Bring a recording
Use an audio file you own or have permission to work with. Start with a short, clean excerpt.
Kostenlose Audio-Werkzeuge und praktische Anleitungen. Musik trennen, Aufnahmen verstehen und eigene Mischungen im Browser erstellen.
Kostenlos · Ohne AnmeldungChoose a layer, then press play. Nothing plays automatically.
An original, synthesized 8-second loop. These prepared layers demonstrate mixing, not AI separation quality.
Independent audio tools by NettoYang. Made for trying, learning, and listening.
Separate vocals and instruments into 2, 4 or 5 stems. Listen closely, adjust the balance, and save a track or your whole mix.
Before you process: check duration, decoded sample rate, channels and sample peaks. See the waveform and understand what the numbers mean.
Use an audio file you own or have permission to work with. Start with a short, clean excerpt.
Try 2 stems for vocals and backing. Choose 4 for bass and drums, or 5 when you need a separate piano estimate.
Compare the separated tracks, adjust their levels, then export a stem or Mix. Separation can leave audible artifacts.
Product-specific guides with real settings, trade-offs, and things to check when a result is not what you expected.
Guides in English
What each model produces, where guitar and piano may end up, and why more tracks are not a quality setting.
Read the guide FILES & EXPORT · 3 MIN READThe actual Stemix output settings, realistic size estimates, and the difference between saving a stem and saving Mix.
Read the guide TROUBLESHOOTING · 2 MIN READA listening checklist for separation artifacts, with practical adjustments and clear limits on what this tool can fix.
Read the guideBegin with the vocal/backing model. Check the chorus for remaining vocal reverb before exporting.
02Start with 4 stems and lower the drums or bass. Keep the original nearby as a reference.
03Inspect the waveform and sample peaks before deciding how to export or mix.
Yes. Separation, mixing, local history and MP3/WAV exports are free. You do not need an account, credits or a subscription. Ads help support the site.
The tools process selected audio locally. Models and components are downloaded, but your audio, filenames and results are not sent to XSL. Advertising and site delivery make separate network requests.
Stemix accepts files up to 256 MB and audio up to 10 minutes. Processing also depends on available memory and a compatible WebGPU browser. Audio Inspector has a 64 MB limit.
No guarantee. These are model estimates of a mixed recording. Reverb, overlapping instruments and lossy compression can leave bleed or a watery texture. Read the guide on separation artifacts.
For Stemix, start with an up-to-date Chrome or Edge with WebGPU and hardware acceleration. Audio Inspector uses the browser’s audio decoder and does not require WebGPU.
You are responsible for permission to process and reuse a recording. Separating it does not grant ownership or permission to publish. See our terms and use content you are authorized to work with.
No account. No upload queue. Just your browser, your audio, and room to experiment.