One upload, both tracks — the clean acapella and the karaoke-ready instrumental. Vocal isolation and vocal removal in a single pass. Preview 30 seconds free, no signup.
Have several recordings? Choose them all at once, then sign in to process them in one go.
A real split of this song by the same AI model you use. Mute, solo, or drag through it.
Preview the first 30 seconds free · Sign in to download 1 minute free
Vocal Isolator is an AI stem separator for songs and videos: one upload returns two sample-aligned tracks, the vocals as an acapella and the music as an instrumental. A BS-RoFormer source-separation model makes the split, not phase cancellation, so bass and kick stay in the music instead of cancelling out. Both tracks download as 44.1 kHz, 16-bit WAV.
Upload the song, let the AI split it into vocals and music, then download the instrumental to remove vocals or the acapella to keep them. It works on any finished song, with no software to install.
Add your song. Upload your song as an MP3, WAV, FLAC, or other audio file (or a video) up to 2 GB.
AI splits it online. Vocal Isolator separates vocals from music with a BS-RoFormer model on cloud GPUs, so there is nothing to install.
Preview both tracks. Solo the vocals or the music and hear the first 30 seconds free.
Download what you need. Save the acapella, the instrumental, or both as lossless WAV or as MP3.
Vocal Isolator uses AI source separation, not the old phase-cancellation trick. A BS-RoFormer model trained on multitrack recordings reads your mix in short overlapping windows and estimates, for every slice of frequency and time, how much of the sound is voice. It is the same family of models that professional stem-splitting tools use. Those estimates are how it separates vocals from the instrumental: the voice goes to the vocal track and everything else to the music track.
The two tracks always add back up to your original song. Because the model recognizes what a voice sounds like instead of where it sits in the stereo field, it works on wide, double-tracked, and off-center vocals, and it keeps the bass and kick drum that phase cancellation throws away.
| Method | How it works | Typical result | Use it when |
|---|---|---|---|
| Phase cancellation (center channel removal) | Subtracts the left channel from the right | Removes only centered vocals; bass and kick disappear too, so the instrumental sounds hollow | You only have a basic audio editor |
| AI source separation (Vocal Isolator) | A trained model separates voice from music in every frequency band | Clean vocal and instrumental tracks; faint traces on heavy reverb or live recordings | You want both stems from any finished song |
| Official stems or instrumental | The artist or label releases the separate tracks | Perfect separation | The song has an official instrumental or stem release |
No hollow "phase-cancelled" instrumentals and no ghost vocals left behind. We use BS-RoFormer, which scores 18.10 dB instrumental SDR and 11.79 dB vocal SDR on the public MVSEP Multisong leaderboard, where the older Demucs model scores below 8.4 dB on vocals.
See the MVSEP leaderboardSeparation runs on dedicated cloud GPUs. Your 30-second preview starts as soon as the upload begins, so you hear the result before you pay for the full song.
Upload MP3, WAV, FLAC, AIFF, M4A, OGG, or video (MP4, MOV, WEBM, MKV) up to 2 GB. Download both tracks as lossless 44.1 kHz, 16-bit WAV or as MP3.
Works in your browser on desktop or phone. There is no software to install and no GPU needed. Files auto-delete after 24 hours.
Yes, you can remove vocals from an MP3. Vocal Isolator accepts MP3, WAV, M4A, AAC, OGG, OPUS, FLAC, M4B, MKA, AIFF, plus most video files, and returns both tracks as lossless WAV or as MP3. A lossless source such as WAV or FLAC gives the cleanest split, but a normal 320 kbps MP3 works well too.
| Spec | Details |
|---|---|
| Audio input | MP3, WAV, M4A, AAC, OGG, OPUS, FLAC, M4B, MKA, AIFF |
| Video input | MP4, MOV, M4V, WEBM, MKV |
| Max file size | 2 GB |
| Output | Vocals and instrumental as 44.1 kHz, 16-bit WAV; MP4 with the new soundtrack when you upload a video |
| Karaoke mode | Removes only the lead vocal and keeps backing vocals |
| Speed and key change | Instrumental maker and karaoke maker: speed 50% to 150%, key up or down by up to 6 semitones |
| Tempo and key detection | Acapella extractor: BPM and key estimated during the preview |
| 24-bit WAV | Acapella extractor: 24-bit WAV at 44.1 or 48 kHz |
| Free preview | First 30 seconds, no sign-in |
| Free download | 1 minute after signing in |
| File retention | Uploads and results are deleted after 24 hours |
| Model | BS-RoFormer (vocal SDR 11.79 dB, instrumental SDR 18.10 dB on MVSEP Multisong) |
Model scores: MVSEP Multisong leaderboard · MVSEP lead and backing vocals leaderboard
Only the lead singer removed, harmonies kept, 20% of the lead left as a guide, and the key moved to fit your range.
