Queue minutes buy priority, not access. Lite at 7.50 dollars a month, or 90 a year, adds downloads, batch processing, a 2 GB file limit and 90 Fast Queue minutes a month. Pro at 15 dollars a month, or 180 a year, raises that to 250 Fast Queue minutes and adds the VST plugin, API access and early features. The free Starter tier gives 10 minutes of Relaxed Queue processing with no downloads, which is enough to hear the quality on your own file first. One-time top-ups cover 750 minutes for 50 dollars, 3,000 for 190 and 5,000 for 300. Minutes are consumed by file duration multiplied by the number of stems selected, so budget in stems. All figures checked August 2026.

Who should look at something else

If you need original music rather than the parts of an existing track, Soundraw generates royalty-free tracks with a commercial licence and is the correct purchase. If you need narration or a synthetic voice, a speech platform such as ElevenLabs handles that directly. And if your real job is turning long recordings into short social clips, Opus Clip does the video side of that work, with LALAL.ai useful only for fixing the audio you feed it.

For anyone who regularly has to get inside a finished mix, though, this is a focused tool at a modest price, and the free tier answers the quality question before any money changes hands.