AI Voice Remover
Use this voice remover to split a song into vocals and backing track. Listen to both versions and download the WAV files you need.
You get the instrumental and the vocals as 2 WAV files.
WAV · AIFF · FLAC · MP3 · M4A · OGG · up to 25 minutes and 500 MB
Checking your runs
Your file is processed on a cloud GPU by our processing partner and deleted there within 24 hours. We keep the results for 24 hours, keyed by the file's checksum, so a repeat in that time is free.
More AI tools for your audio
- Stem splitterVocals, drums, bass and other; six stems with guitar and piano on Pro. Mix, solo and download each stem.
- Noise removerRemove hiss, hum and room noise from speech or music. Music runs in your browser; speech uses a cloud model.
- Lead and backing vocal splitterLead vocal in one file, backing vocals in another, the instrumental in a third.
- Echo removerReduce room echo and slapback in spoken recordings. A room uses a cloud speech model; a single repeat runs in your browser.
- Audio cutterTrim a file to a region with fades, and export it without re-encoding losslessly.
- AI masteringDrop a mix, hear a master in about a second, compare at matched loudness, export with a report.
How it works
- What it does
- Use this voice remover to split a song into vocals and backing track. Listen to both versions and download the WAV files you need.
- Model
- A separation model on a cloud GPU splits the song; two stems are the average of a two-stem and a six-stem pass. Measured on six songs with every part known (SDR, higher is better; the untouched song scores −3 to −5 dB): vocals 15.3 dB and instrumental 18.5 dB on average; drums 14.6, bass 13.3, guitar 11.5 dB; piano 4.7 dB as a quiet layer, 12 to 17 dB when it leads. Nothing is clipped, and the files keep your sample rate. Model names and licences
- Output
- One 24-bit WAV per stem at your file's own sample rate. A stem that peaks past full scale (a loud master) comes as 32-bit float instead, so nothing is clipped. Vocals and instrumental add back up to your file at −108 to −114 dB on our test songs; the song without a stem is made from the others and adds back up with it exactly.
- Time
- About 15 s of GPU time for a 45-s excerpt at standard quality, plus 1 to 3 minutes the first time while the model starts.
- Your file
- Your file is processed on a cloud GPU by our processing partner and deleted there within 24 hours. We keep the results for 24 hours, keyed by the file's checksum, so a repeat in that time is free.
How to use the voice remover
1. Choose a source for the voice remover
Drop a file into the voice remover or choose one from your device. Start with the original stereo mix if you have it. An MP3 works, but a lossless WAV or FLAC gives the voice remover more detail in quiet passages, cymbals, and vocal tails. Avoid running the mix through an equalizer or a denoiser first. You can trim a copy of a long recording if you only need a chorus or a verse.
2. Check the file and start separation
Check the filename, length, and runs remaining in the confirmation panel. Once you confirm, the voice remover uploads the mix for cloud processing and estimates two tracks from it. The voice remover shows when the file is uploading, waiting, or being processed. A model that has been idle may need time to start, and a long track can take longer than a short clip. Keep the result open until you have saved the files you want.
3. Preview and download your voice remover results
Play a busy part of the original, such as the chorus, and switch to Instrumental without moving the playhead. Then switch to Vocals and listen for words, breaths, and harmony. The voice remover keeps the preview versions aligned so you can compare the same few seconds. Download either stem from the voice remover as a 24-bit WAV. If you plan to remix the tracks in an editor, put them at the same start time.
What the voice remover gives you
An instrumental for singing or practicing
The voice remover creates an instrumental with the rhythm, harmony, and other sounds it assigns to the accompaniment. For karaoke, check the intro and count how many beats you have before the first lyric. Listen through the end of each verse as well. A quiet harmony left by the voice remover may be easy to work around at rehearsal, but it can be distracting when you use the track as the backing for a recorded cover.
