Snower day

March 07, 2023

Captain, it is Tuesday.

Tuesday really Tuesdayed today. Even though I actually got some things done, despite everything, the day was even more tired and focus-defying than usual. Up to and including lunch, my best and clearest achievement was my ten or so minutes of pushing snow around on the driveway.

Podcasting (or "talking to a friend", as I am told the more general case is called) helped as usual. After that I had generated enough energy to actually focus on and quickly solve a few things which had totally overpowered my mind before noon.

Plenty of snow has fallen, and more is supposed to be on the way, but I was honestly hoping for even more. This is just a nice big snowfall, not something deserving of the warnings they put out before. Not that I would have gone out in a car or tried to travel by train or bus, but still.

Oh look, it is supposed to feel like -18 at times tonight. That, I must say, is something for around here. Perhaps we should turn up the radiators and … roast some chestnuts or something?

Audio research

I have started looking around for audio processing stuff, and began by digging up Jason Snell's post about audio processing again. Let us see what a trial version of RX 10 advanced can do to that room sound …

The installer is 1,8 gigabytes?! That is a lot to unpack. I hope the uninstaller works well if this thing is not to my taste … The installation process is by no means Xcode-slow, but it certainly is not fast either.

This interface is … very much its own thing. It looks very much like audio software, the information density is really high, and there is a slight feeling that one careless click can have huge and unexpected effects. But I think I can find my way around it.

Oh, things do not apply in real time. Fair enough, and reasonable, but quite the difference after being used to Hindenburg mostly doing stuff live.

But yes, it does use all my cores! Look at that lovely processor load in Activity monitor, over 700%!

Wow, look, there are specific tools here which sound tailor-made for me: "Voice de-noise" and "Dialogue de-reverb". So much to play with. Just got to wait for this current de-hum application to finish …

The Mp3 encoder only uses one core. Nobody can have multi core support everywhere, I guess.

But yes, there is a difference. Fun stuff! Wonder what it will do to some noisy track from a poor microphone?

(Time passes.)

Oh yes, there is quite a bit of difference to be had here. And also quite a bit of learning to be done.

For my first test track - the raw recording for Kodsnack 515 - the "Repair assistant" did a great job with the settings it entered itself after listening to some sample audio.

For my second test track - the raw recordring for Kodsnack 451 - the "Dialogue isolate" tool removed a lot of background hiss without completely destroying the voice, and a quick test of "Spectral recovery" even produced a bit of depth which was never there to begin with.

All of this is guaranteed to just be scratching the surface. I should probably both read up quite a bit, and play around with really short pieces of audio to get a feel for it. As it is now, I keep getting tripped up by being used to live or near-instant processing in Hindenburg, selecting longer sections of audio and then waiting a minute or two to hear the result. No good, need shorter iterations.

To sum up: great research results so far 🎉 Please get in touch if you have tips for tools I should try.