Dialogue and voice cleanup in iZotope RX

Start with the voice you need to keep. For an interview, podcast or voice over, identify whether the distraction is noise between words, noise over speech, room reflections or a short mouth sound. A quieter recording is only useful if the words still sound natural. Keep a recoverable original and compare repairs on both clean and damaged phrases before treating the whole recording.

Match the problem to the module

Hiss, fans and a changing noise floor. Voice De-noise is a useful first candidate for speech; its Adaptive mode follows changes in the noise floor. Spectral De-noise offers finer control and also has an adaptive mode. For a learned noise profile, choose exposed noise without speech. Our Voice De-noise versus Spectral De-noise comparison explains the tradeoffs; a changing background does not automatically rule out either tool.

Busy backgrounds and speech mixed with room sound. Try Dialogue Isolate when separating the voice is more useful than learning a noise profile. The RX 12 module and plug-in provide separate voice, noise and reverb controls in Standard and Advanced. Listen for damaged consonants and an unnatural voice before asking for more separation.

Audible room reflections. Dialogue Isolate's Reverb control is one option for speech. De-reverb also processes voices as well as musical material. Its Learn function needs the direct sound and reverberant tails, with some room tone, rather than a noise-only sample. A distant-sounding voice alone is not enough to diagnose a reverb problem.

Hum, buzz and pitched interference. De-hum targets tonal noise: Static suits a few harmonics, while Dynamic handles more complex tonal patterns. Compare it with spectral noise reduction when noise overlaps the voice; it is not automatically the cleaner choice.

Mouth noises and clothing movement. Use Mouth De-click for lip smacks and mouth clicks, De-plosive for low-frequency puffs, and De-ess for intrusive sibilance. De-rustle addresses clothing noise around a lavalier microphone. Treat distracting breaths with Breath Control selectively; removing every breath is not the goal.

Choose the processing order from the recording

For a voice with mouth clicks, hum and hiss, audition the local click repair and tonal cleanup before broadband reduction. Skip processes for problems you cannot hear. One documented dependency is to run De-plosive before high-pass filtering. Noise reduction and reverb reduction can need a different sequence; compare both orders on a representative passage instead of treating one chain as a rule.

Check phrase endings, strong consonants and quiet words at comparable playback loudness. If a repair produces a watery or hollow voice, reduce the processing, revise the profile or try a different tool. Two gentle passes are not guaranteed to beat one pass. For detailed noise-profile decisions, use the background-noise removal guide.

Listen across edits too. Ambience Match in RX Advanced can supply matching room tone where an edit exposes a gap; it adds ambience and does not remove existing noise. If further processing makes speech harder to understand, consider another take, a pickup or less reduction.

Dialogue cleanup questions

Does a changing background always need Dialogue Isolate?

No. Voice De-noise and Spectral De-noise both offer adaptive processing. Compare them with Dialogue Isolate on the actual voice and background, keeping the version that preserves speech best.

Should I remove reverb before noise?

There is no universal order. Compare noise reduction before and after reverb reduction on the same passage, then listen for clarity, natural phrase endings and processing artifacts.

Do I need RX Advanced for dialogue cleanup?

Not for every job. RX 12 Standard includes Dialogue Isolate; Advanced adds tools such as Ambience Match and De-rustle. Choose the edition for the problems and workflow you actually need.