Just think of a world that did not have procedurally generated content:

No factories, no assembly lines, no microchips, no computer generated graphics, no automated anything.

If there was no factory and no procedurally generated microchip, we did not have a computer to use and brag about what we did not like.

If there was no procedurally generated 3d graphics, there were no elder scrolls, and no other 3d game.

Because of this automation, we can now play games that a while ago seemed impossible, so let's not say blindly "no" to further automation of game mechanics, without thinking about the possibilities.

I do not say that the current methods will become completely extinct, because we still play 2d games, but I'm saying the future for blockbuster games would use a lot more procedurally generated game content, almost in any aspect of game design.

And I do not say that the future games would be random, because today's 3d games are not random, but the automation helps the game developers with mundane tasks and let them spend the saved time in any place that needs their focus the most.

And procedural landscapes or quest-lines do not mean all the places or quests look similar, generic or under-detailed, because those were the problems of the past implementations, not the future implementations, and we can create procedurally generated unique places or quest-lines, if we know how to produce them with seemingly unlimited variations that could change shape and adapt to any situation.

And above that all, those procedurally generated landscapes, populations, events, and quests would be just a base and foundation for the designers to start their work over them, and they could wipe an entire area, and start from scratch if needed, or add enough individually distinct features to any place, event or quest to make them unique enough.

So let's not veto an idea just because previous attempts were not that successful, as future is still to come and will surely see better technologies, and hardware horsepower.

I want to go into some details in the following text, but not too deep, as it would be too long to post here, but if someone wants more clarification on any part of the details, please ask me for it.

=== === === === === === === === === ===

Let me sketch for you a game with well-designed procedural content, and it can be done, as I already now a way to do that, but because those ideas are not implemented yet, I am not sure if current hardware can do the job, but at first I need to go into some technical detail about the procedure so that you understand how it happens, but if you do not want to read the details, just agree that it is possible, and jump to the next part.

First of all, do you know about a Photoshop plug-in that converts a picture into a fractal formula, and drastically reduces the size of that picture, because instead of lots of pixels that cover the entire area we have just a formula that defines the content of that area?

And if we did not need the picture to keep its entire detail exactly like before, and a rough estimation was enough for us, the formula could become definitely smaller, as fractal formulas tend to fill the area with their own detail, to the infinitely small specks of dust.

We can have landscapes generated by fractal formulas, and please do not think that it is not a good idea, because as I have said before, the future is not the past.

The entire area of landscape is inside, let's say, an object oriented landscape generator(LG) that uses fractal formulas and smaller more specialized landscape generators to create the landscape.

The smaller LGs are responsible for smaller areas inside the bigger one and they use the same method, i.e. they use their more specialized fractal formulas and smaller and more specialized LGs to create their content, and this procedure continues to the finest detail that are required.

The bigger LG can be called "Continent" and smaller ones can be Mountain, Jungle, Desert, Swamp, Valley, River, Ravine, Hill, Rock, Tree, Branch, Leaf, Town, Village, Settling, House, Cottage, Shack, Farm, Road, and so on... and those are just generic names, because we can have "cobblestone road" or "mud road" and so on...

Each LG can have a position and a boundary that defines its place on the map, and a random seed that affects the final shape of the landscape under its management, as the fractal formulas can have random elements, and the choice of sub-LGs under the LG that manages them, can also be affected by the random seed.

So if we have an LG covering a specific area, each time we give it a specific seed, we would get the same resulting landscape, so although it is generated procedurally and randomly, but with fixed random seeds, we would get fixed landscapes.

I do not want to go into more details, because it can rapidly become boring for non-geeky people, and in fact those LGs are "Identity Oriented" instead of "Object Oriented", but for now let's say that it is the same thing.

=== === === === === === === === === ===

The designers can start with a flat landscape and paints a rough height map over that landscape, like a painting, the black areas are the bottom of the sea, and the white parts are the tips of the mountains and the gray areas in between are the rest of the landscape. They can use semi-randomized fractal generators for assistance as well.

This is a detail-less rough sketch of the landscape, and at this point they can jump to the conversion phase or they can start to add more detail to the height map as they like, so in a 3d environment they can push or pull the land upward or downward, with specialized height map manipulation brushes, until they are satisfied with the general shape of the land, but the details would remain for the phases after the conversion.

After that they can place some high level landscape generators, like mountains and jungles over some specific areas to help the procedure of conversion phase a bit, and they can also place smaller detail, like valleys and ravines to be more specific, and they can draw the boundary between the sea and the land to make a precise shape for the continent, and so on...

