3

Background: Am using Isotope RX7 Standard suite tools to perform audio restoration on dialogue files with multiple data corruption issues in the sequence listed below:

(a) Declip to restore the waveform peaks,

(b) dehum to eliminate low frequency harmonics and dc offset

(c) spectral repair to eliminate 8kHz frequency noise band

(d) spectral denoise to eliminate broadband and tonal noise

(e) voice denoise to clean up further

(f) equalizer to restore integrity

(g) gain to amplify as needed

Question: I am a novice in audio restoration and RX7 and have the following questions:

(a) Is there a best practice of sequencing of above operations? Or all of them independent of sequence? For example, does it make sense to apply dehum first to eliminate the dc offset (instead of declip first)? Does this affect the declip algorithm effectiveness in a positive or negative way?

(b) Is there a "sweet spot" scenario for using spectral denoise vs. voice denoise? Is voice denoise gentler for dialogue restoration?

1

The list of activities you describe is pretty much the right way to approach this, although data corruption is not an easy artifact to correct.

I would recommend removing any DC component as the first step, then proceeding to declip so that declip can operate symmetrically across the signal.

As long as you are gentle with the denoise you are probably going about this in the best way possible. You may get a better result using EQ to remove the 8KHz artifact.

Aside from that there's probably not much more you can do.

  • 1
    Mark - above approach has been built upon your recommendation @ sound.stackexchange.com/questions/46417/…. Very grateful for the tool suggestion as well as the strategy. For files at hand, 80% of the value appears to be from declip - Izotope is unbelievable in recreation!!! Just wondering if a dc offset in data affects the declip algorithm adversely...in which case I can do dehum first. Thanks! – rambommakanti Oct 29 at 22:36
  • 1
    It is an excellent suggestion to use EQ for the 8k band...and yes, per your suggestion in earlier thread, going very gentle with noise slices - 2dB at a time – rambommakanti Oct 29 at 23:58
  • 1
    Yes, in retrospect I think you will benefit from removing DC prior to declip - simply because of the way that declip works in izotope. The declip usually works symmetrically around the signal level centre. – Mark Oct 30 at 11:27
  • 1
    Confirms my suspicion and validates some differences between files - many thanks Mark for the lead! much appreciate the help... – rambommakanti Oct 30 at 18:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.