"I'm interested in this."

The Journey game is looooog project, and a lot of what i'm doing now is pre-production in a way. The story in the game is linear but based on vast areas of open landscape, with several important structures, and some not so important, but as you're in such a large area (And everything human is ridiculously dead by now, you find clues to this shortly after starting the game)

I love the idea of slightly leftfield game-attitude, the journey will, one hopes, blow many a mind with some sequences and i'll be using the finest bleeding edge technology I can find for it (And ive been reading stacks of academic papers for clues on how to go about it

The issue is systems, im a one man job, i need to make systems that cooperate in the game that i have to do very little authoring of - say clouds, oceans, river and streams, what wildlife is left needs to react appropriately. Your communication with the outside needs to estabish quickly, as in its manner, and then as things develop the whole awakening psyche of the character needs to be modelled, turned into a self sustaining system, with enough wits to escalate or modify the system depending on proximity, and then what will happens. A volcano using a voxel grid needs to be self sustaining and escalate suitably over time to reflect what I suppose could be a finale. What I can do proceduraly means no more tiny level authoring which takes forever. That said, it seems like a lazy way to do it, but actually its not at all, like, say if you stick to a brdf on your ocean, then you can get a proper colour and specular return, particularly the one i'm using which calculates the sky radiance (which is the sunlight bouncing off water droplets in the atmosphere) and the irradiance scattering indirect light onto the ocean. This approach means when my new 'system' being worked on, will integrate it rather awesomely, which is terrain, and the guys latest shaders are amazing. You unify all your models in physically based shading and end up with a relative small number of well understood inputs, and get predictable output, that look the bomb (I said it before but The Order is a great example of what compute shaders and pbr will get you)


The systems mean the game looks fucking great, so i'll be putting a lot of time (all of my time) into it hence my obsession with water, currently, its a very relevant system and masses of work has to go into for any game, cause crappy water just looks crappy, and in war-fucked washington state you need to see this destruction and how much is filled or not filled with water. And while i'm talking Oceans (and my ocean will take a LONG time, lots and lots of features and modes) there can be all sorts of rivers with accurate flow behaviour, you can paint that in on the water surface, with vertex colours or you can simulate it with houdini but rivers, they fascinate me, the lakes (much different dynamic to an ocean) very shallow pools or channels you can splash about in like crysis 3 - looking forwards to that too.

Nevermind awesome fun stuff I had using Dice's translucency trick on a bed of water (i had to figure the lighting model out a bunch myself though, dicks), you'd get light scattered from behind the water waves, might be pretty

And I mix all this cause why not, with quite a heavy free fourier transform algorithm Ocean users could have really spectacular waves (fft is pretty much the byword for cool oceans) at the cost of some frames, you could switch that out for a simpler perlin crossed for more wavey shapes and then add fine detail with normal maps - in fact the set of controls i'll be using will sport without doubt reflectance masks, i wondered what these were then remembered the classic way to get reflections from water, and these masks eliminate sides of the waves from the reflection so it makes it look like its lit properly - will be adding more pbr approaches and hopefully a marmoset skyshop integration which would be fancy

Well there's a nice block of text regarding water - the final result i want to achive, and god bless shader forge for this is actually pretty possible and i'm quite psyched about that

This is a long post, it could be longer but i'll spare you. If any of you have Unity skills going to waste i have several prototypes you can build on as i dont have time for it, and if theres questions, lemme know

Also, should the idea please you you could sign into the 'make marble madness insanely good' initiative, in this case though, unlike the first few levels, you can go crazy, use weird fabrics and surfaces, you can decide how much traction, bounicess slippiness quite easily and i think that would be cool having people muck in. You give me the idea and some drawings and i'll make it

This is probably very boring to read, i'll think of some screenshots for a bit!

Diluted Dante wrote:

" 'compiling'."

Look at this workstation! What a complete slob.

