LALAL.AI is an audio separation service that splits songs into vocals, instrumentals and individual instrument stems, and cleans noise, echo and reverb out of voice recordings. It has a free tier for previews and two paid plans, billed monthly or yearly and metered in processing minutes. What decides whether it suits you is how many minutes you process each month and whether you need downloads, the plugin or the API, which come with paid plans.
What is LALAL.AI?
LALAL.AI is operated by OmniSale GmbH, based in Risch-Rotkreuz, Switzerland. It runs in the browser, with desktop apps for Windows, macOS and Linux, mobile apps for iPhone, iPad and Android phones and tablets, and a plugin for music software. The site comes in English, Chinese (simplified and traditional), French, German, Italian, Japanese, Korean, Spanish, Dutch, Portuguese and Brazilian Portuguese.
The separation runs on LALAL.AI's own neural networks: Rocknet, Cassiopeia, Phoenix, Orion, Perseus, Andromeda and Lyra. LALAL.AI is a proprietary service; its GitHub repository holds API client examples, not the models.
What LALAL.AI does to your audio
The Stem Splitter takes a song and separates it into vocals, an instrumental and individual instrument stems. Each file is processed on its own and split into its parts, and results export as MP3, WAV, FLAC, OGG or AAC.
Voice Cleaner removes background noise, background music, mic rumble and plosives from speech while keeping the voice intact. Echo and Reverb Remover is a separate tool for cleaning room sound out of a recording.
Video works too. You can upload AVI, MP4, MKV, MOV or M4V files, and LALAL.AI returns the cleaned file as video. On paid plans, batch processing lets you select and upload up to 20 audio or video files at a time in the desktop app.
Processing normally runs on LALAL.AI's servers. On the Pro plan, the desktop app adds local processing with Lyra, the on-device model, and works offline apart from occasional subscription checks.
LALAL.AI is built for separating and cleaning, not for producing a finished podcast. It does not set loudness targets, cut filler words or remove silences, and it does not mix several speaker tracks into one. It works on uploaded audio and video files, not live calls.
Working with other software
The LALAL.AI plugin runs locally inside music software in VST3 format on Windows, macOS and Linux, and as an Audio Unit plugin, in beta, on macOS. It works in any DAW that supports VST3 and has been used in Ableton Live, Reaper and VST3-enabled versions of Audacity. AAX is not supported, and the plugin is on the Pro plan.
A public API, with documentation and an OpenAPI spec, lets developers build stem splitting and noise reduction into their own software or service. API access is included on Pro. Results are downloaded; there is no publishing to podcast hosts, YouTube or cloud storage.
LALAL.AI pricing: what each plan gives
The free Starter plan is permanent and gives 10 minutes in the Relaxed Queue, where wait times vary, with a 200MB upload limit per file. Starter shows free previews of results but does not include downloads, batch processing, the Fast Queue, the plugin or the API. On iOS, free users can batch up to 5 files. There is no time-limited trial of a paid plan.
Both paid plans include unlimited minutes in the Relaxed Queue plus a monthly allowance of Fast Queue minutes, which reset each month if unused. Lite costs $9.99 a month, or $7.50 a month billed yearly ($90 a year), and gives 90 Fast Queue minutes a month. Pro costs $19.99 a month, or $15 a month billed yearly, and gives 250 Fast Queue minutes a month, along with the plugin, the API and local processing in the desktop app.
A limited first-year offer cuts Pro's yearly price to $10.50 a month, renewing at the standard rate afterward. One-off Top-Up packs add Fast Queue minutes without changing plan: Master gives 750 minutes for $50, Premium 3,000 minutes for $190 and Enterprise 5,000 minutes for $300.
Privacy and security
LALAL.AI's privacy policy is direct on training: the company "does not use user files for artificial intelligence training", and it does not share uploaded media files with third parties. OmniSale GmbH is the data controller under the GDPR. Support runs by email at support@lalal.ai, with a help page alongside.
Who LALAL.AI suits
LALAL.AI suits musicians, producers, DJs and video editors who need clean vocal or instrument stems from a finished mix, or who need noise and echo taken out of speech in audio or video files. Pro fits people who work inside a DAW or want to process files offline, and developers who want to build separation into their own product through the API.
It suits less well anyone hoping to work for free, since Starter shows previews of results without downloads. Podcasters looking for loudness targets, filler word cutting or multitrack mixdown will not find them here, and the plugin, the API and offline processing all need Pro rather than Lite.
Frequently Asked Questions
How much does LALAL.AI cost?
Starter is free for 10 Relaxed Queue minutes, with previews of results but no downloads. Lite costs $9.99 a month, or $7.50 a month billed yearly, for 90 Fast Queue minutes a month. Pro costs $19.99 a month, or $15 a month billed yearly, for 250 Fast Queue minutes a month. Top-Up packs start at $50 for 750 minutes.
Can I download results on the free plan?
No. The free Starter plan shows previews of results, but downloads come with the paid Lite and Pro plans.
Does LALAL.AI work as a plugin in my DAW?
Yes, on the Pro plan. The plugin runs locally in VST3 format on Windows, macOS and Linux, with an Audio Unit version in beta on macOS. AAX is not supported.
Does LALAL.AI train AI on my files?
No. LALAL.AI does not use user files for AI training and does not share uploaded media files with third parties.
