Stem Analyzer

See the vocal, drum, bass, and harmonic balance of your track — visualized as a radar chart.

Drop your MP3 to analyze stems

or click to browse — MP3, WAV, OGG, FLAC

How stem analysis works

Stem analysis estimates the energy balance between four musical components: vocals, drums, bass, and other (harmonic/melodic content). We use frequency-domain analysis to estimate each band's energy contribution.

The analysis examines frequency bin ranges from an FFT decomposition: sub-bass and transients for drums, low frequencies for bass, mid-range for vocals, and the remainder for harmonic content. Smoothing prevents noise from skewing results.

This is an estimation, not source separation. For true stem isolation (removing vocals, isolating drums), you'd need a dedicated ML model like Demucs. Our analysis tells you the energy balance — how much of the mix is devoted to each element.

Frequently Asked Questions

Is this real stem separation?+
No — this is frequency-domain energy estimation, not source separation. It tells you the balance between vocal, drum, bass, and harmonic energy in your mix. For actual stem isolation (extracting vocals, removing drums), you need a dedicated ML model.
How accurate is stem estimation?+
The estimation is most accurate for tracks with clear instrument separation. Heavily compressed or distorted tracks may show less distinct separation. The results are best used as a relative balance indicator, not absolute measurements.
Can I use this for mixing decisions?+
Yes! The stem balance view is useful for identifying if your mix is bass-heavy, vocal-dominant, or well-balanced. Compare your track against reference mixes in the same genre to calibrate.
Why does my track show high "Other" energy?+
The "Other" category captures harmonic and high-frequency content — synthesizers, guitar harmonics, ambient textures, cymbals, and high-frequency reverb tails. A high "Other" value often indicates rich harmonic content.

Related

Key Detector

Find the musical key and BPM of any song.

Read more →

Mood Analyzer

Science-backed mood detection for any track.

Read more →

How Stem Separation Works

Learn about frequency-domain and ML-based stem isolation.

Read more →
Analyze a track — free