Keep posting Funn, this is your thread to do with as you please. I'll read it, need to get back in the swing anyway.

g.man wrote:

tldr

I don't think you're the ideal audience to be honest, this thread may just bore you to tears

The compiling joke made me laugh but i try to keep busy on my netbook if my desktop is grinding. Seems whatever field you fancy researching there's a shitton of material for it, often with very nicely supplied code to give you the idea. Guys like Eric Bruneton who do amazing work at making science make very pretty things (and as i keep saying, to any tech-adept artist or art adept programmer, go to the crc press website and download the supplemental code for gpu pro books, its magic) - ive spent most of the day still working on water, mainly work that stemmed from Erics work on a physically based lighting model for ocean surface lighting based on sun direction/atmospheric conditions. I was quite adamant about getting this in, even though it's definitely not the most practical of solutions in a game-play sense - but saying stuff like fft, brdf, ibl, scattering in participating media, get the graphics nerds excited so they make a bunch of cool stuff and gamers get to benefit from it

While mentioning Eric, here's Proland, more of his and others' work in a more-or-less entire planet simulator. The guy who ported erics water to unity is now working on porting proland over, this is reallly quite an exciting prospect

Theres quite a lot to water, but i think i got the hang of what I was doing, which is merging two 3d ocean approaches into one mighty beast of a water package, there should be updates soon, will hopefully make advances with the shader graph after I next wake up (This just to get them playing ball on a similar level), then the arty pretty options to add (these are easy and fun) and then the more sciencey stuff for those who simply will not do without a giant pile of realism. The fft water I explained is prime target for this as wether it be actually useful or not, fft models for calculating waves are still seen as the top of the pile - hence reading about light absorbtion in water (i'm assuming theres a formula for this based on depth, lighting and water quality, i read somewhere that the visibility goes down by thirds based on where youre looking bt whatever i'll get it.

While i work out how to do that i'll do the arty stuff which is fun and means i can feck around in shaderforge for a few days. So! When the water's wrapped up onto the next thing i think i need to learn - Probably go back to doing game art for a month or so cause i really need material to even benefit from all this new lighting gubbins, but the water will take a bit and the upshot is i think a bnch of grind and frustration is out of the way - so pretty things will come!

More pics less words plz. And there's not enough sex or drugs in this diary.

I really ought to start using Unity again...

Oh and dont worry Spacegazelle, sex and drugs are pretty much some of the cornerstones of my life, in fact ive found amphetamines make you ridiculously productive. So i'll bring up the subject in detail eventually, i tend to keep my sex life to myself but if something spectacular happens and its both sexy and relates to my work, i'll chuck it in.

Side note: If any of you are texture painters, especially as environment artists but even modellers/texture folk, get hold of filter forge this damn instant, its absolutely the bomb for endless amounts (assuming you have the filters) of material for your textures, and thats not just diffuse colour, but also your bump (Use them for heights in vertex animation or height-needed shaders like POM or tessellation), normal (you should know what these do), ambient occlusion (Theres a fun trick you can do with these to make maps for translucent surfaces), and even, and these took me surprise because they never occured to me - but reflectance masks - you use these when you're using these textures generated with bumpy surfaces, particularly waves, but you're using the old mirror method for getting reflections into your waves. Also a good source, PARTICULARLY if pbr/ibl appeal to you is gametextures.com

Using the mirror method for waves gives you nice reflections of everything but if the sun is on the opposite side of the 3d reflective surface, it still looks like the surface is reflective when it shouldnt be! terrible right? You use these masks to apply to your other generated textures to mask out these reflections = much more realistic reflections

If you want realistic dynamic reflections without the mirror method, you use an environment cubemap, Unity can generate these via its API, which is handy, but the face size needs to be tiny or no more performance  for you - my plan is however to use quite a low resolution cubemap for the environment to give good general reflections which are positionally accurate but very low res, and blend it with the mirror approach, giving the best of both worlds

Theres also a method called screen space reflection which I may look at in future, it has a lot of issues but looks beautiful.. combining that would give great reflections for ay data that there is on screen, mirror reflection and cube/environment masks will take care of the rest

And on the subject of reflectivity, and again physically based shading, i'll show you an example of how unity vanilla can be frustratingly shit with lighting, but asset store stuff is the fecking bomb (somethings on there cost 100+ dollars but many of them are incredible and would be worth a fuckload more if it was an art team doing a bespoke system for use in only one game)

Example:

Just Unity lighting



Now Unity lighting combined with image based lighting and physically based shading



Next up, when I can be arsed, i'm going to talk about water in great detail i think. The more academic amongst you might find it interesting. I wont talk about fft wave generation but i'll link you to an article or two now the JONSWAP spectrum used for the fft ocean i'm using

Here's a picture of the Phillips spectrum (I wont be using):



Actually I might add the above as an option if feasible, it does run reasonably quickly

Here's gerstner waves, which I irrationally hate



And if you REALLY want to know about ocean simulation, read Tessendorfs seminal paper on the subject which is pretty much the basis for all serious business research into wave generation, cept the boring ones

