Lift the voice. Keep the music.
Upload an audio recording and our Hybrid Transformer neural network separates it into Vocals, Drums, Bass, and Melody — plus a clean, phase-aligned Instrumental mix.
Drop a song here
or browse your files · MP3, WAV, FLAC, M4A up to 60 MB
Technical Overview
How Hybrid Transformer source separation works
Traditional vocal removers rely on simple mid-side phase cancellation, which often strips out center-panned bass and snare drums. Sagewave uses a dual-domain Hybrid Transformer architecture (Demucs) that analyzes both time-domain waveforms and complex spectrograms simultaneously—cleanly isolating vocal harmonics while leaving drums, basslines, and acoustic accompaniment intact for practice, backing tracks, and music education.