Heaven's Vault: Creating a Dynamic Detective Story
Вставка
- Опубліковано 5 лис 2019
- In this 2018 GDC talk, inkle's Jon Ingold outline the strategies inkle has to create the sci-fi detective game Heaven's Vault.
Register for GDC: ubm.io/341ZiaZ
Join the GDC mailing list: www.gdconf.com/subscribe
Follow GDC on Twitter: / official_gdc
GDC talks cover a range of developmental topics including game design, programming, audio, visual arts, business management, production, online games, and much more. We post a fresh GDC video every day. Subscribe to the channel to stay on top of regular updates, and check out GDC Vault for thousands of more in-depth talks from our archives.
I've spent the last year thinking about the same stuff, but this guy is light years ahead of me. A very valuable watch.
Jon is light years of just everybody else.
For anyone who is interested in emergent storytelling for sure.
Easily one of the best GDC talk i've ever seen. I have a new favourite designer now. I would check those other talks, but i want to play the games first
Daniele Mazzali his company had a short lived podcast called Inklecast that was interesting
@@michaelswietek8211
That's awsome, i'm definitely check it out, thanks
Same, these talks actually got me to purchase some of the Inkle games.
Reminds me of GMing tabletop. The players muck about on the scene, and when they do something I think is clever I reward it. If they seem bored I make up a reason to move on. They're simulating verisimilitude in place of realism, because realism is resulting in boredom.
I watch these talks mostly for GMing 😛
The point about LA Noire for the creation of LA with nothing to do in it is a great point. A crazy amount of effort went into effectively making a stage for car chases and driving around.
Is that a complaint? Honestly, the banter between detectives as you go from place to place might not have been possible without a place to drive around.
It is bad in terms of cost (and stress). Deadly Premonition did the same with lower fidelity (i.e. cost) and achieved the same experiential function.
This dude is my favorite GDC speaker. A free-thinker with inspiring ideas and a great sense of humour. I deeply enjoy his snarky comments about other games. 😂
The line about the conservation of narrative momentum at the end is pure gold. It's one of the biggest plagues of narrative games right now
11:48 • “Neither of those games are charting here.”
I guess you could say they’re...Uncharted.
But… but Uncharted was charted
@@nintendude794 • Shhhh, just let us enjoy the pun, haha.
I played the game and I had very mixed feelings about it. I LOVED it on one side and felt somehow frustrated aboout it on the other. Having seen this presentation i can now see the developer's message and undersood that what was frustrating me was the gamer's logic i was trying to apply to Heaven's Vault! I really loved this speech and i TOTALLY share its content!! Good work guys! You made
an EXCELLENT job!
I was so sucked into the talk that by the end I was impatient for the game to come out "later next year" as he was saying. Your comment made me realise it's "3 years later" and also got me intrigued on the actual reception. Rushed on Steam, and reading through the reviews, it feels like a hidden level to add to this talk. Some of the assumptions in the talk are key points of player criticism (locking them out of revisiting, pushing them out of a level after there's nothing left to discover, etc). It all looks like a beautiful mess and there's a lot to unfold and learn from. Will definitely give it a spin and I totally get how confusing it can feel without being presented with the novel underlying system that it's all based on.
@@RobertMaximVerdes definelty worth a try! i'm glad my comment was somehow useful!
I finished Heaven's Vault and while there were some aspects of the narrative that really didn't resonate with me, notably the ending, I found the immersive experience to be absolutely impeccable. Sometimes in real life, you mean to say something then the conversation moves along and it's forgotten forever, or you say something you regret and you just have to make do. I felt like I was truly inhabiting Aliya. I wish she wasn't so hard on Six, though.
My guess was that the closest to freedom you can get is "What do I WANT to do now?", given the previous category is asking what a player CAN do (asking about the possibility space). The issue being that, for the most part, only pencil and paper RPGs can actually achieve that level of player freedom, as there are virtually no limits on what can be attempted and you can create "content" on the spot. This kind of freedom is essentially what video games have been trying to achieve and, quite hilariously, boasted to have achieved ("Be anyone, go anywhere!") in the past. I'm wondering if it's a goal worth striving towards given it might not be possible, and whether games should instead embrace their limitations... but "Yikes what have I done?" is an interesting way to flip that question on its head a bit and put how games are played front and center, so it might actually be the answer to some extent.
