AI acapella extractor
Upload a song and this acapella extractor gives you the vocals on their own, plus the instrumental, drums and bass. Hear one song a day free with no sign-up; the whole song is $0.50, with no subscription.
Drop a song here or choose a file
MP3, WAV, FLAC, M4A or OGG · up to 10 minutes and 90 MB
Free: 1 song a day, 5 when signed in · full song $0.50
Hear the acapella
A real 30-second excerpt, run through the acapella extractor you use here. Play the original, then switch to the vocals to hear the acapella that was pulled out of it.
Original. The excerpt as released, voices and instruments together. This is what the acapella extractor starts from.
“Persephone (ft. Vidian, Dimitri Artemenko)” by snowflake, ccmixter.org/files/snowflake/22364, licensed under CC BY 2.5. Stems separated by Stem Split AI.
The excerpt runs from 0:30 to 1:00 of a Creative Commons song. An acapella pulled from a mix can keep a faint echo of the instruments, most of all where the voice is in the same range as a guitar or a keyboard.
Four steps
Add an MP3, WAV, FLAC, M4A/AAC or OGG file up to 90 MB and 10 minutes long. The acapella extractor takes files, not YouTube or other links.
The separation runs on our own servers and the preview comes first, usually in about a minute. You can leave the page and come back.
Play 30 seconds of the vocals, from 0:30 of the song or from the start for a song under a minute, and compare them with the other stems. You get 1 free song a day without an account and 5 when signed in.
Sign in and unlock the whole song for 125 credits, which is $0.50. The download has the vocals and the other four stems as MP3 (320 kbps) and WAV (16-bit, 44.1 kHz).
What an acapella is for
An acapella is a voice with nothing behind it. People pull one out of a song for these jobs.
Put a new beat, a new key or a new genre under a vocal you love. Get permission before you release a remix of someone else’s song.
Blend the vocals of one song with the instrumental of another. The acapella is the half that stays, and the instrumental stem of the other song is the half you add.
With the instruments gone you can hear every line the singers take and practise each part, which is much easier than picking it out of a full mix.
Singers use the acapella to match phrasing, breath and timing, and to hear where the original singer places each word.
Vocal coaches and students listen to a voice alone to study technique, tuning and tone, free of the arrangement.
The other half of the split is the song without its vocals. Take that too if you want to sing it yourself.
A cleaner acapella
A single lead vocal over a band separates best. Choirs, heavy layering and effects like long reverb leave more of the instruments in the vocals stem.
A WAV, a FLAC or a high-bitrate MP3 holds more detail than a small, compressed file, which helps the acapella extractor find the voice.
Play the vocals stem alone on headphones and listen for a faint hi-hat, guitar or keyboard behind the voice. If it is there in the preview, it will be in the download.
For a release-ready vocal, trim the breaths and silences, and use a gate or an equaliser to remove what remains. Keep in mind the song’s licence.
An acapella has no beat under it, so set your project to the original song’s tempo before you load it. Otherwise it will drift against a new instrumental.
Free to preview, pay to keep
Previewing an acapella is free: 1 song a day without an account and 5 signed in. Downloading it is 125 credits, which is $0.50, and $5 buys 1,250 credits for 10 songs.
Good to know
Upload the song, wait for the preview and play the vocals stem, which is the acapella. If it sounds clean enough, sign in and unlock the whole song for 125 credits, which is $0.50, to download it.
The split and a 30-second preview are free: 1 song a day without an account and 5 songs a day when signed in. Downloading a whole song costs 125 credits ($0.50), and $5 buys 1,250 credits for 10 songs.
It uses Demucs htdemucs, an open-source AI model from Meta AI research, run on our own servers. The model has learned what voices sound like and estimates the vocals stem from the mixed song.
No. It is an estimate of the voice taken from the finished mix, so it can carry a faint trace of the instruments. It is very usable for remixes and practice, but it is not the original studio vocal.
The other four stems: instrumental, drums, bass and other. Each stem downloads as an MP3 (320 kbps) and a WAV (16-bit, 44.1 kHz).
MP3, WAV, FLAC, M4A/AAC and OGG files up to 90 MB and 10 minutes. It takes uploaded files only, not links.
The preview usually comes in about a minute, and a four-minute song is fully split in about three. You can leave the page and come back, and unlocked songs stay in My files.
Only if you have the right to the original song and its vocal. The acapella extractor does not change who owns the recording, so get permission or a licence before you release or sell a remix.
Uploads are deleted after 1 day and stems after 7 days. Files are only for the person who uploaded them. Copyright owners can send a notice to contact@stemsplitai.com.
Yes, for 1 song a day. The preview plays without signing in. Sign in to split 5 songs a day and to download the acapella and the other stems for 125 credits, which is /bin/zsh.50 a song.
It is the opposite half of the same split. A vocal remover keeps the music and drops the voice; an acapella extractor keeps the voice and drops the music. Both come out of one split.
Same split, other jobs
The acapella is one of five stems every split gives you. The others are one click away.