Sidenote Mic for recording smallish meetings of people Omnidirectional mics make more sense than cardioid mics (Especially bad since I was a triple major). This never went up on my blog at the time because I got hit by the college truck. If you are really interested, I wrote up a bunch of this stuff way back when (TODO: edit and polish). It planted the seed in me, in other words. The filetype restrictions really do make this program pretty niche).Įver since then, I’ve been determined to find a way to have algorithms magically do audio cleanup for me so that I don’t have to spend any time doing it manually myself. I still think the writeup on that page is an excellent jumping-off point into the issues here, and while it’s been a while and I don’t remember all the specifics, I do believe the program worked, at least to a degree. Since this was a single-track setup (using a sensitive condenser mic to pick up multiple people at once), I couldn’t normalize different speakers in isolation.Īt this point in time, I came across a program called Levelator that claimed to do magical things to single track audio files to even out the volume between speakers. After a couple recordings, it became clear to me that people closer to the mic appeared noticeably louder on the recordings than everyone else.

Using a USB audio mixer, I set up a boundary microphone to capture the audio of everyone around a table. You’re good to go! The initial impetus behind all thisĪ while back, I held an in-person Bible study. But in case you’d like to stop reading here, if you have ffmpeg installed and on your path, to clean up the audio on your video, just run: ffmpeg -i inputfile.mp4 -c:v copy -filter:a dynaudnorm outputfile.mp4 Is it dynamic range compression? Normalization? Do I need to record multitrack audio to get things to a “good enough” state, or is that overkill and more bother than it is worth?Īll of these things will be covered below. I spent many hours learning about audio to figure exactly what it is I want. I want anyone watching my content (even including my toughest customers: any of my old listeners out there with hearing degeneration) to be able to clearly make out what we are saying on the recordings – to be able to make use of the content with basically zero practical disadvantage, even if it isn’t the most beautiful-sounding audio mix they’ve ever listened to.

This is in line with the content-first, function-over-form paradigm I am trying to adopt in all things. What I want is for all speakers on every recording I have to be roughly volume-matched (so no one is noticeably louder or quieter than everyone else – kind of a tall order given that I’m a pretty loud person), for the audio to be relatively clear (without background noise getting amplified to the extent that it obscures speech), and without clipping or distortion anywhere, because that’s just annoying. I could care less about audio balance, audio color, EQ, and so on – an audio engineer can have my head. I want to have excellent, clear audio on all the videos recordings I upload.

Use levelator to increase volume