An isolated vocal for hearing the performance
Use the voice remover’s vocal track to study phrasing, make out a lyric, or fit an a cappella into a remix. Follow a line from the first consonant to the end of its reverb. With strong vocal effects, the voice remover can pick up some of the surrounding mix. For editing, listen at normal volume and leave a little space around each phrase before cutting so you keep the breath and note ending.
A custom blend when you want some of the voice
Sometimes a quieter singer is more useful than a fully instrumental track. Open the voice remover’s mix controls to lower the vocal stem while keeping the backing at a usable level. Keep a little of the original singer for cueing in a rehearsal, then turn that stem down further as you learn the entrance. Solo each stem in the voice remover before exporting your blend to check which track contains the sound you want.
Voice remover workflows for karaoke and remixes
Make a backing track for karaoke
Start with the instrumental from the voice remover when the song is in a key you can already sing. If the chorus sits too high, check the key and transpose the backing separately. Start with a small change and listen for roughness in sustained notes before exporting. For a live rehearsal, save the voice remover output to the device you will actually play it from. Test the first cue and the ending in that player.
Reuse a vocal in a remix
Save the voice remover’s vocal stem and the original so you can line up the first downbeat against a reference. A singer can come in ahead of the beat, and the first word may be a poor alignment marker. Find the bar lines using the original drum or bass pattern, then set the voice remover export to the same offset. If you need the drums or bass on their own, the stem splitter gives you separate instrument tracks.
How to check the quality of a voice remover split
Check reverb, harmonies, and sounds that overlap
The voice remover works from a finished mix, where instruments and voices can occupy the same frequencies at the same time. Doubled vocals, a long echo, distorted guitar, or a choir under a lead can make the boundary harder to hear. Play a loud chorus and a quiet ending in the voice remover to check for traces of the singer or a missing instrument. Try a higher quality source or a short excerpt to see if the result fits your project.
Match your source to the job
A clean studio mix gives the voice remover a different starting point from a phone recording of a gig. Crowd noise and room reflections can cover the singer, and drums recorded through the same mic may be hard to separate. The voice remover is designed for songs. For spoken dialogue mixed with effects, use the dialogue tool. For a voice you want to keep while reducing background hiss, use a noise remover.
Voice remover FAQ
Can I use the voice remover for free?
Yes. The voice remover shows your available runs and when they reset. You can start without an account, and a free account has a monthly allowance. Check the counter before submitting a new song, especially if you are comparing several takes.
Which audio files can I upload to the voice remover?
Common formats include WAV, AIFF, FLAC, MP3, M4A, and OGG. The voice remover accepts files up to 25 minutes and 500 MB for a standard split. If a file is too long or its format is unsupported, trim or convert a copy and keep the original.
Can I download only the vocals?
Yes. Select Vocals in the voice remover to preview the stem, then choose its download option. Keep the original mix if you are taking the voice remover export into another project. It gives you a timing reference for the pickup, beat changes, and the last line.
Can the voice remover isolate the lead from backing vocals?
The voice remover groups singing together in the vocal stem. The dedicated lead and backing vocals tool can separate the main singer from harmonies. Listen to both tracks because doubled lines can be shared across the results.
Does the voice remover run in my browser?
You choose the file and preview the result in your browser. The voice remover separates the audio in the cloud after you confirm the upload. Save the output you need to your device. The privacy policy explains which providers handle the audio and how long files are retained.
Why can I still hear some singing in the backing track?
Layered harmonies, reverberation, and instruments that follow the melody can leave traces in the voice remover’s instrumental. Listen on headphones to identify a word or its echo. A less compressed source may help, but a voice remover split from a stereo mix is an estimate.
Will the voice remover’s tracks stay in sync?
Both voice remover stems come from the same source and follow its timeline. Put them at the same position in your editor to line them up. If you trim one stem for a project, note how much you removed so you can align it again later.
Can I publish tracks I made with the voice remover?
Check the rights to the recording and the composition before sharing or selling any voice remover output. For a public project, start with music you own or have permission to adapt. Keep a record of the license if you plan to distribute the separated tracks.