Try the karaoke makerEvery vocal layer as 24-bit WAV, with the song's tempo and key estimated during the preview and an optional reverb-free version.
Try the acapella extractorEvery vocal removed, then slow it to 50% or shift the key up to 6 semitones — the download keeps the change.
Try the instrumental makerRemove hum, fans, traffic, and room echo from voice.
Try the remove background noiseRemove background music from videos and keep the speech.
Try the remove background musicMake a backing track for any song, even when no official instrumental exists. The karaoke maker keeps the harmonies and moves the key to fit your voice.
Sing over a clean instrumental that matches the original. To rehearse first, the instrumental maker slows the backing track to half speed without changing its pitch.
Grab acapellas for remixes, mashups, and samples. The acapella extractor estimates the BPM and key, and the acapella stays sample-aligned, so it drops straight onto your grid.
Hear the lead vocal alone to learn melody, harmony, and phrasing. Loop the vocal track to hear breaths and runs up close.
Upload your MP4 and drop the vocals so lyrics don't clash with your voiceover. Upload the MP4 directly; there's no need to extract the audio first.
Isolate vocals to hear every word clearly. Isolated vocals make fast or mumbled lines easier to hear.
Each card is a need we hear from singers, DJs, and video creators, with how Vocal Isolator handles it.
Cover singer
“The song I want to cover has no official instrumental, and the old center-channel trick leaves it hollow.”
The AI split keeps the bass and kick in the instrumental. On our demo song, its low end below 120 Hz sits only 2 dB under the whole track.
Clean instrumentalsTikTok cover creator
“I record and edit everything on my phone. I cannot install desktop software or run a GPU.”
Separation runs on cloud GPUs and the whole flow works in a phone browser, on iPhone and Android, with nothing to install.
100% onlineTravel vlogger
“I filmed on a busy street and the traffic is louder than my voice. I do not want to rebuild the edit.”
Upload the MP4 and get the same video back with the noise removed. On our street demo, the quietest moment between words fell from −28 dB to −75 dB.
Remove background noiseSongwriter
“My demo is unreleased. I do not want it sitting on someone else's server or open to other users.”
Uploads and results are tied to your browser or account and deleted automatically after 24 hours.
Private filesDJ and remixer
“I need the vocal alone with no kick bleeding through, lined up so it drops straight onto a new beat.”
On the demo song the acapella carries 37 dB less energy below 120 Hz than the instrumental, and both tracks start at the same sample.
Acapella extractorHobby musician
“I do not want to pay before I know the split works on my song, especially the chorus.”
The first 30 seconds of both tracks preview free without signing in, "Try another part" previews a later section, and signing in adds 1 free minute of downloads.
Free previewVideo editor
“The music under an interview clip clashes with my new voiceover, and I only have the finished MP4.”
The background music remover keeps the speech and the rest of the scene and takes out the music, straight from the video file.
Remove background musicKaraoke host
“I want the lead singer gone but the harmonies kept, so the room still has something to sing over.”
The karaoke maker uses a separate model that removes only the lead vocal (10.23 dB lead-vocal SDR on MVSEP) and leaves backing vocals in the music.
Karaoke makerVoice student
“I want to hear the lead vocal on its own to learn the runs, the breaths, and the phrasing.”
The vocal track keeps the singer and drops the band, so you can loop a line and hear every breath. Download it as WAV or MP3.
Acapella extractorMusic producer
“I work from WAV and FLAC masters and do not want an MP3 round trip between me and the stems.”
Upload WAV, FLAC, or AIFF up to 2 GB and download both tracks as 44.1 kHz, 16-bit WAV, or as MP3 when you only need to share them.
