You must be registered to see the links
You must be registered to see the links
stuff for the stuff gods!!111one
Hello!
Took me quite a lot of time to sit down and make this post, for some reason I felt really anxious about it and procrastinated for literal weeks, but today I finally forced myself to sit down and make the darn thing. Sorry for the delay in communication.
Here's what I've been doing in the last month and a bit:
AI (this part is delving into my tests with various AI models, probably way too boring, feel free to skip unless AI production is something that interests you)
Tried and tested a lot of different AI solutions to figure out where the techs are and if and when I'll be able to use them in this work.
As far as visuals go, I did some tests for visual background noise, like for example generating crowds, cityscapes, empty streets etc. Unfortunately, while I did manage to get decent enough results some times, the amount of tweaking and the amount of learning needed to produce such results is still nowhere near what I'd consider as a passable alternative to the current work flow. It's still much more preferable (at least for me, with my limited experience) to build sets from scratch that can be re-used without an issue than to resort to using AI for that.
I also tried using AI for audio work, and with that kind of work I surprisingly had a much better time exploring. I tried various solutions for voice generation, sound FX and even generating music. The field suffers from various issues and needs some more time to mature, but judging by what I've seen I believe audio AI exploration will provide more fidelity compared to visual AI, as far as consistency and sharpness or accuracy is concerned. In my tests, the audio effects were pretty impressive, however they suffered greatly from diffusion. I could manage to produce a ton of unique, high fidelity sounds in a very short amount of time, but the white noise remained a constant issue. Maybe if I knew how to "clean" the audio, I'd have a ready made, 2 clicks away solution but so far that still remains beyond my reach.
Voice generation has also advanced quite a bit. I've found several high fidelity voicetovoice solutions, but the best ones out there are nothing more than voice altering masks. As far as texttovoice goes, it remains quite "neutral". It's good for announcers or news caster voices, but if you need someone to act elated, angry, mournful, it's still way off.
Lastly, music solutions. There were a couple of promising services I tried, but the ones that offered ready made, clear background music were no more than scripts disguised as "AI" that shuffled harmonic segments combined into relevant topics and re-hashed them. Services that offered original sound tracks couldn't reproduce long segments of music, with fidelity and consistency rapidly falling apart the longer the number was, an issue similar to trying to get consistent visual AI outputs.
All in all, out of all the three types of services I tried, the background sound FX seems the most promising, provided I manage to figure out a way to clean up the diffusion artifacts in the background.
Lastly I tried using AI for literary work and the results were just bad. I know some authors say they can use it for bouncing off ideas, but I'd rather watch a decent movie and wait for something to pop up than deal with the level of writing I managed to produce in my elementary (and I was a terrible writer).
Anyway, if someone has experimented as well with any type of AI feel free to share their thoughts, advice, recommendations etc.
Rendering and a broken AC!
I drastically underestimated the amounts of new sets, NPC's, clothing, props etc. that I'd have to use for the current release. Just tracking down the items that I needed and testing them out took a huge chunk of my time, not to mention actually building all the stuff that I needed. I easily increased my asset library by almost 200GB of various assets and lost quite a bit of time.
Other than that, I've been rendering as much as I could. When I say as much as I could, my AC broke around a week ago. The service guys can't come by until mid/late August because a bunch of people didn't maintain their AC's (not me though! I did everything prim and proper, regularly cleaned it and maintained it... ah, who am I kidding, yeah, I didn't touch the thing for like six months and then cranked it up like a moron...). So yeah, the temp in my place hovers around 31-34 degrees and when I'm rendering I have to move my legs otherwise the heat being emitted from the PC housing is unbearable. There's also the smell of burning plastic when I render stuff, so I mostly toss up a render or two in the early mornings while I can still have the windows open and before the temps in my place become unbearable... I don't have much rendering left, but the progress is currently glacial so we'll see. Still hoping for a release until the end of the month, but realistically speaking, I might as well hope for some snow to fall.
Comments
