Odyssey trailer - thoughts?

[dannysquid]

I think the problem isn't so much creating the water effects (sinking ship or not for example), and rain just needs to be a graphical effect plenty of games have done - although all sorts of rain on all sorts of worlds, different liquids, different gravity, different atmosphere, that'll get interesting. It's generating it convincingly. Randomly generated terrain with a fixed sea level isn't too hard - that's been done plenty of times, but if you want rivers and so on, calculating where they go on the fly (too many worlds to predefine them all), giving tributary systems that don't cross each other, convincing variation in size, working out where a lake actually is - that all sounds like quite a challenge. A very interesting challenge to be sure, one that sounds like it would be very interesting to work on.

Chucking plants and animals on top of that's the easy part, although ideally you'd want to procedurally generate all of them too to give each world unique lifeforms.

Yes, fixed water level is very easy (this is how most games do it), but proper water levels and rivers and tributaries etc... is borderline impossible with PG worlds that need to be updated 60 times a second. At least I can't think of a way of easily doing it.

This just furthers my feelings that if a gaming company can't compete with Rockstar, they should do something else

Rockstar has billions of pounds and thousands of employees working crazy hours for years and years and years.

You can count the gaming companies that can compete with Rockstar on the fingers of one hand.

 

[[PS4] Filthymick420]

Yes, fixed water level is very easy (this is how most games do it), but proper water levels and rivers and tributaries etc... is borderline impossible with PG worlds that need to be updated 60 times a second. At least I can't think of a way of easily doing it.


Rockstar has billions of pounds and thousands of employees working crazy hours for years and years and years.

You can count the gaming companies that can compete with Rockstar on the fingers of one hand.

This just means other companies need to step up their game

 

[Helvellyn]

Yes, fixed water level is very easy (this is how most games do it), but proper water levels and rivers and tributaries etc... is borderline impossible with PG worlds that need to be updated 60 times a second. At least I can't think of a way of easily doing it.

I've one or two ideas, although they wouldn't be perfect and might result in noticeable sameiness (and large areas without rivers). Wouldn't solve the problem of ending up with dry basins that you'd expect to be lakes though.

 

[dannysquid]

This just means other companies need to step up their game

Rockstar's power is down to practically limitless amounts of time and money, which isn't something another company can just whip up overnight. It takes decades of success. Even off the back of a megahit like Witcher 3, CD Project Red clearly ran up against reality with the release of Cyberpunk 2077.

 

[dannysquid]

I've one or two ideas, although they wouldn't be perfect and might result in noticeable sameiness (and large areas without rivers). Wouldn't solve the problem of ending up with dry basins that you'd expect to be lakes though.

For rivers I wonder if you procedurally generate them first, then create the landscape based on the location of the rivers. So you're calculating height etc... based on a pixel's proximity to a river. One of the things I love about ED's planets is the way they (reasonably) seamlessly get more and more detailed, but I think if you were using the same system with rivers, the more detail was added the more the paths of the rivers would change (as the slope of the landscape would keep changing) and you'd end up with them squiggling about all over the place as you got closer.

 

[deckroid]

Meh. I didn't buy this for FPS.

I like to fly ships in space.

Everything else is superfluous.

 

[Bantha]

Critics is good however could you explain what you expected?

Good question.
It’s going to be very difficult for me to articulate this without emphasising what I didn’t like, or simply saying “better this and better that” but I’ll try my best...

So to start with, I actually thought we’d see more than just the FPS element of the game.
I thought we’d see a bit of everything - pew pew yes, but also players stepping off their ships, that first step for mankind, those first footprints... this is a big thing right? Actually stepping on to terra firma, it’s been a long time coming.
Also, interacting with the environment once on foot, the SRV and Ship, alien/geographical sites, crouching down and scanning an artifact, picking up materials - something like that. Plus pew pew, of course

But, given that the demo was purely FPS, then based on that alone, this is what I expected:-

The main issue I had, and my initial reaction, is what it looked like.:-

• So, first is the graphics. I expected to see something modern, on a par with the rest of the game, and also on a par with the better FPS games already out there.
• Possibly more importantly, the animation. I expected good animation, with natural looking human movements. In particular walking and running.

Then the AI and interaction with NPCs:-

• I assumed the AI would act intelligently and be interactive - taking cover when under fire, reacting to suspicious behaviour, or approaching the players to scan them for example.
• I hoped that the pew pew itself would be more tactical - using cover, shooting from roof tops, going for headshots...

Also, the player movement options:-

• I expected strafing, crouching, rolling and jumping, and peering around corners.
• I expected stealth. The option to creep up and stun opponents e.g. creeping up to the workers and strangling them or knocking them out
• I expected some form of melee

Hope that answers your question.

 

[Helvellyn]

For rivers I wonder if you procedurally generate them first, then create the landscape based on the location of the rivers. So you're calculating height etc... based on a pixel's proximity to a river. One of the things I love about ED's planets is the way they (reasonably) seamlessly get more and more detailed, but I think if you were using the same system with rivers, the more detail was added the more the paths of the rivers would change (as the slope of the landscape would keep changing) and you'd end up with them squiggling about all over the place as you got closer.

