RNNoise settings. (Adobe and OBS different voice recordings)

[Indigo128]

Hi everyone. I'm using OBS a lot for streaming and I like it very much. Even supporting it on Patreon. Now I try to make some video montage with Adobe Premiere and having some troubles to make my voiceover sound similar in OBS and Adobe Premiere. I've made same VST effects stack, but there is trouble with RNNoise, because it doesn't have any settings in OBS. And I can't make similar settings in Adobe because of that. My voice just sounds differently on the same PC with the same mic. Can anyone give me some advices to make voices similar? Maybe I should disable all effects in OBS when recording and activate them only in Adobe for my next videos? Can this help? And what can I do with my current video anyway? Maybe problem not only in RNNoise, but these two programs somehow record voice with different default settings which I don't know either?
Appreciate any help, because I got tired already tuning settings in different ways. :)

 

[AaronD]

That's been a thing in sound ever since the beginning of sound processing at all. If you wanted a particular sound, then you had to use the specific gear that *had that sound*, even if there were scads of others at every price point that all did that same basic function. It's not unheard of for someone to use one, just for "its sound" by being in the chain at all, but otherwise set to do nothing.

Often, it was a particular combination and use of a tube and/or transformer, neither of which was anywhere close to transparent like we have today, or some other physical thing that has turned out to be difficult to model even with modern tools. The classic LA-2A compressor, for example, had a light-based gain element that was fed by a glowing panel (before the modern LED), the specific response of which was itself part of the sound, and it took a lot to figure out the details of how the physics actually worked before it could be emulated accurately.

So the idea that a setting-less digital algorithm has a particular sound, is not a new concept at all. If you want *that sound*, then you have to use *that plugin*, however you have to jerry-rig things to do it. Mathematically-perfect plugins can be mixed and matched with exactly the same results, but not everything is written that way.

And this is why you record dead-raw and do ALL of the processing later. If you record after some processing, then you're stuck with whatever that processing is, whether you like it or not. If you don't like it, then you might compensate for it in some way or another - the better you know the internal details, the better you can compensate, or know to give up - but it will never be as good as if you didn't have it in the first place.

 

[Indigo128]

Thank you very much for your answer! So... If I don't use any filters in OBS will it record raw mic sound? And... I like how noise reduction works in OBS. Can you advice any good vst noise supression? Because I choose RNNoise in OBS and everything works perfectly fine even without any tuning (there is no settings in OBS in this effect). But if I try something similar in Adobe it sounds worse. :(

 

[AaronD]

If I don't use any filters in OBS will it record raw mic sound?

It'll record what OBS got, with no further processing. A fancy driver may still do something though, before OBS gets it.

I like how noise reduction works in OBS.

If you KNOW you're not going to want to remove it, then you can record after that. It's not a crime or moral failure; it's just that you CAN'T remove it after the fact.

I've gotten to the point now, that I've always applied the exact same processing to my raw recordings, so I've started to record after that processing to same some time in post. But that's after a bunch of sessions of recording raw and doing that in post, and then seeing it converge onto something specific and stay there.

Can you advice any good vst noise supression? Because I choose RNNoise in OBS and everything works perfectly fine even without any tuning (there is no settings in OBS in this effect). But if I try something similar in Adobe it sounds worse. :(

I like the adaptive version of this:

GitHub - lucianodato/noise-repellent: Lv2 suite of plugins for broadband noise reduction

Lv2 suite of plugins for broadband noise reduction - lucianodato/noise-repellent

github.com

Home

Lv2 suite of plugins for broadband noise reduction - lucianodato/noise-repellent

github.com

But it appears to be for Linux only. Specifically, it's in the LV2 format, not VST, with no other version available.

 

[Indigo128]

I appreciate your help very much!
I've finished my current video I was making with some flaws because of different processing. But now before I start making next videos I want to define my next steps to be able to record my footage in OBS and then edit it in Adobe Premiere (rerecord separate words or even sentences on the montage phase). So I tried to find out if voice recording is the same in Adobe Premiere and OBS (without filters). Hopefully both programs don't monopolize mic, so I managed to record the same footage in both programs at the same time. Then I tried to synchronize precisely both footages and then made kinda checkerboard. I made cuts even in the middle of words to hear if it sounds some kind different (volume differences or something like that). And to my surprise and gladness the checkerboard sounded like it was whole and uncut recording. So the first step is to record footage without OBS filters.

 

[Indigo128]

My stack in OBS:

- Standard Built-in RNNoise supression. No settings allowed at all.
- Standard Built-in Compressor (Ratio 4,00:1; Threshold -20 Db; Attack 6ms; Release 60ms; Output Gain 15 Db; Sidechain/Ducking Source - None)
- Marvel GEQ Equalizer VST (with my own freq. setup)
- LoudMax64 VST Limiter (-8 Db)

I tried to make exact the same stack in Adobe Premiere.

I use there:
- wernan RNNoise supression for voice (custom, not built in). It has it's own settings (in attach).
- Standard Built-in single band compressor (with the same setup as in OBS).
- Marvel GEQ Equalizer VST (same freq. setup).
- LoudMax64 VST Limiter (-8 Db).

So, we see that two first built-in OBS effects don't have exact analogues in Adobe Premiere (even though compressor looks mostly the same, it has to be checked). So if I used in OBS only VST effects, maybe I've reached almost the same or exactly the same sounding.

I like my OBS settings very much and want to make Premiere sound very similar.
I'll try to check OBS sources, maybe I'll find settings for RNNoise...

I'll write back later. Maybe others will be interested in my discovery too...

 

[AaronD]

So, we see that two first built-in OBS effects don't have exact analogues in Adobe Premiere (even though compressor looks mostly the same, it has to be checked). So if I used in OBS only VST effects, maybe I've reached almost the same or exactly the same sounding.

If you use the exact same FX and nothing else, then it should sound exactly the same. In fact, if you put the same recording through each and record the result with the same encoder with the same settings, then those two secondary recordings should even be *bit-perfect* identical.

But, two different FX units that do the same thing with the same list of settings, don't necessarily give the same result. Compressor timing, for one example. Does a 6ms Attack mean that it's *finished* reducing the volume in 6ms? Or that it's reached 70% of the reduction in 6ms? Or 50% reduction in 6ms? I've seen all three, and nothing tells the user which it is. Likewise for the Release time.

Another odd quirk that I've seen is that, instead of applying the threshold and ratio first, to produce an instantaneous signal that the gain element can use, and then slowing down either edge of that separately like most analog compressors do, some digital ones will apply the timing first, and then the threshold and ratio. Effectively slowing down the detector itself. That makes a subtle difference in how the compressor behaves, which may turn into a massive difference with just the right source material.

So when you copy the settings from one processor to another, you're only getting into the ballpark. The exact settings might need to be a little bit different to produce a more similar result. Always check it by ear, and do the final tweaks by ear.