Audio or video, MP3 or WAVAI separation is an estimate, so some recordings leave faint traces. These are the common cases and what to do about each.
Vocal Isolator is the quickest option when you want both stems online without installing anything. Desktop tools make more sense if you process hundreds of songs offline.
| Option | Setup | Separation method | Best for |
|---|---|---|---|
| Vocal Isolator (online) | None, runs in the browser | AI (BS-RoFormer) on cloud GPUs | Getting both stems from any device in a few clicks |
| Ultimate Vocal Remover (desktop) | Install the app and download models; a GPU helps | AI, many open-source models to choose from | Power users who batch-process songs offline |
| Audacity Vocal Reduction and Isolation | Install Audacity | Channel-based (phase cancellation) | Quick edits on songs with centered vocals |
| DAW or audio-repair plugins | Paid software and editing skill | AI or spectral editing | Engineers repairing stems by hand |
Every song gets a free 30-second preview of both tracks, with no sign-in. Once you sign in, clips up to 30 seconds download free (10 a day), and you get 1 free minute of full downloads. After that a song costs its length in minutes, rounded up, and that one charge gives you both the acapella and the instrumental. Pro includes 20 hours a month, not unlimited separation.
Separating songs for personal practice, karaoke, study, or your own music is fine. Publishing or monetizing stems from copyrighted songs needs the rights holder's permission.
Yes: a vocal isolator and a vocal remover are the same split seen from opposite sides, so one upload returns both tracks. Keep the vocal track for an acapella, or keep the music track to remove the vocals and get an instrumental or karaoke track.
You can preview the first 30 seconds of any song free, without an account. After you sign in, your first 1 minute of downloads is free. Longer tracks use minutes from a minute pack or a monthly plan; see the pricing page for current prices.
Usually not. Hollow instrumentals come from phase cancellation, which also removes centered bass and drums. Vocal Isolator uses a BS-RoFormer model that scores 18.10 dB instrumental SDR on the MVSEP leaderboard. Songs with heavy reverb, live crowds, or low-bitrate audio can still keep faint traces of the voice.
Upload audio in MP3, WAV, FLAC, AIFF, M4A, AAC, OGG, or OPUS, or video in MP4, MOV, M4V, WEBM, or MKV, up to 2 GB per file. You get two tracks, vocals and instrumental, as 44.1 kHz, 16-bit WAV files or MP3s that line up exactly with the original song.
Yes. Upload the MP3, then choose MP3 next to the download button to get the vocals and the instrumental back as MP3s, which are small enough to send from a phone. Choose WAV instead when you will edit or mix the tracks, since it keeps every detail of the separation.
Yes, on the instrumental maker and the karaoke maker: the speed control plays the song from 50% to 150% without changing its pitch, and the key control moves it up or down by up to 6 semitones without changing its speed. You hear the change in the preview, and the download keeps it.
Yes, as an estimate on the acapella extractor, which reads the tempo and key while you preview and shows them, for example "≈ 124 BPM" and "A minor". Songs that change tempo or key, or have no clear beat, can read differently, so check by ear.
No. Vocal Isolator runs entirely online: the separation happens on cloud GPUs and you upload and download in the browser. It works on Windows, Mac, Chromebook, iPhone, and Android without an app, a plugin, or a graphics card.
Your uploads and the separated tracks are stored only to process and deliver them, and are deleted automatically after 24 hours. Guest uploads are tied to your browser, and signed-in uploads to your account, so other visitors cannot open them.
Yes. The standard split moves every voice, including harmonies, to the vocal track. The karaoke maker uses a separate model, 10.23 dB lead-vocal SDR on the MVSEP lead and backing vocals leaderboard, that removes only the lead singer, so backing vocals stay in the music for a more natural sing-along.
Not at the moment. Vocal Isolator splits a song into two tracks, vocals and everything else. If you need drums, bass, and other instruments as separate stems, use a multi-stem tool; you can still use our instrumental as a clean starting point.
Yes, with a different model. Vocal Isolator's song models treat everything that is not singing as music. For podcasts, interviews, and voice memos, use the background noise remover, which is trained on speech and removes hum, fans, traffic, and room echo while keeping the voice clear.
Upload a song or video and preview vocals and music free.
Upload a SongLast updated: