Topic: Hints for using cleanup aud recordings in RX?????

[nassau73]

I don’t know if this thread will garner interest in the future. I would like to ask other RX users for their work procedures so that others can try.

So, just to start…

I record basic two channel aud recordings from eyeglass mounted mics. I can usually get close to or in a sweet spot for overall presence.

When I master my recordings, I do it mainly for me to be able to be happy to hear it in my earbuds when walking etc. If I’m happy with what I hear, I will torrent to others at several bt websites to share.

So, for me, the first process  means as much as possible eliminating women screaming as if they’ve just been raped, men whistling for no reason at all except to show everyone around them they can whistle loudly.

That’s just the first part in “cleaning” up my just recorded files.

So moving to actual helpful tips and questions…

For the screams. In the spectrogram you can see the main scream as well as harmonics. In the past, I have found most success using the paintbrush tool to highlight the main “scream” I would the either manually select the harmonics or use the harmonics tool to highlight the others and then select delete in spectral repair. I WAS setting the pixel size of the paintbrush to be just small enough   to barely cover the overall size of the scream display so as to do a “surgically remove”of just that instance. However, that would seem to leave a sort of a hash display in the spectrogram and still have the remnants of the scream when listening back.

I’m experimenting and starting to find that if making the size of the paint tool much larger that the scream or other item you are highlighting and then replace and attenuate – that seems to be more efficient. (at the same time, experiment with “partials and noise”.

Anyway, I don’t know if folks are interested in continuing a thread like this. But at least it’s a start. ‘Cause I’d love to hear your approaches for various solutions other than just the Izotope tutorials.

[nassau73]

 To help possibly move this thing along (tell me if you want it to stop)

You may have other solutions that we would love to hear.

Most of us hate to hear the loud crowd noise in between tracks of a live show recorded in simple stereo.

For me, I first eliminate claps using declick with random clicks set at 5.7 sensitivity and wide 3.8 (depending on the recording)

Again, if you have a solution, let us know. What my learning experience has helped for me is that when I look at the in between songs spectrogram, I can see all the whistles. I span across all of them with the paintbrush until all are highlighted. Then select the harmonics tool to select at least 3 or 4. Then spectral repair, delete and the attenuate (to get rid of the hiss).

Finally, highlight a few of the “brightest” applause/shouts in the spectrogram. Replace and the highlight the remaining applause you can see (being careful to avoid any in between chatter from the artist) and “Attenuate”.

[capnhook]

Put a lasso around a scream, then use Spectral Repair - Replace and mess around with the "Surrounding Region Length" slider and see if you can get something acceptable

I used it today to turn a big "POP" into a little "woooosh"

[nassau73]

One feature which I was not aware of until recently was zoom.

For example, let’s say you are willing to spend the time to meticulously edit your files. Is there somebody talking during or in between songs? You can see the dialogue in RX as well as the harmonics of their voices.

One thing I recently learned was you can use whatever highlight tool you wish  to focus in on  Then click the magnify icon.

Once that is done, you can see the various harmonics in a much larger and separated display of the offending person’s comments dialogue.

You can then highlight several of the harmonics and delete.
Then go back and select and delete some more.

Yes, it’s time consuming and you won’t be able to totally eliminate the dialogue but it can be down to almost not hearing it.
It all depends on how much effort and time you want to spend to edit.

[robgronotte]

I'm happy to post questions and hints in this thread if people start using it, but I did try to start a similar thread a few months ago that didn't get any traction.

https://taperssection.com/index.php?topic=202654.0

I have learned a lot since then though, both from other users and through trial and error.  I've started doing so much work on my recordings that I don't get finished with very many of them, but it's kind of a fun learning experience.

[tim in jersey]

marking thread...

[daze]

Thanks for starting this thread.  I find it's a lot of trial and error, and what works for one set of screams/woots/whistles doesn't necessarily work for the next . . . embedded screams and woots and whistles are my nemisis.  I need to start working at this more regularly and will share any revelations here.

[nassau73]

Thanks for starting this thread.  I find it's a lot of trial and error, and what works for one set of screams/woots/whistles doesn't necessarily work for the next . . . embedded screams and woots and whistles are my nemisis.  I need to start working at this more regularly and will share any revelations here.

