What stems can Song Creator Pro extract from a song?

Extract can separate a mixed song into individual instrument tracks, typically vocals, drums, bass, guitar, keys, and other layers. The exact stems available depend on what the model detects in your source audio.

Can I extract stems from any song?

Yes. Extract works on any MP3, WAV, FLAC, or M4A file, including your own recordings, commercial songs, live recordings, or tracks you've generated in Song Creator Pro. Quality is best on clean studio mixes; heavily distorted or poorly mixed sources are harder to separate cleanly.

What is Extract useful for?

Common use cases include creating karaoke tracks (by removing vocals), sampling loops from existing songs, isolating a vocal for remixing, learning an instrument part by listening to it in isolation, and preparing stems for DAW projects. It's also a great companion to Song Creator Pro's Remix and Layer features.

Does Extract run locally?

Yes. Like the rest of Song Creator Pro, Extract runs entirely on your GPU. No uploads, no cloud processing, no subscription. Your source audio never leaves your machine.

How to Extract Vocals and Instruments from a Song

Every finished song is a mix of individual layers: vocals, drums, bass, guitar, keys, and so on. In music production, these layers are called stems. Normally you can only get stems if you have the original project file, but AI-powered stem separation can pull them apart from a regular audio file.

Extract is Song Creator Pro's stem-separation tool. Upload a mixed song, and it produces individual tracks for each of the song's components, all locally on your GPU.

This guide walks you through extracting stems from a song.

When to Use Extract

Extract is the right tool when you want to:

Make a karaoke track: remove vocals and keep the instrumental
Isolate a vocal for sampling or remixing
Pull a drum loop out of a song you like
Study how an instrument part is played by listening to it solo
Prepare stems for a DAW project so you can mix or rearrange them
Feed clean stems into Remix or Layer for further creative work

Extract is also the first step in a powerful workflow: extract a song's vocal, then use Layer to build new accompaniment around it for an instant cover reinterpretation.

Step 1: Upload Source Audio

Click the Extract tab at the top of the interface. In the Source Audio field on the left, click to upload your track.

Any common audio format works: MP3, WAV, FLAC, M4A. For best results, use a clean studio mix. Heavily distorted, mono, or lo-fi sources can still be extracted but the stems may bleed into each other more.

Once uploaded, your track is ready to be deconstructed.

Step 2: Deconstruct the Track

Click Extract All in the top-right of the Stem Board. Song Creator Pro begins separating your track into its component parts. Extraction runs on your GPU and takes a few seconds to about a minute depending on track length and hardware.

When it finishes, the Stem Board populates with the individual tracks the model was able to isolate from your source.

Step 3: Audition and Save Stems

Each isolated stem appears in the Stem Board as its own row.

Some tips for listening:

Compare stems against the original mix. Toggle between the full track and individual stems to check the quality of separation.
Listen for bleed. Extract works by modeling what each stem should sound like. Some crosstalk between similar instruments (kick and bass, for example) is normal.
Re-run if needed. Extraction is non-deterministic, so if a stem doesn't isolate cleanly on the first pass, run Extract again.

Extracted stems drop can be saved to your chosen output folder and are ready to load into a DAW, upload to a sampler, or you can use them with other features like Remix, Revise, or Layer.

What Extract Is Good At (And Not)

Extract is strong on:

Well-produced studio mixes with clearly separated instrumentation
Isolating vocals, usually the cleanest stem
Pulling out drums and bass

Extract is weaker on:

Heavily layered or overdubbed parts
Sources that are already lo-fi or distorted
Very short clips (the model has less context to work with)
Genres where instruments share a lot of frequency space (extreme metal, dense orchestral)

Like all AI features, Extract benefits from iteration. If the first pass isn't clean, try again.

A Creative Workflow: Extract + Layer

Here's a pattern that unlocks a lot of creative possibilities:

Upload a song you love to Extract.
Deconstruct and save the vocal stem.
Switch to the Layer tab.
Upload the vocal stem as the new Source Audio.
Write a description for a completely different genre, e.g. "orchestral cinematic score" if the original was pop.
Generate.

You now have the original vocals with an entirely new accompaniment wrapped around them. This is one of the most satisfying workflows in Song Creator Pro.

Related Guides

Remix: reinterpret a whole track in a new style
Revise: regenerate a specific part of a track
Layer: add new instruments and accompaniment to a track
Getting Started Guide: the basics of generating songs from scratch

Ready to start extracting? Get Song Creator Pro on the Microsoft Store for $49.99. One-time purchase, unlimited AI music generation, runs entirely on your PC. Also available on itch.io for $44.99.