How To Use Izotope RX's Spectral De-noise to clean up noisy podcast recordings.

[@tkelong3569]

Thanks for taking time out to make the video. Sounds pretty good but that 2nd pass on the example in the 8 minute range, sounds boxy, 1st pass sounded better.

[@jesse.mccune]

I agree that the second pass wasn't ideal. Capturing a noise profile in that way is challenging, especially when doing it live, and the goal was to demonstrate how to capture a profile in that scenario. Ideally, I would use Voice De-noise as the second pass to clean up the noise embedded in and between the words. These days, I'm not even using Spectral De-noise, or RX's standalone editor anymore because I can get comparable results in less time with newer tools. The other thing to keep in mind is that I'm not going for perfection when I edit and clean up podcasts. People are more likely to stop listening because of noisy audio than if it sounds a little boxy, so that is what I'm aiming for.

[@living-wisdom]

@@jesse.mccune I'm curious to know what the 'newer tools" are that you find preferable to RX editor, etc.

[@jesse.mccune]

@living-wisdom When I say that, I'm talking about plugins for de-noising and de-reverbing dialog. I'm still using some RX modules like Mouth De-click, Voice De-noise and De-plosive as plugins. The newer tools are Supertone Clear, Accentize dxRevive, and Cedar Audio VoicEx. You can check out this video where I compare them against the new Dialog Isolate plugin from RX11. kzbin.info/www/bejne/l4arnZh5jNWla6M I will say that I edit podcasts, so plugins work for me. If I were doing something like an audiobook, I would use the RX editor for a number of things. The RX editor has its place, but for my day-to-day needs, I prefer to do all my work in my DAW.

[@James-tk2yl]

Thanks a lot!

[@jesse.mccune]

You're welcome! I'm glad it was helpful.

[@ericfischer5605]

Nicely explained. I would have have first run a high/low cut EQ to roll off some of the very high end, and all of the garbage under the fundamental of the voice with a very steep low cut filter. Removing these frequencies first makes for a little less work for the spectral noise reduction module to deal with. Not sure it matters, but it makes me feel better :-).

[@jesse.mccune]

Since I run De-plosive later in my chain, I don’t run a low cut. I typically just handle that as part of my EQ during the mix phase. Running it before or after shouldn’t make a noticeable difference to the result of the de-noising, but the De-plosive module needs that low end energy to work properly.

[@BlackLeafSkate]

Hey! Great video thanks for this! Ive got a problem and i was wondering how you would tackle it. I was recording a classical concert on a new dell laptop, and it experienced a buffer underflow a couple of times. I found out where the problem is and took care of it but the audio now has this machine gun sound over the audio, and the worst part is, it goes from 0 to 20khz. Just a dash over the audio, could the spectral denoise work on it? I havent managed to try it out yet, but i fear it might degrade the audio because it goes through all frequencies. Thanks again!

[@jesse.mccune]

It's tough to say without hearing the audio in question, but I'm pretty sure it will compromise the audio to some degree. Is the sound consistent across the entire recording or only when you had the buffering issues? If it's intermittent, I would probably start by trying Spectral De-noise and see how it fares. If that doesn't work, I might try De-click and De-crackleon those sections or even attempt to manually paint it out . Regardless if which method I attempt, my goal would be to do as little as possible to take the edge off the noise while maintaining as much of the underlying music as possible. If the sound is throughout the entire recording, I'm not sure that it's going to be salvageable without compromising the music too much since it runs the full spectrum. From the way you're describing it, I wonder if SpectraLayers might be able to help. I haven't used it, but my understanding is that it tries to separate the different elements into different layers, similar to something like Photoshop or other graphic design programs if you're familiar. Essentially, it would try to split your recording into one layer with the music and another with the noise, but like I said, I don't have any experience using it, so I don't know how well it works or how easy it is to fine tune.

[@OneStepToday]

how to use it for live streaming audio, the spectral denoise?

[@jesse.mccune]

You won't be able to use Spectral De-noise for live streaming. It's built for post-production and not capable of processing live audio.

[@OneStepToday]

@@jesse.mccune its working for me nicely via Light Host, on Voicemeeter Banana. Its not too heavy.

[@danzai]

I used steelseries Sonar for this. It's perfect at denoising live.

[@TawakalGTC]

How will you then export the audio? Tried to export it multiple times but the result non audio empty paper like thing.

[@jesse.mccune]

Here's a short video showing how I export a file out of RX. www.loom.com/share/535c3f033b27460b9f33190f71a734a9?sid=c7837880-5747-4216-8eee-a6768c0890fc

[@jefjaeger]

Everyone keeps saying that you have to find an area where there is only noise to take a sample from. Some recordings do not have that option. I have a voice and some cell phone clicking getting on the lav mic. The noise occurs during the voice. Apparently, I can't remove the noise then, only mitigate it. Frustrating.

[@jesse.mccune]

Spectral de-noise isn’t the best tool for your situation. Have you tried Voice De-noise? If that doesn’t work, you could try de-click or maybe painting it out in the spectral editor. Intermittent sounds like that are the worst. It’s something I don’t bother with for client work. In my opinion, the ROI just isn’t there.

[@jefjaeger]

@@jesse.mccune A combination of declicker and decrackle have been the best so far, but I was hoping to go surgical. The spectragram shows the noise takes up a lot of real estate. Painting it out effectively seems unlikely, but I could of course experiment. But yeah, that would have to be on my time, not the client.