tl;dr I'm remastering NWN2 OC and MotB from scratch.Not sure that it's the right place to post, but what the hell.I thought that NWN2:EE was lacklustre. It hardly fixed anything except for maybe Multiplayer and a few bugs. This doesn't look like $30 worth of work.Then I started wondering: what would an actual decent remaster look like?My first thought was using my modpack I created and debugged back in the day: better UI, custom portraits, better character sheet, better inventory, better companion AI, a bit of a facelift when it comes to textures and so on.But that clearly isn't enough.The big part of NWN2 is the original campaign, but it's buggy as hell and has a lot of cut content.IMO the absolute minimum for the remaster is to fix all of the known bugs in the OC. And it wasn't done.The next step is to fix some of the glaring campaign issues that are technically not bugs: for example, some of the influence gates are impossible to clear without the debug menu and your companion dialogue tree is static: you can ask one of your companions "Who are you?" in the final dungeon.Then the AI should be reworked. In fact, there is a must-have AI mod that does just that: your companions and your enemies are no longer dumber than bricks and you have a DA:O-like tactical menu for fine-tuning your companions.It should've been integrated into the main game, but alas.Add all that and I see myself paying $30.So I decided to remaster it myself. If I, one idiot with an LLM, can do it, then Aspyr is officially incompetent.
So far I've gathered 300+ bugs, built a vision-based harness for automated testing and fixed a few bugs.My goal is to make this number provably zero, at least all of the parts that don't require patching the engine. I've already decompiled it, but that would be the next big project.
I think we need to coin terminology for a very peculiar variant of AI psychosis, one in which clueless people start thinking they're programming gods capable of tackling tasks far beyond their means. A type of inflated hubris.I can guarantee that even with good harnesses, oversight, prompting and so on, you by yourself+Claude isn't enough to remaster this without also introducing quite a few new bugs and issues.
>>4141740It's okay to be confident.
>>4141740Thing is, this project has lived rent-free in my head for years at this point, predating generative AI.It's just that now I have no excuse not to start it.>you by yourself+Claude isn't enough to remaster this without also introducing quite a few new bugs and issuesWe'll see. The first step is to create an automation framework for testing. I wouldn't say it's easy, but I'm making good progress.Also I doubt I'll introduce many regression bugs until I start tinkering with the engine itself.>>4141772I'm not especially confident, but the worst-case scenario is that nothing comes of it.
>>4141740AI will do these thingshttps://www.youtube.com/watch?v=zxMZyVn0y3k&t=75s
>>4141845AI indeed will, one day, but that day is absolutely not today. And if the tidal wave of slop has been any indication, even when AI can do all these things, taste (which most people seem to lack) will still be a barrier.
>>4141859This project requires almost no creativity from me.Improving companion AI was already done via the mod, I just need to include is seamlessly.Reworking dialogue trees into the MotB-style is also pretty straightforward. I probably need to think a bit more about the impossible influence gates, but that also looks doable.The rest is just a mechanical work.
>>4141157So what are you exactly doing? A reimplementation of the engine to use all existing resources, no new data?
>>4142222First of all, I want to fix all of the non-engine bugs here: >>4141158Needless to say, it's impossible to do in any reasonable amount of time without leaning on the automation really heavily, so I'm currently creating an auto testing framework. It's a pain in the ass, but I've made some progress: I can reliably fire up up to 10 NWN2 instances on demand and I can do some light automation with the limited vision to test end-to-end scenarios. It was funny to see that the 9070xt started actively cooling itself only after I'd shoved 10 NWN2 instances into it.Then I'll apply some of the fixes that I mentioned in the OP-post.Then I need my automation engine to play the whole game end-to-end.At that point, I'll have a testing engine robust enough to catch a substantial number of regression bugs which will inevitably appear in the process of fixing the engine.As for the assets, I assume that the NWN2:EE team upscaled them using AI(most likely), which drastically increased loading times and didn't improve their looks. I actually consider that to be a regression. Maybe I'll ship a small curated list of visual mods that provably don't collide with one another. I've seen some conflicts, the funniest was Safiya growing an ugly hairdo when I installed a floor tile mod or something like that.I've created a small one-pager to track the progress:https://nwn2reassembled.fyi/
I also want to restore some content, but that's so far off right now that it's not worth talking about.
