StemVocal lets you choose an audio file and hear its parts separately. The tool uses open-source separation models to estimate vocals, drums, bass and other sounds. It also provides an instrumental backing without the estimated vocal part and a drumless track without the estimated drum part. These are estimates made from a finished recording; they are not the original studio multitracks.
You get two free jobs per UTC day across StemVocal music separation, audio transcription and video transcription. Free separations of songs up to 10 minutes run in our cloud: Vocals and Instrumental return those two tracks, and Drumless and Four stems return all six, as 320 kbps MP3 and WAV. On a computer, free songs are separated in your browser when the free cloud allowance is unavailable, with all six parts as stereo 16-bit / 44.1 kHz WAV and optional 320 kbps MP3 copies made on the device; on a phone or tablet the page says when free separation is available again. Paid separation uses a credit or Unlimited, runs in our cloud and returns all six parts in both formats. Music files may be up to 80 MB and 10 minutes. Free and paid cloud separation use the same models, so paying does not buy cleaner separation; the browser uses a different model, and its support and speed vary by device. The player compares parts at the same position, while the mixer lets you adjust their levels and export a combined WAV.
The single-track preview also has optional speed and section-loop controls for listening practice. They stay in the current tab and leave the separated tracks at their original speed and length.
Save tracks individually or, for a six-track result, choose Download all tracks, then Save ZIP, to save all six together in the selected format. ZIP preparation runs on this device, makes no new separation and uses no extra credit. It keeps the original speed, key and levels.
Cloud separation uploads audio to private temporary storage, with result access ending after one day. Separation in your browser keeps audio in private local browser storage; download those results before closing or resetting the tab. The privacy page explains cleanup and retention. For backing tracks, start with Remove vocals; for individual parts, use the stem splitter.
Uploads are sent in parts and continue where they stopped when a connection drops. After three minutes without a connection the upload pauses; while the file is still selected in this tab, Continue upload sends the rest for the same separation, without using another free job or credit. If the connection is interrupted later, choose Check again to recover the existing separation without uploading again or using another credit. A paid track may require Restore purchases first. Expired results need a new separation.
If you choose Get all six tracks after a two-track result, those ready tracks stay playable and downloadable in this tab while the upgrade runs or is retried. Save them before closing or resetting the tab.
This is an independent site. If something fails, or you have a question about your audio, write to hello@acapellaextractor.com. Do not email us your audio file unless we ask for it while investigating an issue.
Practice before you separate
Choose a file to play it locally before uploading. Practice a section offers playback speed, start/end loops and a key change from −12 to +12 semitones. Speed and key are independent. Key changes run in your browser where AudioWorklet is supported; otherwise you can listen at the original key. These settings do not change the individual downloads.
To save a whole track at your chosen speed and key, open Download with speed & key and choose Export track, then Download WAV. This uses your original or the selected native stem WAV; older results use MP3. It runs on this device and uses no extra song credit. The export uses the full track, without section loops or listening-volume changes. Supported files are mono or stereo, 8–48 kHz; decoding support depends on the browser. Keep the tab visible while it works.
Record a practice clip records up to 60 seconds from the selected player while you listen, including speed, key changes and loops. Keep the tab visible. Stop recording or Pause finishes the clip, and Download practice WAV saves it as a stereo WAV. This is a recording of playback; an MP3 preview does not regain lost audio detail by being saved as WAV. It does not record the four-stem mixer or your microphone. Switching tracks, replacing the file or leaving the page discards the clip; download it first if you want to keep it.
Show waveform adds an overview of the original recording's loud and quiet sections. Tap or drag it to move the playhead, then use Set start here and Set end here for a practice loop. It reads the file on your device when requested. The overview continues to show the original when you listen to a separated output. You can still use playback and section times if your browser cannot read the waveform.
Find BPM estimates the beat from up to the first 90 seconds of your original, without an upload. Some music has an unclear or changing beat, and a result can count at half or double the tempo you hear. Correct it with Half, Double, tap tempo or the number field. Analysis requires at least eight seconds and a browser that can decode the file. You can keep listening or cancel while it runs; playback and downloads keep their selected settings.
Open Analyze a section & key to analyze a short section as it plays. Up to 20 seconds of playback are analyzed on your device, with no upload or microphone. The result suggests the track's original tempo and a possible major/minor key. It can be uncertain or unknown, and an excerpt may differ from the rest of the song. Correct the fields or try half/double tempo if needed. These notes do not change playback.
Transcription, with the same credits
Use our audio studio for speech recordings and our video studio for footage and subtitles. Sign in with the same email to use the same credit wallet. Each started 30 minutes of paid transcription uses one credit; free transcription files can be up to 30 minutes.