Overview

Extract Stems is a browser‑based audio stem separator that lets you split a single track into multiple stems without installing software. The service accepts common audio files you own or are authorized to process and outputs the separated stems in several formats.

You begin by uploading an MP3, WAV, FLAC, M4A, or OGG file. The web interface asks which split you prefer: two stems (typically vocals and accompaniment), four stems, or six stems. Each split produces a separate track for each component; for example, a six‑stem split might separate vocals, drums, bass, guitars, keyboards, and overheads. The tool limits the input to 10 minutes of audio, so shorter songs or sections of longer tracks are suitable. If you want to work on a longer piece, you would need to cut it into smaller segments.

After the JSON‑based separation routine finishes, the page displays a preview player for each stem. You can play each isolated track in the browser individually, compare them, or skip to the next. When you are satisfied, you can download each stem as a separate MP3, WAV, or FLAC file. For convenience, the service also lets you download a ZIP archive that bundles all the stems together. All processing runs on the server, so the browser only handles uploads, previews, and downloads.

Who benefits most? Producers and remixers who need isolated tracks for edit, pitch, or loop work will find this helpful. A songwriter trying to replace a vocal part or a DJ riffing on a track might also use it. Because the tool is free for the first five minutes of processing after verification, hobbyists can test it without commitment. Those working with very long tracks, heavily compressed audio, or needing advanced features (like pitch‑shifting or noise reduction) might find the platform's scope limited.

The listing does not provide pricing beyond the mention of prepaid minute packs, and no recurring subscriptions are offered. It also does not state whether the separation algorithm is based on machine learning or a specific model, nor does it detail the exact handling of metadata or sample‑rate conversion. The website and submission text show no guarantees on separation quality and no support for files larger than 10 minutes.