Well, we already see it in machine learning already. As a system becomes too complex to fully grok, it seems to develop interesting and emergent properties that surprise and delight us.
That's really hard to do in a reasonable time frame using strictly linear techniques that flow seamlessly, as opposed to very short time frames using exponential techniques that flow like a pile of rocks. Their studio's approach seems to bridge the two, somewhat, and allow for ironing out the kinks with repeated simulation and monitoring. Which I think renders the answer as yes.
Ultimately, this game was not on my radar until this talk. Now I really want to play it to see how they are able to execute on their ideas. While a little bit of the detective oriented magic will be missing from the experience by understanding that the game can functionally play itself, it doesn't decrease my desire to experience the narrative-as long as the hook is good. And because of how reactive the game is, as described, I'm terribly interested in multiple playthroughs.
@@gogauze yeah i’m super interested to see what happens to games once AI is able to invent experiences on the fly. We’ll be in that “D&D with a great DM” situation… but with a live onscreen image rather than descriptions painting pictures in the minds eye.
Honestly this whole topic just makes me want to train GPT3 to DM. 😂
@brainswashedthisways it probably could pretty easily already. I've been using it to help write a ttrpg, and after feeding it my ruleset it could accurately create NPCs, items etc. as well as talk about areas I need to develop further.
@@paultapping9510 We tried it yesterday, it still has some issues, and a lot of them are linked to how "compliant" it is, it is really hard to have it make decisions that could harm the player, or even move the story forwards in a definitive way.
The first time we tried to have it simulate a single turn of combat, it told us we rolled a 14, and that if it was a success, this happens and if it was a failure, that happens, we had to remind it that it wasn't our role, but it to tell us whether or not 14 was enough to hit.
We then had it DM a short, classical dungeon, we purposefully took a path that was supposed to collapse, but it didn't, chatGPT apparently replicated the mood and features of the dungeon, but it had no notion of the actual structure of it, of the order in which things are supposed to occur (which isn't that bad, we just wanted to see what it was capable of). No matter what we did, traps would never entrap us, and when we fast forwarded to the big fight against the final boss, interesting things happened. It was really good at taking any intricated idea of attack we could come up with and describing the outcome to us, but no matter what we did, our attacks never really failed (at worst, it reminded us that since the lich is undead, strangling him with our hair would only distract him and restrain his movement but not actively damage him), while it repeated tirelessly how cruel, powerful and deadly the lich was, he never attacked our character in any way, and despite chatGPT's description of the state of the lich declining with each hit, it went on for a very long time without it dying. We felt trapped in an endless fight, and since the lich never attacked us, we figured he perharps wanted to be friends, so we stopped attacking and tried making peace, action after action, the lich went from surprised to intrigued to seriously considering our offer to become business partners and open a French-fry truck/chip stand together (we grew desperate, we really wanted this encounter to end, in any way), but chatGPT wouldn't make a firm decision. In the end, we tried to ask it to "move 10 minutes forward", which failed (the lich was still contemplating his options in life), "move 5 years forward" did the trick, but that was tedious.
Maybe by forcing it to write all the mechanical aspects each turn and to unveil all the typically partly hidden combat mechanisms we would have had better results ? I don't know. But on its own, it really has trouble making decisions that could "harm" the player, or remove any agency from the human operator, so it doesn't make for a good GM just yet. It was interesting to play, though, and it wrote a really great scene at the end, but the novelty of playing with chatGPT was a big part of the appeal, we wouldn't have been as forgiving to a human GM.
If James Veitch was a game dev
Never thought about Heaven's Vault as unfair, but now that he said it, can't help but agree. And I feel weirdly ok with that!
This is a really interesting approach to making narrative games!
And then in the middle of this, they created this world that feels real and lived in
46:33 woops.. I err, may have done a lot of that. It felt like the thing to do, I realised I could scavenge clues and went out scavenging! I wonder if you can end up completing the game then without ever entering an "episode"..
Also he was wrong haha, I loved it! Although I probably did a bit more of this grindy clue hunting than he probably wanted..
Me too! It's an pretty obscure kind of problem the game's designers didn't really think about or couldn't afford to do anything about, and I think it arose for me because:
*1)* I'm an absolute completionist maniac. If the game generates clues for me to pick, then I'll go out of my way to roomba the whole map and make sure I've picked every single clue in existence. I'm not having much fun doing it, this game is absolutely not meant to be played this way. Players like me are a small minority but a pretty vocal and active one (making the guides and walkthroughs), and our playstyle intentionally goes against any common sense and expectations of game designers. It's pretty hard to incorporate such players into the general scheme of things.
