Post

found a whole operation of unmanned youtube channels making completely unchecked long form slop videos where the ai voice simulacra regularly trips up & does this for a full ten minutes every time. all the very legit comments are like "NO! THATS NOT TRUE! YOU LIE HE DID NOT" & never acknowledge it.

3.6K reposts11.3K likes

anyways since this popping off hi i'm ompuco, i make games & art & also stream a lot (including the entire journey exploring the network of channels leading to this clip) if you like hand-made techart, weird artsy games, vulnerable furry thoughts, or cursed explorations, consider taking a look~

also regarding everyone commenting that it sounds like he's turning into an insect: i see that i am being directly provoked bsky.app/profile/ompu...

ompuco@ompu.co

I can finally share the full version of my BUG-OUT segment from my 2022 Funny Rave set Track is «Rewind« by John Nieuwsma (bonezone/hell edit) used emulated NTSC visuals for the text intro and prior, then real composite mixing with my Panasonic WJ-AVE5 mixer

currently having premiere detect cuts across 20 hours of archived videos from these channels so i can then sort by longest (the stutters)

oh my god if you make a compilation of the AI voices completely falling apart you'll save me the trouble (and bandwidth) of seeing how much it happens LOL

Managed to extract about 7 hours of constant stuttering from 20 hours of content. I also tried overlaying 38 of the 10-minute clips and it's wild lmfao

Now that I’ve heard about it def was not expecting the WJ-AVE5 to be so affordable 👀 have an Edirol V-4 and probably get some fun layered effects with the two together (+ finally got my own camera working couple days ago)

I had a completely different involuntary reaction - I couldn't stop laughing. Not like full out laughing but those chuckles that are somewhere between "is this person serious?" and "oh fuck, he might just be insane and kill me". And I mean it when I say it was involuntary.

At first I only watched a bit of this before peeping the rest of OP's comments, and I thought "Oh it's just tripping over it's own words and fumbling hard, I do that all the time! How much worse could it be? :clueless:" I was, in fact, woefully unprepared for the Sounds of Squelching Flesh and Bone

This is the auditory version of when AI images were first showing up andinstead of faces they had those nightmarish voids. I miss those days...

I think it’s keyword based and it searches through a pool of clips or something. Like it’ll just show random clips of the wrestlers that are being talked about, or images of random tweets when talking about something that happened on the internet.

stole this one from Twitter user tiny_tillos which never fails to make me laugh "A few years ago I discovered that siri was completely incapable of explaining what Chagas disease was. It was unfortunately fixed and it was the worst update ever"

look, i hate when chuds call anyone they disagree with an NPC as much as the next person who isn't a complete sociopath, but you look at the kind of person who thinks this is acceptable enough to put where other people can see it and tell me there's a human soul behind those eyes

Almost funnier is the people responding to these videos as though this is something a human is doing, like "dude if you keep turning into Yog-Sothoth in the middle of your videos I'm gonna stop watching your channel"

For some reason, the program glitches out on ‘what’ and due to the context being ‘what what what what’, it repeats it faster and faster cause it seems important. This trips the failsafe of ‘clearing throat’ to continue, but without extensive context it devolves just like the what.

Wow those simulated wet mouth sounds, we need to spread this far and wide so we can get the misophonics who one-star podcasts for a single lip pop as our staunchest allies in the war against AI slop

As someone who listens/records audio professionally, it's not too hard to drink a single glass of water before recording, or put a pop filter over the mic. No one likes hearing it; some people's brains can just tune it out easier. You will never find a client asking for "more mouth clicks" lol

it's mimicking naturalistic human speech, and humans mess up speaking all the time, right? what it's getting wrong is that people usually don't repeatedly fail to say the same word over and over again for long periods of time

I heard some crap like this (not quite as bad) when talking to a corporation helpline last week -- i almost sounded real except for all the repetition and stuttering

nah, i alredy upon "tech" chnnls with same thing also Louis Rossman talk about it, but with "right to repair" topic ENG videos not my main content, but i now anxious about any "new" chnls, cause unfortnate often genrtd voice sounds natural (except this shit in post), as for me not native eng speaker

funny enough, in Warframe 1999 there's a group of enemies called Technocyte Coda that's just a bunch of ai generated content on a computer pretending to be a boyband and they're constantly screeching tts glitches when they're not throwing voice clips of the original artists at you.