Isolate a voice from a noisy recording
- Back to: Examples
Scenario
You recorded an interview in a cafe, at an event, or outdoors. The voice is present, but background sounds are constantly changing.
Goal
Make speech intelligible by emphasizing the voice and suppressing competing sounds.
Inputs
- Audio track (or extracted from video)
- Optional: separate mic tracks if you have them (best case)
Steps
- Confirm the primary speaker
- Voice isolation works best when there’s a clear target voice.
- Run voice isolation
- Start conservative; aggressive settings can produce “gating” and metallic artifacts.
- Check for failures
- Sibilance distortion (“s” sounds), pumping, missing word endings.
- Light denoise (optional)
- If there’s a steady hiss under the complex noise, apply a gentle denoise after isolation.
- Leveling
- Ensure the voice is consistently audible across the clip.
- Export and spot-check
- Listen on headphones and phone speakers (real-world playback).
Output expectations
- Voice stands out more clearly
- Less distraction from crowd/traffic/music
- Some artifacts may remain in very noisy segments
Common pitfalls
- Overlapping speakers: isolation struggles when two voices overlap heavily.
- Music: vocals in music can confuse voice isolation.
- Reverb: echo can smear speech; results may be limited.
When not to use
- The goal is to preserve ambience (documentary sound design).
- The recording is so noisy that words are not intelligible even before processing.
Related pages
- Use cases: Audio use cases, Privacy workflows
- Guides: Audio workflow