16 hours ago, ScumCoder said:Oh God, it's like the response to the GaaS video all over again.
This one is on me though, I explained this part poorly. I left a pinned comment to clear it up. Basically, he would be right if that's what I was advocating, but I'm not, but i could easily see where people thought that.
On 1/9/2021 at 8:42 AM, TangledTwine said:This is the video that finally pulled me into creating an account here, so well done. Love the idea, and it really gets the brain juices going.
To make sure I'm understanding the idea right: Ross is proposing brute-forcing the ripping of 3d model and texture data (mostly for environments) by using an AI to analyze footage of the exploration of a game space and then attempt to rebuild it in an easily usable engine format (say unity or unreal). The advantage of this is obviously that it's universal, but it's also a hell of an ask (and obviously Ross is aware of this).
However, in the spirit of going for the low-hanging fruit first, wouldn't it make more sense to just make a tool that collects, automates and standardizes ripping tools? If you have the right tools for a given engine, it's very do-able to just rip map data, models, animation, lighting etc, convert it over to a standardized format and re-assemble the pieces into a working space again (in an automated, push-button way no less), no AI magic required. Given, this would require individual attention and problem-solving on a per-engine (and maybe even per-game) basis, but fans are already doing that on a regular basis for popular games, and it would be worlds easier than trying to wrangle an AI or work with some kind of simulated photogrammetry.
The website in the video showing off reconstructed Nintendo worlds shows how viable this approach could be. The real opportunity as I see it is standardizing and automating what's already out there, and what has yet to be created in terms of ripping tools. Right now if you wanted pull the worlds from say, the new Spyro remake, that's totally doable, fans have created the tools to largely re-construct and pull it all back into an Unreal project. But it's also a nightmarish hodgepodge of "use this and then this and then follow these obscure steps and then-" etc. If someone took all those tools, wove them together with some scripting and slapped a nice UI on top so it was just "Rip game map" "Play game map" or whatever, there's the dream software.
This still takes real time and effort obviously, and it would have to happen in pockets: only games with a community devoted enough to build ripping tools and a plugin for this "universal" program would be accessible. But again, low hanging fruit. To steal something Ross said in the video: something is better than nothing. Ripping and reconstructing like this would be infinitely easier and more practically do-able than trying to mess with AI.
Just some thoughts though. Like I said, that's why these videos are great, they get you thinking. Here's another thought: if this software was created, it could potentially ride the "just barely legal" wave in the same way emulators do. You can't share the maps, but that doesn't matter so much if anyone can create the maps themselves, at home, as long as they have both the software and the game.
I generally just assumed this method would be a never-ending hell with standards changing all the time, encrypted files, and only the most popular games getting support. Even games that use the same engine can have massive compatibility differences. A ripping-solution from the output would be something that would only get refined over time and not have to keep reinventing the wheel, but I'm happy for whatever gets us more worlds.