And heres one of if not the most important article written on the subject of ocean water rendering

But instead i'll cover the optical effects of water, and how to art them up. I'm still waiting on a guy to implement custom nodes/appending code manually to his tool output and i REALLY REALLY dont want to make the code into nodes because theres a lot of it. But I can go into optical effects for the surface (Both based on research papers and examples and artistic control and with any luck, melding successfully with the standard BRDF of the fft ocean) and the generation of shoreline flow textures which are quite a lot different to open sea waves

Oh yeah, shittons! I was a huge fan of its daddies, Vista on the Amiga and Vista Pro on the PC. I just checked their site, didnt realise there was a version 3! I tend to use Vue for those things because I can integrate it with 3DS Max but I don't do stills anymore so previous versions of terragen became a bit useless to me. I will definitely have a good look at the site though, if you can export heightfields that would be very very useful for use with Unity's own terrain system (Even if just introducing high frequency but very low amplitude noise to a terrain extracted from World Machine or a 3rd party tool for Unity I use called World Composer which grabs real life elevation data and satellite imagery from across the world (pic):



I've had better results than that, mind

Depending on the quality of its output and speed it may well prove to be the exact thing I was looking for to create panoramas for IBL, and those unwilling or have no need to buy Nuaj and just want the pretty static clouds

Why do you ask?

LazyGunn wrote:

Why do you ask?

I used it to make some islands for Op Flash. Eleventy years ago!

Enjoying this. I don't understand half of it but that's most of the fun of reading about it for me. A wee glimpse behind the curtain.

Bounding box diagonal is normally good enough for figuring out rough screenspace coverage and LODing based off of that IMHO. Means you have to have tightly fitted bounds but that should be the general case anyway (to make sure visibility checking is as efficient as possible and isn't drawing things it doesn't need to).

The switch point between LODs is the most obvious point where people will notice something odd going on - you may be able to crossfade between 2 LODs but that starts getting expensive. TBH, as long as the closest LODs look similar enough people probably won't notice (seriously, always just try with the cheapest/easiest solution first because you'd be amazed what people who haven't been primed to look for LOD pops just don't notice). Most jarring is often when the final LOD pops out and disappears completely when it hits a small enough screen size - so having an option to have the smallest LOD alpha fade in/out as it reaches the lower size limit might be a worthwhile investment - just to prevent small things on screen popping in and out.

I found this quite neat for treesville but it does seem to need at least some close up LoD stages, if possible, and they dont seem to react to lighting which would need to be addressed, but at distance maybe perfect, and free!

Here's a picture for you picture lovers

LazyGunn wrote:

Cheers for that! Regarding LOD production, i've been considering several ways to do it, the typical Unity method is just to blend between their treemaker trees to an internally generated billboard at a distance you can set. I find this pretty irritating and i've been looking at a lot of talk regarding alternatives, i'm about to try a shader selection I already own which has a few ideas on the solution, but while I try that out i'm thinking about how relevant dynamically generated impostors and volume objects - not true volume textures, just the original model sliced several times in 3 dimensions and interpolated to give an approximate look, are, and at what stage they should be employed 

I'd always recommend starting with the simplest method, assess the weaknesses (but always keeping an eye on "will the regular punter actually see/care about this?") and if there's a fix or improvement for that particular method, only at that point start to complicate the matter - and only if there's no easy fix go for an alternate method.

TBH, I would have thought impostors would have been fine? If you've got any dynamic lighting you might need the impostor to render out as separate albedo and normal map, so it can react to lighting and shadow changes - but for the long distances it should be close enough, right?

LazyGunn wrote:

You'd actually have to tell me what a bounding box diagonal is aha, but would it be faster to calculate a bounding box than a view-orientated disc/ellipse? Maybe Unity already has a bounds property for objects.

Just the distance from top-front-left of the bounds to bottom-rear-right - Unity is sure to have a bounding box for props.

LazyGunn wrote:

Regarding lod pop-in, thats much the reason I want several stages of LOD, not just 2 or 3, with the most distant lod being faded perhaps (As this is usual Unity functionality I think). Theres a tree creator for Unity called Forester that procedurally generates trees and I wonder if I can use any screen coverage amount to drive the complexity of the trees via API

Probably - but remember you want as much as possible to be pre-calculated/cached, so that you don't stress the CPU at runtime.

I don't have the first fucking clue what Gunn is on about, but I do like me some pictures.