Then they would start to mark places that they want more precision, and leave the rest to be either randomly generated or detailed with more relaxed attitude, as the resulting shape does not need to be the exact replica of the height map that was designed.

After that, in the conversion phase, they start a procedure that tries to build a rough estimation of that height map via fractal formulas, by juggling around the random seeds, using smaller and more specialized landscape generators(LGs) for the hard to create parts, and specifically defining the height of some points on the landscape, and so on...

In those fractal formulas, if we specifically define the height of a point in the landscape, the surrounding area would change shape to adapt to the new change, as it is the nature of those formulas.

This procedure would take its time, and if we want more exact rebuilt of the original height map with smaller resulting data, then it would require longer times, so let's say it might take between a few minutes to several days, depending on a lot of factors, one of which is chance.

The end result would be a hierarchy of LGs that define the shape of the landscape. Each with specific position, boundary and random seed, and maybe some specifically defined heights for specific points within their boundary. And that would be really small data compared to the height map, and also those LGs would know how to fill their managed land area with smaller detail and probable life.

If the designers have previously defined parts of the landscape, and marked them to be more precise when converted to LG hierarchy and fractal formulas, then those parts would be just like before, but with more detail as it is the nature of those fractal formulas.

But if they did not mark areas to be exactly like before, then those areas might have changed a bit to follow the formulas of the more precise areas, and those places might have some new details, as well.

=== === === === === === === === === ===

When the landscape is completely converted to LG hierarchy and fractal formulas, the designers can have more interesting tools to design the areas, for instance, they can move the landscape generators(LGs) around, and change their shapes and parameters.

They can place new LGs in places to add content or make the area more specialized, and so on... and the surrounding landscape would snap to the predefined areas of those LGs, and would change shape to adapt to new changes.

For instance you can place a shack over the uneven surface of a mountainous place and the surrounding area would snap to the predefined parts of the shack and change shape to adapt to the changes without losing the previous detail. I have tasted that with an old dos based program that I had written several years ago, and it works.

If a designer wants, he can loosen some snap points from an LG and that part of the landscape would revert to a previous state, depending on other snap points.

The designers can grab a point on the landscape and move that upward or downward and see the changes in the landscape, or clear some previously altered point on the ground and the landscape.

They can place a hut LG near a road LG and they would snap together and the road LG could extend a branch toward the hut to cover the entrance, or they can place a ravine near a valley with the same effect, and so on...

The "cobblestone road" LG would change the shape and the texture of its managed area to make that look like a cobblestone road, and would know how to add branches to new destinations, and the town LGs would use this functionality to make controlled but semi-randomized towns with houses that are linked by roads in between, but they can have somehow different heights, and so on...

The whole landscape can have this type of intelligent adaptability and mutual understanding of the surrounding LGs, and the character AI can use this characteristic to have more precise path finding or understanding of the surrounding area for decision making, but that is another subject altogether.

=== === === === === === === === === ===

The implementation of the landscape in the real time can be done by different methods, depending on the normal target user's hardware horsepower, so either all the landscape could be rendered in the real-time, from the Landscape generator(LG) hierarchy and fractal formulas, which would give the best flexibility and the ability for real time changes to the landscape, and randomly generated surface areas or natural looking (non-blocky), dungeons, seemingly infinite landscape with seemingly infinite detail, better compatibility and adaptability of expansions and mods, and so on...

The fact that both fractal formulas and LG hierarchy can be interpreted just to the required level of detail and stop as there are enough detail, can help in this subject a lot, so for nearby areas we go more into the depth of those hierarchy branches and fractal details, and for far away areas we stop as soon as there are enough details on the screen.

Or if the normal target users of our games/products do not have the horsepower to render such a landscape in the real time, we can have a pre-usage render of the landscape and store the details on the hard disc, and this procedure can be done each time the mod list of the game changes, so that the ground tears and landscape incompatibilities become something of the past.

For the randomly generated dungeons, this rendering phase can happen each time a player enters a new random dungeon and it would seem like a loading screen with a message that the dungeon is being generated and a percentage number and so on... and after that the dungeon can be saved in a place beside the save game and for that character's game it would become fixed.

But if the landscape could be generated and rendered in the real time, this waiting time would be a lot shorter, and there would also be another characteristic that would revolutionize the games: The surrounding area would change to have new features as the story proceeded, and we could have different seasons, falling leaves, new houses or razed cities, and subtle changes that would game a lot more interesting.