I agree with you on all your points. I find that I have to try a few approaches let's say on the screams and hoots during the music information. For me, most success is using the paintbrush to highlight the offending intrusion. Sometimes, attenuating the main/loudest incident in the spectrogram and replacing eliminates any trace. Other times I'll have to select several harmonics, replace, attenuate and attenuate again. Yet, some other times partials and noise rather than replace works best.

The whistles, since they are so high up it don't seem as difficult. I'll select all the whistles, then use the harmonic tool to get all of them. Replace and attenuate and that usually removes everything. If there are any leftover artifacts, I either replace them or simply attenuate. After you do this after quite a few times, you get a feel for what you want to try next.

It sounds like you and me are working with basic two track audience recordings so there's a limited amount of how we can clean things up without individual instrument tracks. Sometimes there are still some artifacts left remaining but for me, if I'm listening to a nice spacey jam and hear a very, very muffled kind of sound versus a loud scream - it's worth the time and effort for me. Because this is all very time consuming.

[voltronic]

Marking thread, as a longtime RX devoteé who still has a lot to learn...

I second capnhook's suggestions of using the lasso select and Replace, though I tend to go for Attenuate more often if the offending noise is on top of music. I also have had good results with the magic select if you get the vertical zoom right.

Don't forget Instant Process if you have a bunch of the same repetitive things in a project, though daze is right to point out that you can't always apply the same exact tools and settings to every situation. Still, you can Instant Process with your "default" scream/whistle reduction settings and if it's not successful you can adjust.

[robgronotte]

Marking thread, as a longtime RX devoteé who still has a lot to learn...

I second capnhook's suggestions of using the lasso select and Replace, though I tend to go for Attenuate more often if the offending noise is on top of music. I also have had good results with the magic select if you get the vertical zoom right.

Don't forget Instant Process if you have a bunch of the same repetitive things in a project, though daze is right to point out that you can't always apply the same exact tools and settings to every situation. Still, you can Instant Process with your "default" scream/whistle reduction settings and if it's not successful you can adjust.

Could you explain what you mean here?

[robeti]

Marking

[voltronic]

Marking thread, as a longtime RX devoteé who still has a lot to learn...

I second capnhook's suggestions of using the lasso select and Replace, though I tend to go for Attenuate more often if the offending noise is on top of music. I also have had good results with the magic select if you get the vertical zoom right.

Don't forget Instant Process if you have a bunch of the same repetitive things in a project, though daze is right to point out that you can't always apply the same exact tools and settings to every situation. Still, you can Instant Process with your "default" scream/whistle reduction settings and if it's not successful you can adjust.

Could you explain what you mean here?

See #6 here:
https://www.izotope.com/en/learn/hidden-audio-repair-features-rx.html

[robgronotte]

Marking thread, as a longtime RX devoteé who still has a lot to learn...

I second capnhook's suggestions of using the lasso select and Replace, though I tend to go for Attenuate more often if the offending noise is on top of music. I also have had good results with the magic select if you get the vertical zoom right.

Don't forget Instant Process if you have a bunch of the same repetitive things in a project, though daze is right to point out that you can't always apply the same exact tools and settings to every situation. Still, you can Instant Process with your "default" scream/whistle reduction settings and if it's not successful you can adjust.

Could you explain what you mean here?

See #6 here:
https://www.izotope.com/en/learn/hidden-audio-repair-features-rx.html

I don't really understand the Instant Process portion of that website either.  Don't see what it is telling you to do that would make anything faster.

[nassau73]

Marking thread, as a longtime RX devoteé who still has a lot to learn...

I second capnhook's suggestions of using the lasso select and Replace, though I tend to go for Attenuate more often if the offending noise is on top of music. I also have had good results with the magic select if you get the vertical zoom right.

Aloha Voltronic:

I'm working on remastering an older recording and have been trying your preferred method of using Attenuate in addition to what I have been doing with attenuate/replace/attenuate (and variations) but pretty much using the default settings (since Izotope pretty much has their act together). At any rate, I changed the strength weighting to 2 based on your post and it seems to be much more effective when working on a 2 channel aud recording.

Since you are one of the most experienced members of this group and have probably experimented with all the settings, could you expound on the attenuate settings you are using that seem to work most often. Right now, mine is set to Bands 512. Surrounding Region length 100. Before and After Weighting 0 and Strength 2.

Of course, different settings change depending on the recording.

[robgronotte]

My default for attenuation of unwanted noises is also 2.