Vocal Remover

Yes. No account, no watermark and no charge for the first run. To keep it free and stop the cloud bill from being abused, every connection gets a limited number of runs per day: a guest gets one run of up to 30 seconds, a free account gets five credits a day, and Pro subscribers get 1500 credits a month with clips up to 10 minutes long. A run costs 5 credits per 30 seconds of audio, so a 30 second clip costs 5 credits and a full 10 minute track costs 100. Signing in raises both the daily allowance and the length limit. There is nothing to install, no plugin to download and no subscription required just to try it on a short clip. If the separation fails, nothing is deducted, because credits are only taken once a result is actually produced.

Your track is sent to Volcengine AI MediaKit, which runs a trained source separation model over the audio. The model looks at the whole mix and estimates two signals: the parts that sound like a human voice, and everything else, which includes drums, bass, guitars, synths and reverb tails. It is not a simple filter or a phase cancellation trick, so it also works on tracks that were never released as instrumentals and on mono recordings where the old centre channel trick fails completely. Because separation is a prediction rather than an exact subtraction, the result is never mathematically perfect: you may hear small artefacts on dense mixes, and heavily processed or distorted vocals are the hardest case. The output is two independent MP3 files you can play and download separately. For best results use a clean mix at a decent bitrate rather than a low quality phone recording.

You can upload MP3, WAV, M4A, AAC and FLAC audio, or a video file in MP4, MOV, MKV or WebM, in which case the audio track is extracted and separated. The upload limit is 30 MB. The length limit depends on your plan: guests and free accounts can process up to 30 seconds in a single run, while Pro allows one clip of up to 10 minutes. Processing time scales with the length of the audio, roughly 2.3 times the clip duration in our tests, so a 30 second clip usually finishes in about a minute and a 10 minute track can take several minutes. You can separate a long song in several passes, or trim it first with our Audio Trimmer. Nothing is charged when a run fails, so a rejected file never costs you credits.

Yes, this tool is different from our browser based tools. Your audio is uploaded to our server, forwarded to Volcengine AI MediaKit for processing, and the resulting files are stored on our server so you can download them. They are deleted automatically after a set time: 30 minutes for guests, 2 hours for free accounts and 24 hours for Pro. We do not listen to your files, build a library from them or use them to train anything, and they are never published or shared. If your audio is confidential, keep in mind that it does leave your device. For work that must stay local, use our browser based tools such as Audio Trimmer, which run entirely on your own machine and never upload anything. Trim or mute anything sensitive before uploading.

Technically yes, and that is what most people do with it: the instrumental track is a backing track for practice or karaoke, and the acapella is useful for remixes, sampling or transcription. Legally, the copyright in the original recording and in the composition still belongs to its owners, so what you may do with a separated track depends on the licence you have and on the rules of the platform you want to publish on. Separating a track does not create new rights and does not make a commercial release lawful by itself. For personal practice, study or a private project this is normally not an issue. If you plan to publish or monetise the result, treat it exactly as you would treat the original recording. When in doubt, use music you own, licensed music, or a track released under a permissive licence.