Suggestion: A basic "Vocal normalize" or "Vocal expander" filter.

[JD_Mortal]

I know you can sort-of cobble this together with a "compressor", followed by an "expander". (In that order, handled with care, so you limit clipping, before expanding. The limiter being a last resort, after, for poor DB control emergencies.)

Also adding the noise filter here, as a pre-filter. Selecting that filter here, would disable it if it existed anywhere else, as a separate added filter. (You don't want to do this when noise is present. Again, it just removes a bit of confusion from the process. Playing with filter orders and multiple filters.)

Why one single filter?

Simplicity, reduced confusion, reduced trial and error, reduced overlap in settings and one less filter to slow down audio processing, adding more delays.

These settings only. (Avoiding all the technical babble, sticking with real human terms.)

Noise filter: Enabled, [TYPE]
Start: -40 DB (point where volume begins to raise)
Boost: +24 DB (Volume to raise to, as a max)
Speed: 5ms (How fast to raise or lower 1 DB)
Limit: -12 DB (Point at which the boosting should not go past.)

The filter taking care of the Gain and also managing the Hard Limit. Increasing the gain past the start point "Threshold", on a soft curve, if possible. Targeting the users speed value. Before reaching the limit, "threshold", where it soft curves out of the desired gain. Avoiding reaching that peak of 0 DB in the meter. (A forced limit of -6, as a minimum and another max limit that is your start plus the boost value, plus 6 DB. Making sure it is impossible for impossible settings to be used here.)

This is a real specific use, but also the most common use, abused by many, using a combination of 4-5 individual filters, which also seems to compete with one another and create distortion when setup incorrectly.

Having them as one singular filter, above every filter, except possibly a "Gain" filter, would save a lot of us headaches.

Though this is specific to microphone vocals, it could be used in other audio, as that may contain vocals too.

A fuller featured bonus would be the inclusion of a 3 to 5 band "Vocal EQ", applied after the normalization and noise removal. Except possibly a hard high-pass clipping filter and a hard low-pass clipping filter, applied before noise reduction, to trim the non-vocal tones from being included in any processing. (Tones below 40 Hz, and tones above 14 KHz. Tapered so 20 Hz and lower, is 0 DB and 18 Khz and higher, is 0 DB. Most adults can't hear past 17 Khz, but kids can. That is the most unpleasant frequency to everyone. Especially in vocals.) maybe offering a greater adjustment to trim it down more, but not higher than 140 Hz for lows and not less than 9 Khz for highs.

 

[AaronD]

Sounds like you've got some good ideas. I wonder how hard it is to code that up and make a plugin out of it?

One nitpick though:
I think the high- and low-pass filters should have "ridiculous" ranges of adjustment. For example:

• I've run a Behringer X32's channel highpass (called "Low Cut" over there) all the way up to its limit of 400Hz on a vocal mic, and she sounded lovely with that!
• For a custom pure-analog audio rig that I'm in the process of building, I have a design for a channel highpass that can be adjusted all the way from 20Hz to 2kHz. Yes, it's really that high! And selectable slope too. Inspired partly by the X32's limited options and what I *really* wanted to do with that...

And with *that* being said, the developers really don't want to support OBS as a DAW. In fact, there's been some talk about *reducing* OBS's audio capabilities in the app proper, and requiring the community to create a separately-maintained plugin for each general application.

Some would be simple, with very few settings so as to not confuse people who don't do much. Others would practically be their own full-featured mixing consoles. But the OBS devs themselves would not support any of them. They're all plugins maintained by someone else.

 

[JD_Mortal]

Yeah, I have a microphone input that registers from 2 Hz, up to 28 kHz. There is an odd "notch", around 22 kHz and 18ish kHz. (Zero noise floor. Something is trying to cut out specific frequencies that only bars can hear! Just not a flat cut-off, like I would like it to be. Before noise filtering has to process that extended range.)

I try to keep my internal gain in the mic, as low as possible, to reduce undesired noise and frequency extents. (It outputs digital audio with a post-DSP added gain, managed separately.)

 

[AaronD]

Yeah, I have a microphone input that registers from 2 Hz, up to 28 kHz.

Are you sampling at 96k? There's rarely a reason to go above 48k, and that has an absolute hard limit of 24kHz content. Usually closer to 22kHz in practice, just to give the anti-aliasing lowpass some room to work.

I try to keep my internal gain in the mic, as low as possible, to reduce undesired noise and frequency extents. (It outputs digital audio with a post-DSP added gain, managed separately.)

Generally, best practice for least noise is to have all of your required gain right up front, to set the amount of headroom you want, and then keep all the rest of the chain close to 0dB gain throughout. Only at the very end, do you go from that to what the final output actually needs.

The Allen & Heath analog console that I'm basing my rig on, has a range of +10dB to +60dB (roughly 3x to 1000x) at the mic preamp. That's the first active thing at all, and then the signal is all the way up to nominal already. (it also has an optional 20dB pad (divide by 10) before the preamp, to give it a total range of -10dB to +60dB) The faders go from off to +10dB as needed for the mix, but they're supposed to be around -10dB to +5dB or so, most of the time. The mix busses have a slight loss to them as well, just to not be overwhelmed by all the channels feeding into them, and only the final outputs have another (fixed) gain to get from the console's internal design level to the industry standard for interconnected gear.

For digital processing and broadcast, you'd set the mic pre to give about 18dB of headroom or so, unless there's a specific reason for the raw signal to be different (sometimes there is, but usually not), keep that working level throughout everything, adjusting the makeup gains of any dynamic processing (compressors, expanders, etc.) to get back to that same average level again, and only the final mastering turns it up from there to full-scale while ensuring that it never tries to exceed that.
Since you kinda need that mastering anyway, it should be the first to add, and if it also fixes the other problems, that's great! Leave it like that.