The Art of the Shitpost

BEHOLD:

I remade the trailer to Robot Monster, to serve as a trailer for a hypothetical remake of... you guessed it, Robot Monster.

Destruction has come, hu-mans, and its silliness will not protect you.

My thoughts and how-to process blog post under the fold.

"Cool, I didn't realize it had gotten to the point where it can generate a narrative."

To which my response was:

"It can't."

But that statement made me realize that work remains to do to lay out what it takes to make something like this. So to communicate this, I thought it might be illuminating to show some raws and my editing timelines:

The audio timeline is relatively basic at first glance, but this is the last step after compiling an initial version, using that for the Suno prompt, then editing and remastering that version, and that was what was used as the main timeline component here.

And here's the video timeline:

Vidu makes clips of 4 or 8 seconds. Minmax makes six second clips, and neither really does so at a speed or framing you'd prefer. All the transitions and title effects (minus the electric effect on "shocking, etc" is editing in post.

These gifs are built from the raw output. You can see that a lot of what I wound up using had to be carefully cut around the Bethesda-class wonk, sometimes taking just a second or two, reversing speed, flipping the frame, etc.

And a lot just didn't work initially and had to be reworked or approached differently.

Stuff really hasn't gotten that much more stable than in the "Will Smith eating spaghetti" meme era, it's just now things are just stable enough that someone patient and with some editing chops can produce enough material to weave into a narrative.

AI doesn't make videos so much as it makes video clips. The same way that Midjourney can't make comics but it can make parts of a comic. In short:

i am not putting this in the master and commander tag

mandelbrot raven

generated by @philpax using neural network generator midjourney in an iterative collaborative process with me using previously generated images and previous prompts with the above prompt as the final step

!7th Century French Arcade

The opulence of French arcades under the Sun King represents one of the greatest historical gaps between rich and poor gamers, and was one of several flashpoints leading to the revolution.

The image(s) above in this post have not been modified/iterated extensively. As such, they do not meet the minimum expression threshold, and are in the public domain. Prompt under the fold.

Neural nets are at a stage of development that's honestly remarkably like when lasers were first invented - they're a solution in search of a problem. Problem is, unlike lasers:

While the groundwork was done by FOSS enthusiasts and researchers researching for research's sake, a lot of the big loud recent strides have been made under the banners of private companies that want to turn a profit about them, as much and as quickly as possible, and

We're kiiind of on the brink of potentially seeing another tech bubble burst in the wake of 2020 lockdown profits turning out to NOT be a sign of infinite explosive growth, which came as a surprise to absolutely no one except a bunch of rich people who unfortunately were also the ones who had the capacity to fuck around under the assumption that the growth WOULD be permanent and infinite.

So what we have is a situation where people have poured a ton of money into these things - and now they want, nay, NEED, for it to be A Product that they can extract profit from ASAP. Problem is...how?

Making art (illustrations/writing/audio) cheaper? Turns out that if you just let it do its thing you get the most generic garbage imaginable even if it's not AS wonky as it used to be; in order to get something useful out of it you STILL need someone with creative and technical skill running the thing, who still needs to be paid, and it's not like companies were already paying the artists they thought they could replace with a magic button a huge chunk of their revenue anyway, so as it turns out, often enough: money saved < the cost of developing the thing and paying someone to run it. Oops.

Novelty toys? Hobbyist use? Weird experimental stuff? Currently the main thing they're good at! Also, broadly not profitable. Imagine if all lasers EVER turned out to be good for was pointing at things and playing with cats; their invention would likely be seen as a massive waste of money and effort, right? The profit just isn't there. In fact, it's rare for these uses to break even with their operating costs.

Enhancing search engines - or damned near any other "informative" use, for that matter? Worse than useless, it's actively dangerous-

...but at least it's visible, easy to integrate, and looks good on a quarterly report. At least you can sell it to big-money clients. Ding ding ding! We have a winner - at least, in the same sense that you might get by squinting too hard at the word "wiener".