You can control the degree of squiggle - most random noise landscape generation methods have that, so you have control over how rough it appears at different scales. I was thinking of height as a proximity to rivers generated first, the hard part is possibly generating the rivers at a decent scale themselves and not having them cross over, or two large tributaries ending up back near each other with significant height differences. If you had a selection of pre-generated large-scale river maps the smaller scale could be randomly generated within limits to avoid those problems, and they could be distorted without crossing problems, but the sameiness would probably be fairly noticeable.

The problem isn't too bad offline, it's the on-the-fly aspect that makes it tricky. I keep meaning to play around with some ideas someday (not that I've got the ability to scale any of them up into a real-time system useable by a game).

 

[dannysquid]

You can control the degree of squiggle - most random noise landscape generation methods have that, so you have control over how rough it appears at different scales. I was thinking of height as a proximity to rivers generated first, the hard part is possibly generating the rivers at a decent scale themselves and not having them cross over, or two large tributaries ending up back near each other with significant height differences. If you had a selection of pre-generated large-scale river maps the smaller scale could be randomly generated within limits to avoid those problems, and they could be distorted without crossing problems, but the sameiness would probably be fairly noticeable.

The problem isn't too bad offline, it's the on-the-fly aspect that makes it tricky. I keep meaning to play around with some ideas someday (not that I've got the ability to scale any of them up into a real-time system useable by a game).

I'm far from an expert (although I do sort of create landscapes in 3D for a living, just not in realtime!), but wouldn't it be the case that the bottomless detail in a noise function would only be resolved at the resolution you're currently rendering at, so if one pixel is 1000m from orbit and 0.001m up close then the slope (or normal) of that pixel would be very different at different distances and any rivers you were calculating would calculate differently. Or something!

I wonder if the rivers were to incorporate a height stamp that gets subtracted from the surrounding landscape that gets ever wider the closer you get to sea level, so the land is forced to go downwards. Not sure how that would actually look without trying it. Probably rubbish!

Again, it's the having to do it in microseconds (or whatever games people calculate things in) that makes it tricky.

 

[lysan]

Good question.
It’s going to be very difficult for me to articulate this without emphasising what I didn’t like, or simply saying “better this and better that” but I’ll try my best...

So to start with, I actually thought we’d see more than just the FPS element of the game.
I thought we’d see a bit of everything - pew pew yes, but also players stepping off their ships, that first step for mankind, those first footprints... this is a big thing right? Actually stepping on to terra firma, it’s been a long time coming.
Also, interacting with the environment once on foot, the SRV and Ship, alien/geographical sites, crouching down and scanning an artifact, picking up materials - something like that. Plus pew pew, of course

But, given that the demo was purely FPS, then based on that alone, this is what I expected:-

The main issue I had, and my initial reaction, is what it looked like.:-

• So, first is the graphics. I expected to see something modern, on a par with the rest of the game, and also on a par with the better FPS games already out there.
• Possibly more importantly, the animation. I expected good animation, with natural looking human movements. In particular walking and running.

Then the AI and interaction with NPCs:-

• I assumed the AI would act intelligently and be interactive - taking cover when under fire, reacting to suspicious behaviour, or approaching the players to scan them for example.
• I hoped that the pew pew itself would be more tactical - using cover, shooting from roof tops, going for headshots...

Also, the player movement options:-

• I expected strafing, crouching, rolling and jumping, and peering around corners.
• I expected stealth. The option to creep up and stun opponents e.g. creeping up to the workers and strangling them or knocking them out
• I expected some form of melee

Hope that answers your question.

Yes all good a valid points.

 

[CMDR Sophie Joanne]

I'm far from an expert (although I do sort of create landscapes in 3D for a living, just not in realtime!), but wouldn't it be the case that the bottomless detail in a noise function would only be resolved at the resolution you're currently rendering at, so if one pixel is 1000m from orbit and 0.001m up close then the slope (or normal) of that pixel would be very different at different distances and any rivers you were calculating would calculate differently. Or something!

I would assume (assumption is the mother of all F-ups, lol) that the terrain is only calculated once (at the highest resolution to be used), then saved as a string/seed that can be reloaded (exacly the same) at a later time, with lower resolutions being a LOD of the highest resolution, whereas applying noise at different resolutions would result in completely different terrain each time the noise is applied.
I hope that makes sense as I don't really know the correct terminology - I know what I mean though.
As a hobbyist game programmer, that is how I would do it (I might have just explained why I really struggle with terrain/meshes/textures though!)

 

[dannysquid]

I would assume (assumption is the mother of all F-ups, lol) that the terrain is only calculated once (at the highest resolution to be used), then saved as a string/seed that can be reloaded (exacly the same) at a later time, with lower resolutions being a LOD of the highest resolution, whereas applying noise at different resolutions would result in completely different terrain each time the noise is applied.
I hope that makes sense as I don't really know the correct terminology - I know what I mean though.
As a hobbyist game programmer, that is how I would do it (I might have just explained why I really struggle with terrain/meshes/textures though!)

