Mastering Studio Budapest

Speech cleanup

Speech-focused noise reduction with adjustable blend and multiple processing modes.

Try it with my recordingFind help for my audio problem
Audio editor showing a real recording waveform, spectrogram and tools
A real recording in the editor. This screenshot demonstrates the workflow; evaluate treatment results using your own audio.

When should you choose this tool?

Phone interviews, voice-recorder notes and street conversations often contain changing backgrounds. Speech cleanup offers an alternative to profile-based denoising. Try both on separate versions of the same passage.

How it changes audio

Speech-trained noise filtering estimates which components belong to voice. Quiet or unusual speech can also be mistaken for background. Strength blends original and processed audio.

What do the parameters control?

Processing
Basic, Fast · speech and Wideband · speech modes. Fast operates at 16 kHz and loses high frequencies. Wideband processes at 48 kHz; availability is checked in the editor.
Cleanup strength
A 0–100% blend. Lower values retain more original audio. Fully processed audio is useful for comparison, but may not be the best final setting.

Try a short passage

  1. Choose a passage containing soft speech, stronger speech and background noise.
  2. Compare available processing modes on the same Preview passage. Do not judge only the pauses.
  3. Increase strength until the background is less distracting while word endings and timbre remain acceptable. After applying, listen to the surrounding sentence too.

What to listen for

Check sibilants, breaths and distant speakers. A clean but incomplete phrase can be less useful than slightly noisy, understandable speech.

Limits of the method

The tool neither transcribes nor authenticates words. For overlapping voices, consider Speech separation or Extract speaker.

Preview covers a short passage. After applying, review the full treated section; you can undo the step from history.

Related problems and workflows

Other useful tools

Noise reductionReduce hiss, fans and persistent background noise using a learned noise profile or adaptive estimation.Extract speakerFollow a chosen voice using a clean reference passage from that speaker.Speech separationCreate two estimated speech sources as new mono clips.