PEEL

zplane's PEEL is a spectral source separation plugin that isolates individual instruments and vocals from stereo mixtures with sufficient accuracy to enable practical mixing and processing workflows. The plugin analyzes incoming audio across frequency and spatial domains, then presents an interactive visual interface where users can identify and extract specific sources.

The technical approach leverages machine learning trained on isolated instrument sources, allowing PEEL to distinguish vocals, drums, bass, and other elements with reasonable fidelity. While not perfect - harmonic bleed and frequency masking remain inherent limitations of the source separation task - the results are sufficiently clean for most practical applications. Extracted sources maintain acceptable phase coherence and can be routed to separate DAW tracks for independent level adjustment, panning, and effects processing.

PEEL is best suited for remix engineers, podcast producers working with pre-mixed material, and mixing engineers salvaging problematic recordings where individual stems are unavailable. It's particularly valuable for extracting vocals from finished mixes for acapella remixes or for isolating drums when you need to rebalance a specific kit element. The visual spectral display helps identify frequency conflicts and mixing issues across the entire mix.

Compared to similar tools like iZotope RX's music rebalance or specialized stem-splitting software, PEEL offers moderate separation quality with lower CPU overhead. It won't match true multitrack recording quality, but it meaningfully extends creative possibilities when working with finished masters or archival material. The results reward careful listening and realistic expectations about source separation limitations.