Remove background scoring from audiobook and narration recordings

You receive a narration recording from a client or archive. It has a music bed under the voice — either intentional (cinematic documentary style) or accidental (headphone bleed during recording). For your workflow you need just the narration, without the music.

AI vocal separation gets you there in one step. Upload the audio, extract the vocals stem, and you have the narration with essentially no music residue. The output is ready to be re-scored with your own music, encoded into audiobook formats, or delivered as a clean voiceover file.

Audiobook chapters are typically 15-60 minutes. At typical production quality (44.1 kHz, 16-bit mono WAV), that's roughly 80-320 MB per chapter — above our free tier limit in most cases.

Workarounds:

- Process at 24 kbps mono MP3 — sufficient quality for speech content and brings an hour-long chapter well under 100 MB. Convert after upload.

- Split long chapters — process in 15-minute segments, re-assemble in your editor.

- Use the Pro tier — 300 MB uploads cover most audiobook chapters at full quality.

AI vocal separation removes music; it doesn't fix other problems. For a polished audiobook workflow, typical post-separation processing:

- De-esser — tame sibilance that becomes more audible without the music masking it.

- Gentle EQ — restore low-end warmth (music beds sometimes disguise thin vocal recordings).

- Leveler / compressor — smooth out dynamics across a chapter.

- Noise gate — remove low-level room noise between phrases.

Tools like iZotope RX, Adobe Audition, or Audacity's plugin ecosystem handle each of these. Run separation first, then the rest of the chain.