*2)* In this game, procedural leads to new locations have translatable text on them, and in my case that turned into a problem. Players only *need* a limited number of leads, and the procedural generator is designed to give them just the necessary minimum. But the translation is the main gimmick and selling point of the game, attracting players who *want* as much of that as possible (I certainly did because I'm fascinated with the language stuff), and procedural generation gives a potentially unlimited amount of new text to translate. So as a result, we have a gameplay mechanic being used in a way it wasn't designed to be used, and the game experience ends up not as good as it could have been.
@@haunter6682 I have some tendencies towards (1), I'm curious whether you have some game recommendations where that play style felt really rewarding
@@IAMED2 For me, playing like this feels really good in games with handcrafted (not procedurally generated) levels and maps with lots of little secret areas with collectables. I really enjoyed Legend of Grimrock 1 and 2, with their cleverly designed dungeon layouts with puzzles and secret doors, and Tunic, where stuff is hidden literally everywhere but unseen because of your fixed camera angle. Also, weirdly enough, I find that the Dark Souls series is doing it for me with its level design.
those authoring patterns are super interesting. Very similar to how ttrpgs can be approached (and I prefer to approach them):
moment>room>dungeon>scenario>campaign
So good!
Really liked the talk.
I'm not surprised anymore of how I loved 80 days! The game was really built ground-up to be that engaging.
Now I'd like 2 more talks: 1 about that fictional language, 1 about that time you sneaked past the Pooh-wall. :D
The first I can oblige you with: ua-cam.com/video/e1L1qRmYgK4/v-deo.html
Cool!
Hecking masterclass!
Brilliant.
These are interesting concepts but I don't see the connection to Heaven's Vault, a textbook example of a "what am I supposed to do now" game. Seems like the developer cared more about the experience of making the game than the experience of trying to play it. And I say this as someone who liked the game overall.
I got that "what now" only once so far. And to the credit of the game I defaulted to roleplaying: well I the player am out of ideas. What would Aliya do? Well it's been ages since she talked to the professor, and I skipped the funeral... Let's check in on that, and stop by the librarian friend to check translations.
I got a couple of *very good* interesting scenes and bam, 3 locations to check in the world.
I think that experience is exactly what his game is built for. But he fights game logic and that is a roadblock if that's your standard operating procedure :)
This architecture can be summed up that a random walk completes the game, essentially it doesn't matter what the player does just that they do something.
That's pretty much how story-based games work.
how is a nonlinear walk worse than a linear walk?
It matters in that the story of *your* walk will be unique due to *your* choices, which is the stated goal of the project. He's very clear about it not being a strategy challenge.
17:30 I don't know. I think a Game like Braid has to be up there, or Portal 2.
Great talk. Love the Sorcery series, but HATED Sorcery 3, and so got put off and didn't play Sorcery 4. ! and 2 were great though.
I disagree about Horizon Zero Dawn and The Witcher on chart at 11:00. Witcher offers bigger freedom of decisions than HZD, and game structure is so good, decisions made in previous games have impact on gameplay.
The end of the Bloody Baron Questline is a great "What have I done" moment.
@@gormanls Ye. Theres plenty of moment where you can only choose "lesser evil". Which makes Witcher a great game.
It seems like he was talking about the machine-hunting game of HZD (lots of player freedom) as opposed to the narrative game of HZD (basically linear with no meaningful choices).
I'd say that's one of the biggest flaws with HZD, and with a lot of single-player AAA games: it's two games stapled together, and one's kind of more interesting than the other.
They made an exploration game and did everything in their power to prevent you from exploring. They made a game where choices matter and did everything in their power to prevent you from knowing what those choices were. A single misclick can lock you out from a location FOREVER and they force you in a single auto-save, havens forbid you want to take a different choice without replaying the whole game. They had a gem and covered it in a ton of manure: most of your time playing is spent digging through shit in order to catch a glimpse of that gem. 30 hours in, about half of the game, I decided I had enough. What a waste of an exceptional concept...
Those are games honey.
He says a lot of things that appear to make sense, but in the end they made a game where exploration is not good, and story is told in a tedious way