Stem Splitter
Library
Your recent creations and history
AI Vocal Remover: Separate Vocals and Instrumentals
Upload audio to extract vocals, make a backing track, or split instruments into separate stems. Listen to an example before processing your own file.
Hear the original and separated tracks
Pixel Paws · 00:24–00:48
Pixel Paws · Ling Model · vocals and instrumental. Each version uses the same excerpt, trimmed from a completed separation and encoded as MP3 for this demo.
Click a waveform to play from that point. Use Space to pause or resume, arrow keys to seek, and Home or End to jump. Only one example plays at a time.
Vocals
00:00 / 00:24Instrumental
00:00 / 00:24Compare the start and end of vocal phrases with the original: listen for remaining voice in the instrumental and background instruments in the vocal track.
This example shows one recording with one model, not a guarantee for every song. Strong reverb, layered vocals and low-quality recordings can leave residual sound or remove instrument detail.
Separation modes, models and credits
Times are estimates; file length and provider queues can increase the wait.
| Mode | Model | Credits per task | Estimated time |
|---|---|---|---|
| Vocals + instrumental | Demucs v4 | 3 Credits | About 45 seconds |
| Vocals + instrumental | Ling Model | 10 Credits | About 90 seconds |
| Vocals + instrumental | MVSep | 10 Credits | About 90 seconds |
| Vocals + instrumental | MVSep High Vocal Fullness | 10 Credits | About 90 seconds |
| Vocals + instrumental | MVSep High Instrumental Fullness | 10 Credits | About 90 seconds |
| 6 stems | Demucs v4 | 20 Credits | About 90 seconds |
| 6 stems | Ling Model | 20 Credits | About 300 seconds |
| 6 stems | MVSep | 20 Credits | About 120 seconds |
| All stems | Ling Model | 25 Credits | About 240 seconds |
| All stems | MVSEP All-In | 25 Credits | About 600 seconds |
Credits are deducted when you submit a separation task. Failed tasks are refunded after failure is confirmed; a new attempt requires credits again.
Accounts with no paid purchase are subject to a 30-day retention policy and receive notice before cleanup. Accounts with a paid purchase are excluded from this cleanup. Download files you want to keep; this service is not a permanent backup.
Choose the tracks you need
Use separation for a specific part of your workflow, then judge the result by listening.
Vocals and backing tracks
The two-track mode returns vocals and an instrumental version. Use them to prepare karaoke practice, a vocal edit, or a remix draft.
6-stem and All Stems options
Six stems separate vocals, drums, bass, guitar, piano and other instruments. All Stems extracts every track supported by the selected model, including detailed vocals and instruments; the track count varies by model.
MP3 and WAV downloads
WAV export requires an eligible plan. When a native WAV is unavailable, the download is converted to 16-bit PCM WAV from the available audio. Changing the format cannot recover detail lost in the source.
Know what you are processing
Check the example, model and cost before using your own audio.
Public audio examples
Listen to original and separated excerpts without creating a task or spending credits.
A choice of models
Compare the available models for your selected mode; a different model can produce a different result.
Visible credit cost
The submit button shows the credits required for the selected mode and model.
Use your music library
Select an existing MemoTune song directly instead of downloading and uploading it again.
How to separate vocals from a song
1. Choose your audio
Upload a supported file or select a song from your MemoTune library. Use a recording you have permission to process.
2. Choose tracks and a model
Select vocals and instrumental, 6 stems, or all available stems. Available models, credit cost and estimated time are shown before you submit.
3. Listen and download
Compare the separated tracks with the original. Download MP3, or use WAV export if your plan includes it. Listen for residual vocals and missing instruments.
What can you use separated tracks for?
Check the audio quality and the rights to your recording before using the result.
Karaoke and singing practice
Use the instrumental to practice a vocal part. Some vocals or reverb may remain in a dense mix.
Remix and sampling drafts
Isolate a vocal or instrument to experiment with an arrangement. Separation does not grant permission to sample someone else’s recording.
Learning instrument parts
Listen to the bass or drums separately to follow a part. Quiet notes may bleed into another stem, so compare with the original.
Check the audio quality and the rights to your recording before using the result.
Separate the part you want to work with
Listen to the examples, choose a mode, and check the credits before processing your own audio.
Frequently asked questions
Formats, credits, output quality and file storage.
Upload MP3, WAV, FLAC, M4A, OGG or WebM. The upload limit is 100.0 MB and the recording must be at least 6 seconds long. Browser codec support can vary. WAV uploads may be converted to MP3 before processing.
The public examples are free to listen to without signing in. Processing your own file requires an account and credits. Eligible new accounts may receive trial credits; availability depends on account and regional eligibility.
Credits are deducted when you submit a task, before separation starts. A confirmed failed task is refunded. A pending task is not yet a confirmed failure; a new attempt is a new credit-consuming task.
Choose vocals and instrumental, 6 stems, or all available stems. The mode table lists the available models and current credit costs. All Stems supports Ling Model and MVSEP All-In; the number of tracks depends on the model.
The estimated time for each mode and model is shown in the table and tool. These are estimates, not deadlines: longer recordings and provider queues can increase the wait.
No. Strong reverb, layered singing, live recordings and low-bitrate audio can leave residual sound or remove useful detail. Listen to the original and separated tracks before deciding whether the result suits your project.
Not necessarily. WAV export is available on eligible plans. If the model provides a native WAV, that file is used; otherwise the available audio is converted to 16-bit PCM WAV. A WAV conversion cannot restore quality already lost through compression or separation.
Accounts with no paid purchase are subject to a 30-day retention policy, with notice before cleanup. Accounts with a paid purchase are excluded from this cleanup. Save your own copies of important files and check the storage and privacy policy for details.
Separation does not transfer the rights to the original composition or recording. You need the relevant licenses or permission. For music generated with MemoTune, the terms of the plan used to generate it also apply.