Loading page
About this tool
Split audio you own into vocals, bass, drums and instrumental stems.
The Vocal Remover & Stem Splitter runs FFmpeg in your browser to pull four approximate stems out of a stereo audio file: an instrumental made by centre-channel cancellation (left minus right), a vocal pass band-limited to 120 Hz–5 kHz and downmixed to mono, a bass stem low-passed at 250 Hz, and a drum stem high-passed at 250 Hz with 4:1 compression above -18 dB. It is for musicians, karaoke makers and video editors who need a usable backing track or a rough isolated part in minutes. These are deterministic filter approximations rather than neural source separation, so expect some leakage and check the result before you use it.
Open Vocal Remover & Stem Splitter on AltFTool — it loads instantly in your browser.
Upload or drag in the file you want to work with.
Adjust the settings and preview the result.
Download the finished file to your device.
Instrumental, vocals, bass and drums are separate profiles you can render one at a time from the same source file, each written out as WAV.
Every profile is a fixed FFmpeg filter chain with published crossover points, so the same input always gives the same output — no model randomness to re-roll.
The tool states plainly that this is frequency and centre-channel approximation, not AI source separation, so you know to listen for leakage rather than trust it blind.
No — it removes centre-panned content, which is usually where the lead vocal sits, using a left-minus-right phase cancellation. Anything else panned to the centre (often kick, snare and bass) is removed with it, and reverb tails or double-tracked vocals panned wide will survive.
The vocal profile keeps 120 Hz to 5 kHz and folds to mono, bass keeps everything below 250 Hz, and drums keep everything above 250 Hz with a 4:1 compressor engaged at -18 dB. The instrumental profile is not a filter at all — it is the L−R / R−L centre cancellation.
No. Those use trained neural models that can separate overlapping instruments in the same frequency band; this uses fixed FFmpeg filters and channel maths, so a bass guitar and a kick drum sharing 80 Hz will land in the same stem. Use it for a fast usable result, not for a clean multitrack.
No — FFmpeg is loaded as WebAssembly and the file is processed inside the page, then handed back to you as a WAV download. Only use it on audio you own or have the right to process, since stem-splitting a commercial release can still be a copyright issue regardless of where the processing happens.