Like adapting to the game-play style of the players and sensing their immediate needs and supplying them without its being obvious, and so on...

The Randomly generated dungeons can be made by guidelines defined by designers, so that we can have an LG called "Random bandit mud cavern" which is a descendant of "Random Lived-in mud cavern" which follows "Random mud cavern" with follows "Random cavern" and random cavern knows how create inter connected cavern complexes, and random mud cave knows how to make such a complex with muddy effects, and random lived-in mud cavern knows what facilities are essential for a lived-in cavern and include them, and random bandit mud cavern knows how to manage a bandit camp residence and what is required for such a place.

That was the object oriented implementation of the idea, but the "Identity Oriented" implementation would be like this:

We place an entity in the game with the main identity of "Random Dungeon" and it would have the "Landscape Generator" identity as well, and then give it the additional "Mud cavern" identity which would have "Cavern complex" identity as well, and then we would add "Bandit hideout" to the entity and it would have the "Lived-in place" identity's facilities as well, and those identities would cooperate with each other and together create the random dungeon.

In fact with this method a place can have all the "Mountain", "Valley", "River", "Forrest" and "Lich King Territory" identities together and they would all leave their marks on that place, and shape it to be all those places in the same time.

As for the cavern example, the different bricks/units of the cave complex know how to stick to each other and to what other units they are compatible, the "Mud cavern" identity knows how to paint the surfaces, and add what artefacts to the surfaces, and how to populate areas with natural entities like rocks and holes and so on... and a hole when applied to a surface would change its shape to include a hole in there.

The lived-in identity would force the inclusion of living places, and how to stick those places to other parts, and also how to add stairways and generally make the place look like it has been lived-in, and so on...

The bandit hide-out identity knows how to add guard-rooms near the entrance, and how to add places to keep prisoners, and how to populate those living places with bed-rolls and general bandit staff, and how to populate the whole place with a living group of bandits, (by the help of the constant events and their schedules, but again, that's is another subject), and so on...

=== === === === === === === === === ===

I wanted to add details about how to populate an area with procedurally generated life over the available landscape and how to automatically give life to the added cottages and other LGs, and how to let the designers change those additions as they like, and how to add randomly generated or manually designed events to the available life, and how to develop randomly generated or manually designed quests over the available events, but that is too much detail and it was impossible to write all the stuff in one stand so I'm going to add those stuff in chunks and thus I need to add empty posts after this one.

What I write next, if any, might depend of the comments and feedback, as I'm versatile and if an idea clicks with the whole image, it could be included, or you can just tell me to stop right here and do not waste your times, and I would go elsewhere with these ideas.

If I change the poll and add items about procedurally generated encounters and quests, do the previous votes disappear? What if I add another question?

Edit: Please add your opinions about these matters, and I hope to continue with other procedural types of content.

I don't want to play a microchip, I want to play TES V

My problem with random generated content (if that is what you are talking about) is if it randomizes towns and dungeons as that would ruin the immersion for me. A town wil always look the same, any changes will be slow and take years. Especially in a pre-industrial world like Tamriel. I don't want to enter Chorrol with one character only to have it look quite different with another.

Events
Quests
Wildlife
Changes to the Landscape.

That's what, if anything, I'd like to see. I don't dislike a randomly generated game because of Daggerfall being bland in the wilderness, that's silly. It can work, and can be nice to not know exactly what's going on. But, it's like the difference between a photgraph and a painting. It's just more.. unique and taken care of. Hard to explain, but I just far prefer a hand-placed game.

If you had read my post, (I know it is long and hard to read) you would know that random places are just one of the results from procedurally generated landscape, and we can have procedurally generated landscape that is quite fixed, but for instance changes a bit to show the current season, or the sea tide, and so on...

Or we could have the main places, near the towns and important places fixed but have outlandish places that change shape in each play through with new challenges, and new random dungeons, to provide a bit more replay-ability, this way both parties are satisfied, we have fixed game core with hand placed detail, but some random places with random events and quests in the side places for the ones that like this type of content.

And believe me, if the core engine developed in a way to support this type of content, it would not be that hard to make games that satisfy all the different tastes and preferences.

I want the game to be completely hand built. By that I, of course, don't mean something silly like every blade of grass being placed by hand, but I want all items, static or not, in any location, from the smallest campground to the largest city, to be placed by hand. Procedurally generated lands and dungeons cannot completely replicate the feel of hand-placed items and hand-made dungeons. I don't even want content past what would normally be the game-world's borders to be procedurally generated as a "bonus." I don't want the quality of the world to be diminished by content that the devs didn't make by hand appearing past the game's borders. And no, I don't want procedurally generated events changing what the devs made by hand over the course of the game.

