Monday, May 19, 2025

This is the forty-fourth installment of West Wind, your daily drop of thoughts, ideas, and info for this Season.

Here at ZMT Labs we take few things seriously, since there’s an opportunity for a joke in so many places. But one thing we do take seriously is not using generative AI for media and assets. Many of my fellow authors put up a quick generated image for their posts, which helps draw readers in visually. This doesn’t work for us, not only because of the ethics, but because it’s much easier to ask the illustrator: “It’s a pig with turnip leaves on its head and the top half is purple” and have her come up with a quality design in an afternoon, rather than muck around with the abominations that AI image generators come up with.

So the resolution was made to avoid this at all costs. Therefore, when I was looking for a good noise reduction plugin for Audacity, the audio recording and processing app that I use, I was intrigued to see one included in an “AI effects suite.” There are several different functions, from transcription (don’t need that, I’m reading off a script already), to music generation (don’t need that, I have lots of music resources already, and can make more), to noise supression (wait a minute, how is that AI?).

As near as I can tell, it compares the audio track I put in to what it thinks speech is supposed to sound like, and removes all the background noise naturally generated by microphones and cosmic rays and whatnot. You can hear the difference very plainly.

With only Audacity’s standard noise reduction effect:

With standard reduction and the OpenVino Noise Suppression “AI” effect:

I’ve already used it in all of my previous audio work: “End Quotes Save Lives, A Christmas Corral, Dragonfruit Express, Bellowing of the Ox, and of course Avalon Lost.

But the big concern was: Am I helping feed Skynet by giving it samples of my voice? Will the AI Overlord’s giant head have my voice when it declares dominion over the world? No, because the plugin doesn’t use an internet connection. Everything is done locally on my computer, nothing is beamed back to the mothership. So, I get a high-quality, professional-sounding end product, without any of the drawbacks from using AI. Then the question, why is it called AI in the first place? Every application that involves any kind of creative input is marketing new AI functionality, so it makes sense to get on the bandwagon with it. But it’s really more like Diet AI; it’s made with similar software design principles to AI, so it works like AI, with none of the drawbacks like collectively destroying human creativity.

An even better question: does this mean that the audio production of Paper & Feathers is finally underway? The answer is yes! Honestly, this goes beyond just an audiobook, since I’m not going to be simply reading the story out loud and publishing it. There’s going to be music, sound effects, and a couple of other fun tricks included too. It will end up being the definitive version, rather than a derivative version. So much of the story is not only what the characters say, but how they say it, and I can only convey so much on the page. Through this new media, I’ll be able to bring the Euphony of the Seasons to life in an entirely different way.