LALAL.AI

The Andromeda model is LALAL.AI’s sixth generation AI, built on transformer architecture and trained with four times more data than its predecessor Perseus. It delivers up to 40% faster processing and a 10% improvement in Signal to Distortion Ratio (SDR). For producers and engineers, this means cleaner isolated stems with fewer artefacts and virtually no cross bleed between vocal and instrumental channels.

Unlike many competitors that cap out at four or five stems, LALAL.AI splits audio into 10 distinct stem types. These include vocals, instrumentals, drums, bass, piano, electric guitar, acoustic guitar, synthesiser, voice and noise, and lead versus back vocals. This gives remixers and sample creators granular control over every element in a mix.

Beyond music separation, LALAL.AI provides dedicated tools for spoken word content. The Voice Cleaner strips out background music, mic rumble and ambient noise from recordings. The Echo and Reverb Remover tackles room reflections and unwanted reverb tails. Both tools are essential for podcasters, filmmakers and anyone cleaning up dialogue tracks.

LALAL.AI now offers a VST plugin that brings stem separation directly into your digital audio workstation. Available on the Pro plan, the plugin allows producers to split vocals and instrumentals without leaving their session. This eliminates the round trip of exporting, uploading and re importing files.

The platform accepts 12 audio and video formats including MP3, WAV, FLAC, AIFF, AAC, AVI, MP4 and MKV. Batch upload lets you queue multiple files at once, making it practical for professionals working through large libraries or entire album sessions in one go.

LALAL.AI runs on web browsers, Windows, macOS, iOS and Android. This means you can start a separation on your laptop and check results on your phone. The desktop app also supports drag and drop for quick local file processing.