What? Re-read my post. I said these things are nice, but hand placed terrains just feel nicer, and are generally nicer to look at. (Human error may be present, but computer made perfection isn't always the best thing)

I prefer a large gameworld with most content being randomly generated with some hand-placed content than a small gameworld with all hand-placed content and no computer generated content. Mostly because there's more content in the first one.

Quality over quantity.

But I think your post has more too it, I'll have to comment again once I read the whole thing.

It may just be me, but quality>quantity.

I know randomly generated content can give good quality, but this doesn't seem a very good opinion. :shrug:

Let's simply answer your thread title: "Why do people hate procedurally generated content?"

And the answer is: Because people don't understand what it is.

A procedure is nothing more than a set of rules on how a given situation evolves. The rules of physics are nothing more than such a procedure; according to them, stars and planets form, and on planets like the Earth oceans and continents, and on them mountains and rivers and rocks and minerals and life. The whole of evolution is doing nothing more than procedurally generating more life out of whatever the starting conditions are. Almost everything on Earth is basically procedurally generated content. the Alps, Yellowstone, Amazonas - all procedurally generated by the laws of physics running for a long, long time.

In computer science, we're quickly approaching the capability where our procedural generation tools are very close to what nature does. Last-gen procedural terrain generation techniques basically involved taking a 3D model of the rock layers, with all the relevant physical properties, and then applying the force of wind and simulating billions of rain drops (each of them individually) to erode this and form rivers and lakes. Today, people are taking it a bit further and adding simulation of chemical and biological processes into the mix.

Sometimes, I think people hear "procedural terrain" and are thinking "Perlin noise" or some other old technique from last century. We left these primitive methods long behind us.

Personally I'd like to see beth use procedurally generated quests and events, and also take advantage of it to generate terrain (but not make it so large that they could not manually fix problems that something like SpeedTree might cause like floating trees).

For quests and events, I'd like to see them have a variety of random events/quests made with procedural generation.

Like a system where:
step 1. NPC is spawned.
Step 2. Based on NPC factors (factions, how much money they have, starting location) they can take part in events A-F or X-Z.
Step 3. An event type is chosen.
Step 4: Game chooses between a positive and negative version of the event. (So, you could meet a person on the side of the road who wants an escort to the nearest town, or you could meet someone on the side of the road who just wanted you to stop so they could steal your horse)
Step 5: Game picks out various details based on the previous steps to determine dialogue.
Step 6: Game looks at NPC to determine reward if one exists.

I don't know enough about procedurally generated content to give an informed opinion about it. From what I've experienced, though, I can say that I don't mind procedurally generated landscapes, so long as some tweaking is done to make the map resemble the landscape as it was previously established in the lore. Most everything else should be hand placed, IMO.

Hand-crafting everything is not the way to go, there's got to be generation to get a decent scale for the game. If they could bring back the generated quests to the series, it would be epic.

Every time I want to appreciate how far things have come, and have yet to come, I play the regular Nintendo or play a few games on the first version of the Playstation. Then I look at our existing games and appreciate how far we have come. Yes, we lost length of content in favor of beauty of content, but some at least try and bring this back, like Bethesda. Seeing those 8-bit games, some with over 50 levels of entertainment, and we never thought the day would come as quickly when you see worlds generated that look like a photograph.

This reminds me of using Lindenmayer systems to predict plant growth.

While I like some games that use procedural generation unless there is some reason it can't / shouldn't be done by hand (Daggerfall's size or randomizing levels in a Roguelike for instance) it detracts from the game. It's like somebody programming a machine to paint and then selling the paintings as their own work. In a way I consider games to be an artwork, and if it isn't put together by hand it says the developers didn't really care about their work to me.

That is a pretty inaccurate way of looking at it. If we stay with the art example, think of it like a parade float.

You have to make an awesome parade float with fifteen thousand dollars. One team had a flatbed truck to begin with, the other team didn't. Is the team that has to buy themselves a truck before they can get started in a better position than the team that already had it? Then think of it like this: What if you had the engine, but had to hand craft the truck yourself? That is a major timekiller. Probably hurts the budget too.

