AI Stem Splitter

AI Stem Splitter is a web-based stem separation and vocal removal tool for karaoke tracks, remix work, and DJ use. It splits songs into isolated parts and can also extract clean acapellas, remove drums or guitar, and create instrumental backing tracks.

The service can separate a song into up to six stems: vocals, drums, bass, guitar, piano, and other. It also offers a YouTube and SoundCloud vocal remover, a karaoke maker, an AI acapella extractor, a guitar remover, an AI drum remover, and a voice isolator for spoken voice in noisy recordings, interviews, calls, and field audio. Additional free tools include a key finder, a nightcore maker, a pitch changer, a slowed reverb maker, and a TikTok voice generator. The site says tracks are separated in under 60 seconds, and that a 4-stem mode has better audio quality while 6-stem output is available as well. It also provides automatic tempo and musical key detection, DJ mode with beat grids and Camelot notation, Rekordbox XML export, and browser-based stem preview with mute, solo, and scrub controls.

Supported input formats include MP3, WAV, FLAC, M4A, AAC, OGG, and WEBM, with a maximum file size of 50 MB. Users can also paste YouTube or SoundCloud links instead of uploading files. The system is powered by htdemucs, described as a Meta AI model, and the page notes that core stems come from htdemucs_ft while guitar and piano come from the 6-stem model htdemucs_6s. The service is available through a web interface, a REST API with OpenAPI 3.1 support, and seven first-party SDKs for Node, Python, Java, Go, PHP, Swift, and Lua.

Pricing is pay-as-you-go with credits that never expire. A free start is available with 3 free minutes of processing and no credit card required, and a Pro plan is listed at $9 per month.