How can I improve this chain of filters to make a better voice with Audacity

Question

I'm a true beginner in sound design, and use of softwares to edit sounds. However I started to make a chain of filters in Audacity (File > Edit Chains).

I'm recording podcasts with only my voice as a sound (male voice, just mono) and I tried to use Audacity to edit the original soundtrack in order to have a better, clearer and warmer voice. I created this filter:

You may wonder how looks the curve called deep_3, here it is:

I am pretty satisfied of the result but I have few questions:

1. I'm not sure about the current order of the filters. Is it correct? Maybe some filters must be applied only after or before other ones.
2. Is there any filters that are redundant? Or that I should replace with better ones?
3. Should I change some parameters of the current filters?
4. Any additional filters, other suggestions?

/!\ PLEASE It's my first post here and I am a beginner, so maybe for experts this chain is crap that's why I want to improve it. Also, I'd like to keep my software Audacity because it's convenient for me, I don't want to overthink that process.

Answer 1

Processing always has the negative side effect of either increasing noise or taking away from the quality. The secret to processing is you want to do as little as possible to achieve what you want, and in this case that means removing steps that essentially do the same thing.

As for the order here's what I would do.

1. Normalize - This will amplify it to the right level and remove any offset for the rest of the processing. Always do this first.
2. Equalization - This will clean up the sound and you want to do it before messing with the dynamics. But the curve needs some work. Human voices don't have much in the 100hz range, so your boosting something that isn't there. Even with a male voice I would roll off everything below 120hz and maybe even do a 6db or so cut to everything above 12khz. Boost between 200-500hz a little to get that warm feel. You can also play around with a boost in the 1 - 5khz range to give it some presence. Save it as a preset and you can use it in the "CurveName" field.
3. Noise Reduction (maybe) - If you need noise reduction, you would do it here. The compressor will amplify the noise so you want to get rid of it before then. However, If you are recording with the proper level and mic placement into a decent recording system you wont need it at all. Noise reduction can kill audio quality if overdone, so use as little as possible to accomplish what you need.
4. Compressor - The compressor reduces the dynamic range, meaning the difference between the loud and quieter parts become less. Leveler does the same thing but less control. You'll have to play with the settings for your particular setup but try a ratio of 3.5 to start. The ratio is how much it compresses the sound, the threshold is at what amplitude that compression kicks in. When normalize=yes it will amplify the output to the proper level for you so you don't need an amplifier plugin.

The only other thing you might want to play with is a little reverb. Keep it subtle, but a little bit of good reverb can help fill out a voice an make it sound a little bigger.

Answer 2

I'm guessing there are going to be as many variations on this chain as there are answers. [It could reasonably be claimed to be "too broad" but here goes...

My brain doesn't work in numbers, it works in sounds, so I'm not going to comment on your individual plugin settings - they will all likely need changing if you swap to my recommendation anyway, as I would get the gain stages in a completely different order.

I'd start with this & see how it fares...

• Normalise - get the general peaks right before you go any further
• Equaliser - before compressor, so you're comping what you already want to hear rather than getting your comp right then EQ-ing the result. This is even more important if the comp is multiband, imo. [I have a side bet on this being the most contentious issue, comp/EQ or EQ/comp. See my note in comp...]
• Compressor - for the above reasons. Note: Normally I'd have a soft comp earlier in the chain - probably on my input to 'tape' [usually in the pre-amp itself] & my 'big' comp here.
• Amplify - I'm assuming you're using this like some kind of channel gain, though tbh, you could do this at the output of the comp or in your Leveller & drop this element altogether.
• Leveller - I'm guessing what this plug does, but I'm assuming some kind of "broadcast compressor". Whether it's brick-wall, multi-band, etc is likely to affect the result, but in any case it feels like it ought to be last.

I've left out Noise Reduction - because where that appears in the chain really depends on how well it works between each stage. You just have to keep moving it down the chain & see where it does the job best. I'd guess somewhere between Normalise & Leveller, but that's as close as I can guess. I wouldn't put it first or last in the chain.

Late thought - you could probably do all of this with the noise reduction & one good multi-band compressor, though getting the settings right might be more difficult.

Answer 3

The Audacity manual on this page has this to day about the order of applying effects:

• Amplification or normalization may be done before or after Noise Reduction.
• Do any Notch Filtering or Click Removal before doing noise reduction.
• Do any compression or any other effects not mentioned above after doing noise reduction, not before.

Also, please see the yellow-highlighted section below the above text in the manual, near the bottom of the above-referenced page.

Answer 4

While I agree with Tetsujin's answer and some of Steven Smith's answer (I'd compress before EQing), I have to disagree regarding noise reduction.

I've always removed all unwanted noise before any other processing is done, even normalization. It's usually the very first thing I do. I would amplify before noise reduction, but only if the original signal had a significantly low amplitude.

I believe it's important in noise reduction to have the noise at the original, consistent level, this makes the removal process cleaner and easier, and you can then start the subsequent processing with a clean signal (I'm assuming this noise reduction isn't a simple noise gate). If it is a simple noise gate, then I may apply normalization just before it. Amplitude adjustments don't really make much of a difference in this situation anyway. It's just the way I like to do it.