Interior design

[Bambooza]

Wee this is not even scope creep this is full on scope swell.

View: https://youtu.be/l_ZW6zxYwmk



I mean don't get me wrong I dig the idea and love the expanse of the possibilities. But like planet tech changed the game this too has the potential of significant stalling features.

 

[Vavrik]

I think I agree with you here. I don't mind them thinking about this, something has to be done with some of the cities. But not at the expense of current deliverables and not before gameplay related deliverables.

 

[Lorddarthvik]

Agreed, this is something on the "nice to have" list, not the "must be done" list. I do want this to be a thing, truly, but it's fine if we only get this 2-3 years after release (I'll be dead by then lol) as a big update.

I only watched a 2 min breakdown of the vid so I might be repeating things they already said.

For those who don't pay attention to these vijeos or just forgot, this has been a thing since forever.
The OG pitch from CR was that every building will have some sort of interior (obviously primitive, generated, not handmade) where missions can take place and we can hang out and the usual fever dream stuff...
Later when we got Lorville, nothing was said at first, like it wasn't ever a thing. But I distinctly remember that at the Area 18 release the topic got revisited. Then the pitch was that things like bounty missions will be generated as usual but those will temporarily generate an interior for the building the mission is played in, so the buildings won't just be empty boxes to get crashed into by the autopilot.
This didn't happen either.
Around this Citcon (2022) we got the word that Lorville will get an update. Great, who gives a f... anyways, the point is that we have seen the tools they use, the workflow for making the buildings and such. It's really simple and does NOT allow for interiors with the current exterior modules they use. And by current I mean the ones they spent at least half a year work on at this point for the updated Lorville!
The simple fact is that ALL non-enterable building modules will need to be reworked. Remodeled, retextured, adding real windows, real doors... a bunch of stuff. Also they will have to make a new clipping solution for the engine. Imagine having to load in and render ALL the buildings with interiors in Area 18 with windows you can see through! Yeah, that would crash even my system with 128gb ram.
Aaaand we haven't even gotten to the issue of floor level. What happens there? Can you walk out onto the streets? Then we will need roads! That's something that has been in the works for god knows how long, and the last time we heard about it was the one guy who did the river building demo mentioning that yeah, it's still a thing, using the same tech as the rivers, but nowhere near as usable as the river tool so it will be ages till it's done.

TLDR.: currently non of the assets support what they want to do with them, and neither does the engine. This is a great goal to have, on a long term basis. But not for 4.0, unless they already know that 4.0 won't be out for at least 3-5 more years.

 

[Pistol]

I long ago renamed the RSI/Star Citizen icon on my desktop to "Space Matrix"

 

[marctek]

I think making this update will be way more than just interiors. They will need to add city streets, bridges, and such as well for us to access the buildings. Then they will need to spawn NPC to populate and make those areas more alive. I know some of the current tech can be reused for this but I'm not sure they aren't creating another performance issue that will take a long time to fix on top of the new tech needed to populate the buildings.

But a giant city to wander around and get lost in would be fun.

 

[Sky Captain]

Agreed, this is something on the "nice to have" list, not the "must be done" list. I do want this to be a thing, truly, but it's fine if we only get this 2-3 years after release (I'll be dead by then lol) as a big update.

Great point. Do we have clarity on the 'must have' versus 'would be nice' development features list for Star Citizen at this point? As a long time backer, understanding what CIG sees as minimal viable product vs. being possible within their timeframe and budget would be good to understand better at this point 10+ years into development.

 

[Lorddarthvik]

Great point. Do we have clarity on the 'must have' versus 'would be nice' development features list for Star Citizen at this point? As a long time backer, understanding what CIG sees as minimal viable product vs. being possible within their timeframe and budget would be good to understand better at this point 10+ years into development.

It would be awesome if we had any sort of clarity in this matter, but as far as I can see, they don't have any either. It's just "what's the next thing we can rework or add" and they go with it. This nature of not having any definition of "minimal viable" is what's been bothering me the most 10 years in... They just move the goalpost every time it even starts coming into view, because they can. I think we will only get a definition when they Have To work towards an actual release because we stopped giving them money and they are about to go bankrupt. Until then, nothing will change.

