Load a song. rebasser separates it into stems, transcribes the bassline note by note, and replaces everything below the crossover with a synthesized sub that plays the same notes one or two octaves lower. The output is mastered loud with a clean true-peak ceiling.
- SeparateDemucs (htdemucs) splits drums, bass, other, vocals
- TranscribepYIN pitch track, energy and pitch onsets, tuning-aware snapping
- Synthesizesine sub with light harmonics, octave-shifted into the 24–58 Hz register, ducked under the kick
- Masterglue compression, loudness target, look-ahead true-peak limiter
- Privacy
- Your audio stays on this device. Nothing is uploaded.
- First run
- Downloads a 90 MB model once. After that it's cached.
- Speed
- On a GPU (WebGPU) separation is fastest. CPU-only takes about twice the song's length on a 4-core machine. Re-renders after a settings change skip separation and take seconds.
- Listen on
- A system that can reproduce 25–40 Hz. On laptop speakers the new low end will simply be missing.
- Decode
- Model
- Separate
- Transcribe
- Render
0:00 / 0:00
Bassline
original note sub note sub register
Drag to pan, scroll to zoom. Hover a note for details.
Spectrum
original rebassed