When it comes to making terrain, the more of it you have to handcraft, the more bland it's going to end up. The more time you spend making a great rock formation, the less time you have for everything else, which means you're quickening the process for the rest of it. Now if you had a generated terrain that's basically what you want, then you have all of that time you were going to spend creating every little thing, and spend all of it making creative awesome environments. Handcrafting everything in the terrain I would see as an insult because using the procedural generation from the beginning lets you create tons more creative environments without the filler busywork they'd be copy/pasting things into anyway.

Good example and reasoning. :goodjob:

It seems that either people do not like to read the lengthy OP, or it is too technical to be appreciated, so I like to write some easy to understand examples about the great benefits we can get by migrating to an engine that uses the procedural content generation for every aspect of the games, in the editor, or on the fly and in the run time.

It would replace the most part of the OP, and those detailed technical info would be moved to the next post.

I don't mind proceduritlyly (shutup) generated content... I just want them to pay attention and hand place things at important locations etc... generally give some areas the look that they put a lot of time and effort into the details/atmosphere.

You can tell when an area has received attention and detail from a person.


EDIT:
I voted for:
OK, procedurally generated changes to the landscape, like season effects, razed down villages, and so on... (Sorry for the late addition)
OK, procedurally detailed land outside of the province which is the focus of the game.
OK, procedurally generated main land as long as designers make them unique enough afterwards.
OK, procedurally generated wild-life in the main land.
OK, some randomly generated dungeons.

Basically I like the Idea of procedurally generated content BUT ONLY IF the developers use the time they saved through this process in the first place is used to detail more important hand designed areas.

I don't hate it. If done properly, it makes each playthrough unique

Oblivion didn't do it properly really. A lack of hand placed loot killed it for me. Even if it's just a retextured fork, I'd be exstatic to find it at the end of a long dungeon crawl

Also, more bosses would be nice. We had Umbra, and the Uderfryke, and... Umbra...

One thing that I loved about Morrowind was how [almost] every dungeon or cave felt unique. You never knew when you would find some awesome artifact that you could have read about in a book or heard a rumour about previously. Continuity is a great thing.

However! I do like your ideas and I think they could work. But I would rather see it implemented in a more diablo-esque hack n slash game.
Some of the stuff you talked about that Npc's could do. One day you may run across an NPC who needs a lift. Or one who wants to steal your horse.
They did that in Red Dead Redemption and it was pretty cool. But it got to the point where I didn't trust anyone on the road and shot first asked questions later.
Which in turn made the game feel more western to me.

It's good, I just don't know if I want that in my Elder Scrolls personally. :foodndrink:

Procedualy generated content is awsome if done well, and adds ALOT of replayability since every time you play its different. (or maybe im confused with randomly generated content). But the problem with generated content is the lack of detail.

I am skeptical towards randomly generated things in games because I fear that they will become not as unique and interesting as those designed by hand, that sort of thing simply does not have as much character as hand-made locations can be, and in a game that focuses on exploration, you need that to keep it interesting. And I don't think this is something that will change just because hardware improves, because in the end, its still just a computer generating some random things based on variables it's given, without the help of a designer's imagination, I just don't see how the results can ever hope to be as good as hand-made ones. While it might not be a problem in some games, since exploring an open world is none of the major appeals of the Elder Scrolls to me, anything that would make the world less interesting to explore is a bad idea.

And I'm not going to be convinced that procedurally or randomly generated content is a good idea until I see a game come around that applies it to create content just as interesting as hand-crafted content would be, which, so far, has not happened.



It didn't get to that point for me, but instead it got to the point where all the random events on the road started to seem rather repetitive, it was fun at first, but like all attempts at random events like that I've seen in games, they start to get repetitive after a while.



I wouldn't go as far as to say that, a lot of the dungeons still felt pretty generic to me, but the knowledge that at some point, I might stumble upon some unique artifact or some particularly unusual dungeon helped to make exploring dungeons feel much more worthwhile, because while perhaps not every dungeon would have a great find in it, knowing that if I was persistent enough, I would see something worth the effort helped to make the whole thing feel much more worthwhile, and that is one thing that bothered me about Oblivion's dungeons, since the loot was always leveled, if I find a good item in one dungeon, I know I could have just found it elsewhere, finding a good item is still a reward, but it didn't make me want to say "I'm glad I explored that dungeon or I wouldn't have found that item.", the game could really have benefitted from some hand-placed loot in the game.

Mind you, I did like the random loot in chests. Aside from the fact that you couldnt store stuff in them, the random loot was good as I never knew what I'd find and was suprised with some of the usefull things I found in there.