 

[BUTUZ]

but I'm not sure they aren't creating another performance issue that will take a long time to fix on top of the new tech needed

Congratz you just summarised Star Citizen.

 

[Vavrik]

Congratz you just summarised Star Citizen.

I just want to point out that most of us in the PTU are noticing some performance improvements. Not all, and not everywhere but they are there.

 

[Han Burgundy]

Mark my words, girls and boys; This is a task that will be handed off (largely) to A.I. within the next 5 years. This is the kind of thinking, modular set-pieces that is, that will make iteration via Artificial intelligence an attainable thing. They build out, by old fashioned human hand, several versions of modular pieces that have different aesthetics but identical function. The A.I. then chews on those different variants and combines them to create new ones of its own. Not only do you have architecture; You have decoration, items, loot, notes, NPCs, and any other gameplay element that can be packaged as an item or scripted event. The combinations are endless and will need to be if we are to be kept entertained for decades to come. Classic procedural generation is all well and good, but its big brother will quickly become purpose-built A.I. that is not simply smashing together item-ports of a similar value in repeating blocks, but actually building narrative with the items chosen to populate each space with a meaningful story behind it. A new tide of honest-to-god endless gameplay is now technologically feasable and somebody, somewhere, is going to capitalize on it. Here's hoping it's Chris.

/$0.02

 

[Bambooza]

Mark my words, girls and boys; This is a task that will be handed off (largely) to A.I. within the next 5 years. This is the kind of thinking, modular set-pieces that is, that will make iteration via Artificial intelligence an attainable thing. They build out, by old fashioned human hand, several versions of modular pieces that have different aesthetics but identical function. The A.I. then chews on those different variants and combines them to create new ones of its own. Not only do you have architecture; You have decoration, items, loot, notes, NPCs, and any other gameplay element that can be packaged as an item or scripted event. The combinations are endless and will need to be if we are to be kept entertained for decades to come. Classic procedural generation is all well and good, but its big brother will quickly become purpose-built A.I. that is not simply smashing together item-ports of a similar value in repeating blocks, but actually building narrative with the items chosen to populate each space with a meaningful story behind it. A new tide of honest-to-god endless gameplay is now technologically feasable and somebody, somewhere, is going to capitalize on it. Here's hoping it's Chris.

/$0.02

Yes and no. The issue procedural generation saves is two parts. The amount of designers time crafting a level and the amount of disk space the level takes up.

While AI can reproduce and craft new assets and do a pretty good job of level design procedural generation is great in reducing the complex unique world into a much smaller number. Of course you can take it to far like the original launch of no mans sky with far to little assets game play loops and points of interest. While the world's were vast and unique there was nothing interesting to do and once the thrill of the vista wore of the game was dull. It's in many ways the same with star citizen in that graphicaly it's beautiful the moons and planets are fun to explore but outside of that there currently is not a lot of game play loops to engage in. More than more empty space or in this case empty lifeless rooms in lifeless buildings. I have to wonder if the purpose is forgotten that set dressing should always play second fiddle to the character and stories. And while multi-player games allow for new stories to be crafted its the game play loops and not the stage that truly facilitate that. And so I'm left puzzled by why so much resources are reported on the stage dressing and not on the game loops. Is there something blocking them or are they being utilized heavily in the dark void of sq42 or has SC turned into a tech demo void of fame elements

 

[Han Burgundy]

I wholeheartedly agree that current utilization of procedural generation leaves room for improvement, but that's exactly what I mean. Truly A.I. crafted spaces are not, technically, procedurally generated. (Not in the same way) And i'd wager the reason they are spending so much time putting together the set-dressing is to facilitate their eventual plan to generate a building interior for a one-time mission that changes location and details each time it is generated. For that, you'd simply build missions based off of a few artist-led parameters (Enemy faction, equipment, reinforcements, ect) and tag it to a location that is currently unoccupied. That would make each rooftop in Area 18 is a potential mission site. A site that evaporates as soon as the mission is over and players leave the area. This ability will be paramount to developing a game that won't feel repetetive like NMS. You know what they say: A good set can bring any production to life.

