AI Stem Splitter, 4-Stem Separation
Extract vocals, drums, bass and other instruments from any song
Upload Your Track
Drag and drop your audio file here
or
Want your project in the cloud? See the plans
How to Split Any Song Into 4 Stems
What is a stem splitter? A stem splitter is an AI tool that separates a mixed song into individual instrument tracks called stems. Modern stem splitters use deep neural networks to isolate vocals, drums, bass and other instruments as separate audio files. Producers, remixers, DJs and vocal coaches use stems for sampling, mashups, karaoke, practice and vocal isolation. RemoveVocals extracts 4 high-quality stems for free, on its GPU servers, and a free account is asked only to download them.
RemoveVocals Stem Splitter uses a hybrid time-frequency neural network trained on a large dataset of multitrack recordings. Each 10-second window of the song is processed through two parallel branches: a waveform branch that learns directly from the raw audio, and a spectrogram branch that learns from the time-frequency representation. The two branches meet in a cross-attention bottleneck, then each decodes back into 4 stereo stems: vocals, drums, bass and other instruments.
By default the song is separated on our GPU servers and the stems come back to your browser; if the server cannot process the file, the Vocal Remover still gives you vocals and instrumental. After splitting, transpose any stem, fine-tune the EQ or run stems through AI Mastering.
Free to try, no watermark, free account only to download.
Drum remover: get the song without drums
Once the four stems are separated, mute the drums stem and use the Download Custom Mix (WAV) button: it renders only the stems left active, with the gains you hear in the preview, so you get the track without the drums for practice or for building a new beat under the original vocals, bass and instruments.
Bass remover: isolate or drop the bass
The bass comes out as its own WAV stem. Keep it alone to study or sample a bassline, or mute it before the custom mix download to get a bass-free version of the song and write your own low end.
Guitar and piano: 6 stems on Pro
The free tier separates 4 stems: vocals, drums, bass and other. The Pro plan adds guitar and piano as two more stems, so the "other" stem is split into more usable parts.
Key Takeaways
- 4 high-quality stems from any song vocals, drums, bass and other instruments, separated by a hybrid time-frequency neural network on our servers.
- Private by policy your file is processed for the separation only, deleted right after, never used to train models.
- Free to try no credit card, no watermark, free account only to download.
- All stems as WAV 16-bit PCM WAV ready to drop into any DAW (Ableton, Logic, FL Studio, Pro Tools, Reaper).
- Works on any genre pop, rock, hip-hop, R&B, electronic, metal, jazz, classical. Modern studio mixes give the cleanest stems.
- Use cases remixes, mashups, sampling, karaoke, vocal practice, cover versions, DJ edits, film/video sync.
Stem Splitter vs. Paid Alternatives
About the RemoveVocals Team
RemoveVocals is a suite of free, browser-based audio tools built by a small team of audio engineers, music producers and ML practitioners. We ship and maintain stem splitting, vocal removal lyrics transcription key detection BPM detection AI mastering and more, on our servers or in your browser, with no retention of your audio. Since 2024 we've served thousands of songwriters, producers, beatmakers and DJs every day. Learn more about us.
Frequently Asked Questions
What is a stem splitter?
A stem splitter separates a mixed audio track into individual components. RemoveVocals extracts 4 high-quality stems using a hybrid time-frequency neural network that runs on our GPU servers.
How many stems can I extract?
4 stems free: vocals, drums, bass and other instruments. The Pro plan adds guitar and piano, for 6 stems.
Is it free?
Free to try, no watermark, free account only to download. 4 stems free, 6 stems on Pro.
How does 4-stem separation work?
A hybrid time-frequency neural network processes 10-second windows of the song in parallel waveform and spectrogram branches that meet in a cross-attention bottleneck, then outputs 4 separate stereo stems in a single inference pass.
What is HPSS?
Harmonic-Percussive Source Separation uses median filtering on the spectrogram to separate transient percussive content (drums) from sustained harmonic content (instruments).
Is my audio uploaded to a server?
By default, yes: the song is sent to our separation server for the job and the result comes back to your browser. If the server cannot process the file, the Vocal Remover still gives you vocals and instrumental. We keep no copy of your files unless you are on a paid plan, where your downloads are auto-saved to your project library, and nothing is used to train models.
How is a stem splitter different from a vocal remover?
A vocal remover only produces two outputs: vocals and instrumental. A stem splitter breaks the track into more components: in our case 4 stems (vocals, drums, bass, other instruments). Use a vocal remover for karaoke; use a stem splitter for remixing, sampling, or isolating a specific instrument.
Does it work on all music genres?
Yes. The model is trained on a broad range of pop, rock, hip-hop, EDM, R&B, jazz, classical, metal, and acoustic material. Quality is best on modern, well-mixed productions and slightly lower on very dense mixes or heavily distorted guitars.
What audio formats are supported?
MP3, WAV, FLAC, OGG, M4A, AAC, and most common formats. Maximum length is 12 minutes per upload. Output stems are delivered as individual WAV files you can download one by one or all at once.
Can I use the stems commercially?
You can use stems from tracks you own or have the rights to (your own recordings, demos, royalty-free music). For copyrighted material, you are responsible for clearing sampling and derivative-work rights with the original rights holders, same rules as any DAW plugin.