Split a song into four separate stems.

What is a music stem splitter?

A stem splitter separates a finished song into a small set of coordinated audio tracks. VocRemover creates four: vocals, drums, bass, and other instruments. The tracks share the same timing, so you can line them up in a digital audio workstation, practice with one part muted, or build a new mix.

The “other” stem contains the instruments that are not identified as vocals, drums, or bass. Depending on the arrangement, it can include guitar, piano, strings, synths, and other sounds. The model is designed for common music mixes, so unusual sound design and heavily processed recordings can leave some bleed between tracks.

Preview before you split the full song

Choose an audio file and listen to a free 15-second preview of all four tracks. The preview starts at the beginning of the song and does not require an account. It helps you check whether the separation fits your recording before spending credits on the full-length result.

When the preview sounds useful, choose Split full track. Full processing requires a signed-in account and credits. The completed tracks are saved to your private file list, where you can listen again and download the WAV files.

Ways to use four-track stems

• Practice drums, bass, or another instrument against the original arrangement.
• Make karaoke or rehearsal mixes by muting the vocal stem.
• Balance the rhythm section and harmonic instruments for a remix or edit.
• Study a song’s arrangement by listening to each stem on its own.
• Create backing tracks for lessons, covers, and music production sessions.

How to get a cleaner separation

• Use the best-quality source available. A lossless WAV or FLAC gives the model more detail than a low-bitrate file.
• Upload the original mixed song instead of a file that has already been split or processed.
• Listen to the preview at a moderate volume and compare the four parts; instruments can overlap naturally in a dense mix.
• Keep the stems aligned at the same start time when you import them into a DAW.

Know what to expect

AI source separation estimates which sounds belong in each stem; it does not recover the original studio master tracks. A vocal reverb tail may remain in the instrumental parts, and a distorted bass can overlap with the drums or other instruments. These artifacts are normal for a mixed recording and vary from song to song.

VocRemover uses the Demucs htdemucs_ft model for this four-stem workflow. Results are intended for preview, practice, remixing, and production workflows. Check the output and make sure you have permission to use the source recording before publishing or monetizing a derivative work.