String zing and the Frequency Repair tool

I’m trying to use the FR tool to get rid of some nasty acoustic guitar string zing but it doesn’t work over an area, it would seem. I don’t know if it would have done the job but it strikes me as a useful facility.

Incidentally, it turned out in this case that SL identified the worst of it as Vocal and has done a pretty good job.

If anyone’s got any hot tips for dealing with this, please post here!

Hmmmmmmmm it sounds like the resolution is to broad and pixel-y.

Try to play with FFT size and resolution and refinement and try again.

Thanks for the tip! It pushed me in the direction of Components rather than Stems and the results are much better. Granted, you don’t want to be losing all the transients but it’s relatively easy to distinguish the ones you need to keep.

Could really do with a tool that repairs spectra over a range, rather than just a single frequency as at present (afaik).

No problem! Magnifying the FFT Size and Resolution/Refinement is hinted at throughout the manual of producing different results. The reason why (I assume) that none of the tutorials on youtube (about spectralayers) demonstrates this is because optimization of Spectralayers is so poor (and the calculations of increasing the FFT Size/Refinement/Resolution are so cpu intensive) that the screencapture (used to record those tutorials) wouldnt be able to keep up and probably cause the GUI to hang.

FYI: Somewhat belatedly, I’m struck by how various elements in the display come in and out of focus, or even disappear or enlarge massively, as you move the FFT slider. I think it never quite sank in what I was seeing. So I just need to decide whether it’s better to (literally) focus on the harmonics (and keep them) or transients (highlighting the zing for removal…

Glad I could help and bring it into focus for you.

Right! I know exactly what you mean. I’ve suggested a feature like broader selections tool (so you can get the best of both worlds) based off of GUI where you can choose and select the partials/overtones/harmonics/transients, that way you have more of an option to select exactly what you want. That way you choose your selections based off of what you see and not have to be forced to choose between a higher or lower FFT size. I find that when you decrease the FFT size (and this varies from audio to audio from song to song) you get the sharp (natural sounding) attack of the transients(drums and percussions) and vice versa when you increase the FFT you get a more (natural) sound of the harmonics/overtones(instruments/synths/vocals).

1 Like

Ok, I’ve got something I can use. The key decision was probably unticking the Transients option (!) before splitting into Components. I think Noise and Transients (in this case) were too similar so differentiating them turned out to be counter-productive.

I put the resulting Noise channel through RX7’s De-reverb (via the internal External Editor command), which cleaned it up enough that I could pick out the bad noise (the zings) but keep the transients. This method resulted much fewer drop-outs than I had been getting so the Frequency Pencil became a practical option.

It’s a bit laborious but this is an old tape with the perfect groove and totally worth it! Thanks for your expertise - couldn’t have got there without you. :grin: :+1:

1 Like

I forgot to mention that Heal does a better job than Delete. (But not all the time - see next post!)

@Crotchety

I stay far away from the “healing” tools and other tools like “clone stamp” tool because it introduces interpolation and reconstructs audio material that wasn’t originally there. It might not sound like it at first but when you start using those tools, you’ll get the sense that the audio sounds unnatural, it’s subtle but noticeable. Using tools like “clone stamp/healing” also interferes with the phase and throws the entire mix balance off course.

Its best to cut new layer, that way you’re not introducing any interpolation and you’re not interfering with the phase. That way you can always go back and use phase inversion.

What would be great (and I’ve talked about this feature being implemented before) is if you can mold from other layers to fill in the holes in a natural sounding way, that way the layer sounds as natural as possible without introducing any oddities or interpolation. However (currently) there is no way to do an overall selection (based off GUI) and select all the dark spots/holes and then invert your selection to then copy from all layers and then mold back the selections (from all layers) back to the layer you want holes filled in.

Thanks for the advice. I should probably delete that post because I think I’ve already come across what you’re talking about. The frequency pencil doesn’t do a bad job of patching holes but it’s very fiddly and I’d prefer something more similar to the way RX does it with Spectral Repair, although this is far from perfect unless there’s plenty of good audio either side. Not the case here unfortunately, as the zings happen at a slid chord change.

I love messing about with restoring audio (I’ve got plenty of old tapes to practice on!) but it doesn’t half eat into production time. Still, if a job’s worth doing…

Cheers! :grin:

1 Like