>>4142287>Then I need my automation engine to play the whole game end-to-end.>At that point, I'll have a testing engine robust enough to catch a substantial number of regression bugsThis is how you handle the elementary QA in anything more complex than a shoot'em up, at least when you aren't a braindead suit who thinks a bunch of monkeys can test everything million times a month for a handful of pennies.Black Isle, Bethesda, Bioware, Troika and Obsidian were known for abysmal QA. The wtfs found in game resources by fans patching Planescape Torment and Fallout speak for themselves.CRPGs, especially those with a lot of C&C, are not just "games" where clipping through the obstacles or inaccessible locations would be the worst thing that could happen. It's a complex logical system, a large finite state machine which can't be properly maintained without an automated system.In the ideal world your automation wouldn't even start with actually "playing" the game to find a regression, it would be reserved for late stage testing, a very naive approach outside of it. It would start with grinding through the game at the level of scripts and data, or in a special dev mode.
>>4142303>Black Isle, Bethesda, Bioware, Troika and Obsidian were known for abysmal QA. The wtfs found in game resources by fans patching Planescape Torment and Fallout speak for themselves.To their defence, everything was jank back in the day.Also the hardware is an issue too: I have several supercomputers at home by the standards of 2004.>In the ideal world your automation wouldn't even start with actually "playing" the game to find a regression, it would be reserved for late stage testing, a very naive approach outside of it. It would start with grinding through the game at the level of scripts and data, or in a special dev mode.I do that too and I test everything that I can on a lowest level possible.Issue is, an NWN2 save embeds a snapshot of the module - areas, object/instance state, local variables, globals - into its saves, so the area/module/creature changes don't apply to the existing ones.So I need to actually "play" the game to do the robust A/B in many scenarios.But loose script and dialog resources still load live, so override fixes to those apply immediately.
>>4141845We have iwd and nwn in bg, g in nwn 1 and 2 and is iwd in nwn? I know bg is planned as mod after path menzo in bg3 and some iwd2ee guys wanted iwd2 in bg3, are there other crossovers? Bg1, 2 and 4 are low key announced who will do it, cdpr, their bitch offspring or owlcucks? All three suck btw. >>4141157 I'd rather let him and Claude make bg4.
>>4142318>I'd rather let him and Claude make bg4
I set up the pipeline to post e2e A/B bug videos:https://www.youtube.com/watch?v=HUmhlni8eAcThe default youtube recommendation feed is a horror show.
Fixed some of the easier bugs and added counters.https://nwn2reassembled.fyi/It's still far from a whole bug corpus, but if I can't handle these, I 100% can't handle the trickier ones.
>>4141158I honestly don't give a fucking fuck about anything but Mystic Theurge prestige class. Get that implemented and working correctly.Preferably by implementing it as a 20 level base class like the Kobold Press one for PF1e: https://www.d20pfsrd.com/classes/3rd-party-classes/kobold-press-open-design/theurge/I do not fucking care about just about anything else because I'm not interested in playing any other class. And since Kaedrin is such a gaping yeasty cunt, the enhanced edition still doesn't have a functional MT and no one else has been updating still. There were like three different versions of the class created by mods for the old version of NWN2 and yet the enhanced edition remains unplayable not for any other reason but that the mods just don't work anymore.If you get the item crafting / enchanting working with quality of life features, correct pricing and more enchantment variety, then that'd be nice, too.But still. I have zero interest in NWN2 because I can use toybox to gestalt in Kingmaker or Wrath instead.
>>4143522Thanks for the feedback.I'll file Mystic Theurge, but restoring this class isn't the first priority right now.>NWN2 and yet the enhanced edition remains unplayable not for any other reason but that the mods just don't work anymore.I'm basing my version onhttps://www.gog.com/en/game/neverwinter_nights_2_completebecause, frankly, fuck the EE. I do not consider putting in this much effort and then having the gall to ask $30 for it to be acceptable.
>>4143522>>4143533https://nwn2reassembled.fyi/#downloadhttps://www.youtube.com/watch?v=NulQhS6IpNcbut that's still that needs to be properly play-tested, which is very far-off.I cannot ever recommend using it as it is right now.
>>4143508Are you saying you'll be finished with a remaster in like 80h or so? This means gen8 ai will bury gaming studios, good, fuck them. Imbeciles.
>>4143508You are going to end up like Cleve Blakemoore "polishing" his Grimoire in 2014.>only 123 issues and 4134 microissues left!
>>4143730>like 80h or solol. lmao.My optimistic estimate is two months *before* fixing the engine.I only posted a fraction of the bugs. The main issue is robust automation, and it's hard.Also keep in mind that a lot of work was done for me by the community, I just compiled the data.>>4143756Could be. Again, the worst case scenario is that nothing comes of it.
>>4143820That makes much more sense