 

[Bambooza]

I wholeheartedly agree that current utilization of procedural generation leaves room for improvement, but that's exactly what I mean. Truly A.I. crafted spaces are not, technically, procedurally generated. (Not in the same way) And i'd wager the reason they are spending so much time putting together the set-dressing is to facilitate their eventual plan to generate a building interior for a one-time mission that changes location and details each time it is generated. For that, you'd simply build missions based off of a few artist-led parameters (Enemy faction, equipment, reinforcements, ect) and tag it to a location that is currently unoccupied. That would make each rooftop in Area 18 is a potential mission site. A site that evaporates as soon as the mission is over and players leave the area. This ability will be paramount to developing a game that won't feel repetetive like NMS. You know what they say: A good set can bring any production to life.

I think there is some confusion over terms and or point.

For me, procedural generation is like you said crafting one-time mission locations using reusable object blocks like legos to craft slightly unique locations. This can be as vast as whole planets and solar systems as seen in NMS and Elite down to what Star Citizen is attempting to accomplish by reusing room segments to craft rooms in builds, outposts, and space station layouts. The uniqueness of each location is going to be bound by the number of available blocks (rooms). On the upper end each block can be used only once and on the lower end only one block is used everywhere.

A.I. crafted spaces for me is not so much the utilization of these blocks (while it is still technically the same thing) as that is more just taking a seed value and running it through procedural code to craft a location, with each location having its own unique seed value. A.I, crafting is more typically referenced in the creation of those blocks, like the creation of the artwork, and requires a far more complex code to accomplish as its trained and then utilized to craft new things. These new blocks typically are not crafted each time you need them as the computational cost and data required to generate the outcome are far too high to do on personal computers in real-time. So what would typically be done at the moment is to use A.I. to create blocks and then saved the generated result. This is the same as what game artists currently do and so it's easily comparable.

My point was that the room blocks have their own costs in the way of bytes in both storage on a player's hard drive, RAM, and VRAM and are thus bounded by the player bases computer capabilities. Sometimes you can get around these limitations by streaming to the player over the network the object but this too has limitations. So while A.I. crafted set pieces can be great in allowing for far more blocks to be utilized by procedural generation it is still bound like artistically created blocks by the intended player's computer capabilities. Where procedural generation can reuse those blocks to create many different unique combinations it does not increase the feel of uniqueness, (it can be argued, and perhaps rightfully so that most subdivisions suffer from the same blandness, even if you factor in micro-changes over time).

What would be interesting to see is taking procedural generation to the room level and having the seed value craft the room itself and everything in it. While it could be argued that even here it would be the same assets reused ad nauseam, it's far more believable in that a lot of our items are mass-produced as it is. What I am not sure about is if there would be any storage cost saving by storing the seed value over the current room block.

 

[Han Burgundy]

What would be interesting to see is taking procedural generation to the room level and having the seed value craft the room itself and everything in it.

This right here is exactly what I was getting at. A shelf is an object that contains a certain amount of "item port" space where set-dressing objects can be produced. These objects are physically modeled by artists. A spray bottle looks like a spray bottle, but several different textures can be generated to coat the outside of that model (Varying tag-lines, colors, products, ect) The A.I. system could then gather several storytelling elements to attach to the space (Poverty/Swanky, Legal/Criminal, Some value in between) and spawn objects with skinsets that fit into those chosen set-piece categories to fill that shelf space. The A.I.'s job is to read the values appended to the "Story" of the space and craft something believable.

I don't know a whole lot about the technical side of streaming things in and out. I just know how good A.I. driven iteration is getting, given a proper training set. Is there something ready-made on the market right now that can do this for them? Nah, not really. Do I think CIG is going to sink funds into developing their OWN A.I.? God I hope not. Do I think that pre-built A.I. design assistance programs are coming to market in the nearish future? Oh yeah. And Chris likes him some new CUTTING EDGE bleeding future tech, so I suspect it would be difficult for him to resist the new shiny. Gota fill those hundred star systems, right?