Audio normalization / mixing for "imperfect" vlog audio

Question

I am trying to (easily) post-process my audio for a vlog which I've recorded. I used the DJI Mic for recording, and have two speakers. As it is an amateur vlog, the recording environment is not really perfect – which is fine for me, I just want to get the best out of it which I can. My main challenges are currently the high dynamic range, e.g. one speaker changed the volume in the middle of one clip, cars driving by or something hitting the lavalier mic. As said, I don't really care so much about the quality – mostly what I would like to achieve is to have a consistent loudness. So if it sometimes clips or there is some background noise – that's okay, I'll approach that in the future :D

I'm editing in DaVinci Resolve, but I guess I could do the same thing in Audacity or some other audio editing software. I've already watched quite some YouTube tutorials about it, but somehow they don't tackle this specific scenario (they talk more about short videos recorded in a studio). My video is about 20 minutes long.

What would be a good workflow for the audio in the screenshot? For example, the speaker in track 2 speaks relatively quiet at the beginning, then gets a bit more excited at 02:07:00 and speaks much louder afterwards. Additionally there are also some peaks.

• I know about compression to reduce the overall dynamic range, but for me it seems that I first have to set the individual clips to the same loudness more or less, so that they afterwards end up the same. Otherwise I don't really understand how I should work with the makeup gain.
• I tried to normalize loudness to e.g. -23 LKFS with -1 dBTP, to run the compressor afterwards. Unfortunately, that didn't work at all because of the loud peaks in the clip. Other clips without any peak get much louder in general.
• Just normalizing to true peak also does not work.
• I tried to manually decrease the volume of the peaks with keyframes; which worked relatively nice (although quite some manual work). Unfortunately, then the problem was that the later part of the clip in track 2 was normalized to -23 LKFS and the first part was much quieter (even though the first part should basically be the “normal” loudness and not the later part).
• On the other hand, I somehow also get the impression that it might be just a manual process, and I have to play the clip, observe the mixer and then see whether the “normal” part of the clip is more or less hitting the target loudness which I want; with the other parts of the clip just being louder (which I don't really mind, I don't want to compress everything extremely much) and getting rid of the large peaks with a limiter?

Any good “real life” workflow help would be appreciated :)

Answer 1

My take on this.

The first to try if you have it (believe it is only in Beta for Studio version) is the dialogue leveler. If that does not work go manual. You may automate the volumes but that works best with a motorized fader. Another take is to cut the cake:

.1) Identify "areas" with very different volumes. Examples are the peaks you see in speaker two or the soft and strong volume areas.

.2) Cut around each area (before and after) using the Razor tool.

.3) Mark the whole timeline and do normalize, selecting Independant.

Each clip is now normalized to at about the same perceived volume. You may now elect to adjust the volume further or to add a fade-in or fade-out. Sometimes further divide the clip helps and normalize

I generally add a high pass filter (say 75Hz for spoken word) and a bit of compression using the dynamics tool.

Davinci Resolve has a lot of tools for detailed work with audio but as with most of the program they may take a bit of experimenting to understand how to use them.

Answer 2

Until about 6 months ago, I would have posted a long, convoluted tutorial on how to achieve this using gates, compressors, limiters, noise removal software…

There's little point these days as Adobe Enhance will do it all for you, for free*, better than I could do it with £1,000 worth of pro software.
It takes mono or stereo wav or mp3, but gives you back a mono file; so send it each channel separately.

*Requires signup & is limited to 1 hour max duration; 500 MB max file size; 3 hours total daily limit.
It's currently in beta, so try it while it's free. They may start to charge for it eventually.

Answer 3

you could try the "clip fix" tool from audacity - clippers shave off the peaks of your audio, also it's a simple tool to use - just mark your audio track + add the effect and turn down the threshold form 100% to maybe 95% first and check how far the waveform peaks got reduced. I would turn the "Reduce amplitude" option to 0 at first but ofcourse you can play with it and find a sweetspot. This would help with the pops and clicks, since they sound nasty anyways, but be aware that the clipper changes the tone of your audio peaks (makes it griddy)... but that should be fine with clicks/ this method gets used a lot in loud bass music like Dubstep for example. If you have to cut deeper into your audio and not just the tips of your audio you might want to try a limiter/ this would do a better job in preserving the tone of the audio parts hope that helps :)