I don't think they could precalculate the whole thing. The free NASA textures you can download are all about 21,600x10800 pixels in size, which is pretty huge as far as texture memory is concerned. And yet the UK is only a few pixels high at this resolution, and the landscapes we're seeing are a lot more detailed than that. I'd guess it's more of a function that asks what_height_at(x,y) and what_material_at(x,y) and returns a simple answer even if the stuff going on inside is pretty complex.

Noise functions are usually seeded rather than random, so if you input the same seed you get the same noise. I see that with Odyssey they're also distorting the noise functions to create things like proper mountain ranges and that the texturing is following the slope of the hills, so that will get you a lot more interesting results straight off. It looks to me like there are also interesting things going on with the displacement of the cliffs, which is possibly a side effect of getting the textures to follow the curves. Can't wait to see it in action!

Apologies for incorrectly sounding like I know what I'm talking about.

 

[Helvellyn]

My guess at how the current algorithms work (which don't need to worry about things like rivers flowing downhill) is large-scale regions are defined (hilly, plains etc.) with whatever random noise algorithm being used works with the approriate parameters within those. We also have crater models, which whilst not a predefined heighmap can still be applied on top of whatever the terrain has been generated as anyway. Existing canyons may work in a similar way to proposed rivers except that it's a much easier task - no need to be concerned about criss-crossing, or the bed going up and down.

Height based on distance from previously-defined river (with a random function) - could well work, a scaling factor on the random noise with some attention to the gradient too maybe. Whether or not it is easy or hard to make that run fast enough I've no idea.

 

[Soulstain]

star citizen is impossible to run on most pc's.

Run’s just fine on my son’s 1080 and my 2080ti. The game isn’t even at the optimization stage.

 

[VR StarLightPL]

Elite Dangerous: Odyssey Mission Q&A - Recap

Greetings Commanders! Last week was big for Elite Dangerous: Odyssey news! Not only did we release a brand new look at Odyssey pre-alpha gameplay, we also revealed that the alpha will begin on March 29 on PC (more details on this soon). The gameplay reveal was followed by a live interview with...

forums.frontier.co.uk

for perspective ;-)

 

[cmdr the dude is me]

Run’s just fine on my son’s 1080 and my 2080ti. The game isn’t even at the optimization stage.

thats a very high bar to have. with a 1660ti and 16 gb ram I couldn't even run it.

 

[Jukelo]

I'm far from an expert (although I do sort of create landscapes in 3D for a living, just not in realtime!), but wouldn't it be the case that the bottomless detail in a noise function would only be resolved at the resolution you're currently rendering at, so if one pixel is 1000m from orbit and 0.001m up close then the slope (or normal) of that pixel would be very different at different distances and any rivers you were calculating would calculate differently. Or something!

Indeed: you need to know the terrain to know where liquids will flow, but the terrain changes as your level of detail changes (because you can't afford to generate the whole planet at the maximum level of detail and to calculate the flow, in real time, at the planet's scale). And any change in the terrain can cause a dramatic change in the flow of the liquid, unless you're fine with rivers running uphill, or lakes being contained by empty air (which might be fine in, say, a flight sim where not too much attention will be paid at ground-level details like this, but not a game like Elite).

Doing it the other way around and generating the rivers (and lakes and oceans) first would the only feasible way imo: you get to control the start, end point and all the features in between (as opposed to flowing from high to low points randomly distributed on a noise map over whose result you have little control). But it is no panacea either: it's a lot less data than a whole planet's terrain heightmap, but if we are talking about an earth-like planet, it's still going to be a lot, which your terrain generation algorithm will need to shape its results around. And suddenly you're no longer just displacing the whole planet's surface according to a dumb heightmap (not that this is the extent of how Frontier generate planets, the presence of mountains, ravines and craters is evidence that there's some high level planning of major surface features).

Run’s just fine on my son’s 1080 and my 2080ti. The game isn’t even at the optimization stage.

The typical gaming PC runs a GTX 1060.

 

[Soulstain]

thats a very high bar to have. with a 1660ti and 16 gb ram I couldn't even run it.

MSFS 2020 released last year with a recommended spec of needing a 3080. It’s just the reality of games that push the boundaries. SC is the most ambitious game ever created so you have to expect it will take future tech to run it well. This has always been the case in my almost 30 yrs of PC gaming. If you were around when ED launched, many people complained about it’s performance as well. Many were really complaining when they flew down to planetary bases. ED’s game engine is now 8 or so yrs old. SC’s game engine was completely rewritten recently to handle the much more ambitious ideas they are creating.

 

[Tom Jacobs]

1080ti x3 here and tried SC with a few free fly days on my 1440p ultrawide and its unbearable to play and crashes randomly.
ED im playing with locked 75hz on ultra.

 

[Virgile]

1080ti x3 here and tried SC with a few free fly days on my 1440p ultrawide and its unbearable to play and crashes randomly.
ED im playing with locked 75hz on ultra.

Same. On a 1080. It runs like sh... On HDD it's absolutely unplayable. And still terribly Bad on SSD. I've played Cyberpunk day one, it was Bad but not star citizen Bad.