@coolboymew stapling the n64 lora straight onto the facewrap lora doesn't get useful results, but if i had a few dozen n64 sheets i could train up something that does that specifically. i experimented with fear and hunger spritesheets for that a while back. i'd call it a success because i used a miniscule dataset and insanely unfair parameters. if i chopped up the sprites into individual poses and actually provided text files to explain what the AI was looking at, i could easily make ai-generated sprites of anything and whip up a quick script in stable diffusion to turn them into sheets
@coolboymew and since i'm already rambling about it, i trained up a fear and hunger artsyle lora but i'm still not satisfied with the results. i need to refine it for a stupid project i'm working on. i also realized i could trivially use img2img to make the game look like it was drawn by someone else, and that if i got the artstyle lora to my satisfaction i could reverse the process and combine that with a spritesheet lora to easily inject any character into fear and hunger
we are rapidly approaching a point where the idea guy is the only guy that matters
>we are rapidly approaching a point where the idea guy is the only guy that matters
More or less. Super indie games are either broken up in these categories -Art dude that can only use RPG Maker -Programmer that doesn't have art dudes -Dude who has neither and has to make Mario clickteam fusion games
Also interested (yet scared) to see where AI will take video game music, because my team (really just me and another programmer) don't have a music guy either
@hj@coolboymew nah, it can do it fine, i just fucked up part of the datasets so it looks like shit with all sorts of weird noise and jpeg-like compression around high-detail areas like hands. i know what caused it and how to fix it i just haven't yet
@hj@All_bonesJones The second pic, with patchouli in the fear and hunger like art, has some actually very potable art that you could use in an indie game
@coolboymew there's also music guys and story guys but the issue i've always had is that while i'm proficient in everything i'm not good enough to be able to do it quickly. if i can use AI to fill in for my lack of speed i could knock stuff out in under a year
@hj@All_bonesJones we're aware, but those results are still good enough at making something free and extremely indie vs not having any assets to use at all
@All_bonesJones@coolboymew i mean AI is pretty bad at consistency, i.e. making character rotations or different expressions while maintaining same style/position, maintaining fine details etc. Worst of all - refining existing art, i.e. if someone notices some detail is incorrect like ribbon color or anatomy error, right now you can't tell AI to fix it, you have to regenerate the image.
@All_bonesJones@coolboymew "massively", those arts you posted are quite inconsistent. I mean have you seen those "ai generated" videos? It's almost like watching analog TV with really bad reception - everything moves erratically and there's too much noise.
Bro, again, I'm talking about fucking clickteam fusion, game maker and etc. The kinda stuff you'd see on websites aggregating these. Here's the reality, not a lot of people can art, not a lot of people can even art good enough to the level of Smelda Macarena of Time, which is just made in mspaint
These people have to rip already made assets and that's it. But now, the AI is definitively good enough to be able to shit out "good enough" potable assets for a free ass game
@coolboymew@All_bonesJones "good enough" good enough for what? A programmer's drawing of a kirby is good enough. To me your "good enough", especially given the content y'all provide sounds more like "whatever, it will sell".
Y'all sound like AAA videogame makers in late 90's - early 00's who JUST discovered 3d graphics and decide to use it for everything, not caring if it looks ugly or not. image.png
@hj@coolboymew here's a character lora in action. is it completely 100% consistent with small details like the weird metal thing on her choker? no, but that can be touched up later with either photoshop or inpainting. is it consistent ENOUGH that the average layman will understand this is supposed to be the same character without having to put in loads of effort on my part? absolutely, this took under a minute
@hj@All_bonesJones it's already flooded with trillions of Mario fan games dude. This makes it so that those without skills may actually move on from Mario fangames because there'll be more assets
@hj@coolboymew i don't have any, but yes, you can train those. if you're going to demand that AI gets backgrounds correct and persistent 100% of the time, your argument has changed from "ai isn't good enough" to "ai is not SIGNIFICANTLY BETTER than real artists." even highly praised artists like murata forget stuff like the locations of minor background objects like shelves and which way the doors swing!
also i figured you were going to demand more varied poses as proof the character lora can handle those but you didn't so have the thing i rendered while i took a shower anyway
@hj@coolboymew >those mario fan games look better than that. >stolen image collection learning model. oh, ok. you're just butthurt about AI, don't understand how it works, and are arguing in bad faith :blobcat3c:
@coolboymew@All_bonesJones and those mario fan games look better than that. I don't think people who make most mario/sonic fangames will move on to make original or other games just because there are assets. I mean most of these games were made because people like mario, not because they wanted to make a game and only had mario sprites at their disposal.
Those without skills should obtain better skills, i.e. learn to draw or be resourceful and creative, instead of learning how to write the prompt and make your stolen image collection learning model.
hj is often arguing in bad faith, but it's clear that the dude has no experience with the extremely indie scene of random clickteam and game maker games made by extreme amateurs
I had so much fun with the Biebersoft games, Tatsusoft, French RPG Maker community and the absolute random shit I'd find on personal websites' download sections
I remember, on RPG Maker 2000, a SUPER ambitious action rpg made on the engine, starred Link going into the past from a space age to save Zelda or some shit
Literally used Zelda characters, there were bunnies from Secret of Mana to kill, Link would go in powered up mode every 2 mins, the combat was jank and barely worked for the engine but damn was it ambitious. I remember they did one of those cutscenes where a mountain rise with an entrance and I was like "wtf how did they do that" because that's super complicated to do on 2000. I opened the project and still didn't get it
@All_bonesJones@hj I remember they had a Zelda RPG project, attempt to recreate Zelda in RPG Maker 2000. Obviously this failed badly. Like lmao, the boss was a giant jpg of a Lizalfos that you had to hit in action rpg style
They eventually moved on to game maker. But years later they made a parody of their first attempts and it was so fucking kino. They made a Zelda dungeon based on that bureaucracy hell part of The Twelve Tasks of Asterix (where Asterix has to go to x and x counter to get the right document and etc) and it had the jpg Lizalfos boss at the end. That was amazing
@All_bonesJones@coolboymew am I wrong though? "bad" art or simplistic art at very least is self-aware. AI art bros will swear by god, mouth foaming, that it looks good or at least "good enough".
@romin@PurpCat@All_bonesJones@coolboymew@hj the pizza machines apparently have huge mechanical flaws but it seems like making a mcdonalds burger robot should be the simplest thing.
the last one I read that made any sense to me said "tag anything that's *not* the thing you're trying to train" and that's really the only thing that stuck.