100% Free & Local

Your 100% private
audio toolkit

Eight audio tools, from vocal removal to pitch and voice changing — all done on your own computer. No uploads. No account. Works offline.

  • No upload — audio never leaves your computer
  • No account — open it and use it, no limits
  • No tracking — no analytics, no beacons
  • Runs in your browser — fetch a model once, then offline
Start with the Vocal Remover or scroll on and see what else is here — eight tools in all

This site has no server and no upload endpoint — open it and use it; close the page and nothing is left. Press ⌘K to look up a tool.

Why do it here

Online audio tools rarely fail on quality — they fail on the pipeline: upload first (the song leaves your computer), queue, pay by the minute, download the result back. This site folds the whole pipeline into the browser: the models and the math run on your machine, and there is no upload step at all.

Audio never leaves

Everything runs on your computer’s CPU / GPU. There is no upload step — once the model is cached, pull the network cable and it still works.

Fetch once, run offline

AI models download once on first use (30–136 MB, from our own CDN) and the browser caches them — after that the tool runs offline.

No accounts, no minute packs

No sign-up, no login, no “10 free minutes”. The only limit is physics: memory gets tight on very long files — 20 minutes at a time.

Exports you can use

WAV (16/24/32-bit float), MP3 up to 320 kbps, lossless FLAC — individual stems or the whole set as one ZIP.

Three steps, start to finish

Every tool follows the same flow; they differ only in which model runs and how many stems come out.

Add a file

MP3, WAV, FLAC, M4A and friends — video files (MP4 / MOV) work too, reading their audio track. Drag it in.

Click once, watch the bar

The model runs on your machine — speed depends on your computer, usually the same order as the audio’s length. Cancellable at any point.

Listen, mix, download

One row per stem: Solo on the left, waveform on the right. Export as WAV / MP3 / FLAC, one file at a time or all in a ZIP.

Common questions

The things worth asking before you use it.

Does my music get uploaded anywhere?
No. This site has no upload endpoint at all — the processing happens in your browser (WebAssembly and ONNX inference under the hood). The only traffic is in the other direction: the model file coming down to you. Once it is cached, pull the network cable while a separation runs; it keeps working.
Does it cost anything? Any limits?
Free, no account, no limits in the licensing sense. The only limits come from physics: browser memory caps how long a file can be — 20 minutes at a time, less on the stem pages where several tracks come out at once.
Which formats are supported? What about video?
Audio: MP3, WAV, FLAC, M4A/AAC, OGG, Opus — anything the browser can decode. Video: MP4 / MOV / M4V usually work directly (the browser’s own decoder hands over the audio track). MKV and AVI containers generally do not — export the audio to WAV elsewhere first if you hit one.
How long does a song take?
It depends on your computer. A recent laptop usually takes the same order of magnitude as the song itself (a 4-minute track in a few minutes); more cores or a WebGPU-capable browser go faster. The six-stem split runs a heavier model than the two-stem page. The page shows a progress bar and a rough time estimate as it goes.
Do the model downloads eat my disk?
Each model is 30–136 MB, fetched once by the tools that need it and cached by the browser. All of them together are about 440 MB — use fewer tools and it stays smaller; the cache can be cleared from browser settings at any time.
Can I use the results commercially?
The tools themselves do not restrict what you do with the output — but the copyright in the music is not yours to take: splitting someone else’s copyrighted song and then publishing or selling the stems needs their permission. Exactly the same as chopping up that song in any editor.