This would make a cool art installation. If you can get access to one of those giant screens that you see demo'd at E3. You can visualize the entire Land of Hyrule from Lengend of Zelda. With a tiny Link navigating his way through the world ;)
Even after having seen _a lot_ of (S)NES hackery, and being quite a bit into reverse engineering the consoles and games, this is mind blowing to me. The result is just so perfect. It makes me happy to see that decades after the release of these, there are still people playing with them, in one way or another.
Which brings to my mind an interesting question: Could a ml agent be used to auto-explore with the intention of visiting all parts of the game world?<p>There has been recent focus on ML research trying to win or beat games. What if the goal was merely to have an agent that would manage to explore the largest game "area" possible? Has that been done before?
If you want to see some of the wacky things devs did with NES nametables and scroll registers, check out the Rad Racer pseudo-3D implementation: <a href="http://forums.nesdev.com/viewtopic.php?t=8588" rel="nofollow">http://forums.nesdev.com/viewtopic.php?t=8588</a>
Curious to see how it would handle Dragon Warrior 1 - I bet the dark dungeons probably don't use the scroll registers at all.<p>Also Ultima III: Exodus would be interesting. I suspect it's "black out tiles behind walls and in trees" effect is done through the attribute table but don't really know what it does when elements scroll offscreen.
I almost gave up on this after seeing the first gif thinking that was going to be it, but I couldn’t have been more wrong.<p>This is a very neat trick and it seems like it works really well, I’m kinda surprised I’ve not seen anything like it before.
Quoting the statement on the page:<p>"In the mid 1980s, [...]. Boasting the best sound, the best graphics, and the best games of any home-console to date, it pushed the envelope for what home-gaming could be."<p>I challenge it as a big, fat lie! Yes, the games might be really great but the best sound and the best graphics in the mid 80s were offered by Amiga (released in July'85 while NES was released in NA in Oct'85.) I know that Amiga has been released as a home computer but originally it's been intended to be game console. So here are my two cents.
This looks really impressive!<p>I wonder if this could snapshot the state of the game, use spare CPU capacity (including other cores) to simulate in the background and feed in various inputs, and try to give a preview of the upcoming map? It wouldn't have to get very far; trying a few common input patterns would suffice.<p>I also wonder if this could apply some heuristics to see if the sprite layer scrolls consistently with the background, and if so, provide a preview of upcoming sprites too. If they consistently spawn at the same point, render them at that point.
Why not use these methods to map out a game, create a meta file with the data, and then allow it to be preseeded to improve performance, show the extended screen prior to loading on future playthroughs, and enable game specific heuristics only for the initial mapping, which could then be removed to avoid performance and compatibility hits?