Search This Blog

Saturday, December 31, 2016

Technique Training: Updating Game Guru to HD


So it's no secret Game-Guru is really weak in terms of performance.  I've covered several topics in the past ranging from more advanced lighting techniques to improving the way reshade handles your graphics.

What I've found in the past is it's really difficult to work with a lot of the stock models.  There's simply a cutoff for what you can get from a model that's pushing 512x512 SD textures.

Before we continue, what I'm talking about here is something called 'texture depth'.  I'm sure by now you've probably heard the acronyms "SD" and "HD".

Taken from some chinese blog, but suits the purpose.
Basically the less pixels you dedicate to the quality of the texture surrounding your model, the less 'depth' it will have in the sense of how detailed the picture can look. Here's an example.  The stock asylum graphics are .. acceptable .. by Game Guru standards.

SD is really anything under 720 pixels wide/high. Most games tend to use standard divisors of 8 so you have like 256x256, 512x512, etc.  This tends to max out around 4096x4096.  This is primarily because of the graphic cards capabilities.

However you are using a 2048x2048 texture that's divided up into multiple subcomponent textures.  This basically makes each effective texture 'SD quality'.

Some are more, some are less, but all are SD.

Now to spruce up the texture, what I do is go is open it up and immediately resize the texture to 4096x4096; then proceed to sharpen the picture significantly.
It's hard to see with this scaled down snapshot, but trust me, it's there.

The resultant image has much higher fidelity.

Now the real trick here is fully utilizing ALL of the mapping available (when necessary) for Game-Guru textures.  Game-Guru can handle basically four types of mapping:

  • The standard UV map - this is your _D file.
  • The normal map - this is your _N file.
  • The specular map - this is your _S file.
  • The illumination map - this is your _I file.
Generally speaking, most models only include a _D and a _N file.

In the case of our asylum texture, we have a specularity file as well (_S).  This particular addition gives a slight shinyness to the textures, making them seem wet or slimy.  You can only see them with the entity shader set in the tab-tab options menu set to 'Highest'.

These settings are what I found using in combination with reshade as producing the best visual result for the asylum.

Default walls, unmodified.  Looks ok.

Now that looks .. ok.  That's the best you're really going to manage from stock assets and some really basic lighting.  Now if we modify the specularity file so it's a lot more pronounced on certain areas.  The basic specularity file is pretty clearly just a black and white image of the original _D file which is fine unless you want to really highlight areas.  Now tile generally tends to have a shiny edge because of the enameling on it.  So what I did is took the original image and modified the luminosity settings in to give it much higher contrasting on the black and whites, causing a more clear delineation in the shiny areas of the texture.

Old on the left, new on the right.
Side note: If you don't HAVE a specularity file with your models, you can generate one simply by making a black and white image, saving it as _S on the end of the filename and then modifying luminosity (or using edge detect even) to create bright white spots.  Specularity maps work by using a greyscale of the image - the more 'white' it is - the shinier the surface will be.

The final result, in combination with static lighting ends up looking like this:

 Note the cleaner texture lines, better shine on the walls, etc.  It's really a vast improvement over the original configuration.

While a lot of this is trial and error, it's a big piece of improving your work to at least be within striking range of more up to date engines.

One final note:  I personally run this setting in reshade's mastereffect.h file:
#define USE_HDR_LEVEL        2    //[0 to 2] HDR Level: Rendering bitrate. 0: RGBA8 | 1: RGBA16F | 2: RGBA32F

2 seems to be the highest fidelity setting possible if you're interested in HDR quality imagery.

Addendum to original post added here on 1/4/2017:

Monday, December 19, 2016

A most interesting day

It's not often you hit 10's on everything you are doing.  I really was way off the charts in terms of motivation, energy, focus, etc.  I am happy with this and made great progress on several major projects.   Also interesting when you discover a significant amount of your favorite art was done by one person (without you realizing it!):

Yoshitaka Amano

Good times.  I really have a lot to do now but am happy with the progress.
Now to finish up this project for my wife's Christmas Gift.

Thursday, December 15, 2016

Technique Training: Optimizing Game Guru Levels

So I was working on my most recent project - *famous game* homage for my wife.
Something she could play for christmas.  My son and I were both going to help each other and produce a really masterfully done game for her in the vein of one her favorites.

This would be done in a 3-4 week timeline with about 3-5 nights a week of a good solid 3-4 hour runs on adding stuff.

Sourced from - very accurate assessment of stress, IMO

Well we quickly discovered how unwieldy a large map would be treated by Game Guru.  So I decided to go back and optimize the map in every way I could.  Here's a bit of that process.

Before we begin...Limitations

A key understanding here is that Game-Guru is only 32-bit.  It's also using DirectX9.  These are effective, but outdated infrastructure components.

32-Bit applications are limited to 4GBof memory. But really, considering all the underlying subsystems, operating system data, etc.  It's more like 2.5 GB.  Also, for the uninitiated, memory is basically the short term, high speed data storage your computer runs applications in.

Holy crap he writes, makes games AND uses excel, incredible!
64-Bit on the other hand, can access something like 16 EXABYTES ... some absurdly large number that's out of reach for the time being.  So arbitrarily, we'll use a common value of 16GB for our memory example.

As you can see, your game only takes up a *FRACTION* of available memory on a 16GB system.

Bear in mind though no matter if you're running a 64-Bit version of windows (as I am) that the application itself, in this case Game-Guru, can only ACCESS 4GB of ram at any given time.  And it always considers the OS as part of that.

In real terms, this means that only 2.5 GB of memory is ever available for any given time.

Now consider that every single item you add, every texture, every object, etc - occupies a bit of that storage space.  If you are using thousands of unique objects, each one has to be stored separately.

DirectX9 adds another limitation in that it is vastly more CPU reliant than DirectX11.  So in my case, for instance, I have a quad core AMD running at 3.2GHZ for my primary system.  My video card is a GTX960, with 1gb of video ram.  While it's not top tier it's still an excellent card for any game.

Except, of course, Game-Guru games. This is because DirectX9 relies on my CPU which is at best, mediocre.

Unfortunately this is another pipe-lining bottleneck we have to work around (at least until Lee gets off his duff and codes this thing in DX11!).

Optimizing Memory


From this standpoint of minimizing memory usage, we need to utilize this knowledge in a ground up context.  While a lot of AAA titles these days use incredible CPU and Memory resources to produce an earth-shaking experience, many older systems and titles had to be more inventive.

Some examples of this are:
Silent hill
Resident Evil

What we're going to do here is emulate the design by using a tried and true multi-level system with several small zones off a hub zone.  This is the main world, an area where you navigate from place to place.  However once you arrive off the world map, you open a door to a building and are transported to a new level.  This is done in such a way that the load times are minimal and it typically makes logical sense.

Outdoor zones in fallout 4 are part of the 'world' zone -
This quarry, for instance.

But when you go to a building of any measurable size, you get an 'enter' option that you can click which gives you a loading screen like this:

Side note: Probably the most interesting weapon in FO4.

Then you're loaded to the new zone.  Now we can accomplish a similar trick (albeit without the fancy loading screens) using winzones.  The winzone object in Game Guru allows you to setup a pointer to another level which is then brought along when the game is compiled.  When the player steps over the zone, it ports them to the new level.  You can also port them back to the original level, if you desire.  This is the method we'll be using.
The concept here is very simple.  If you are using one large map, it's seamless - without load times.  But it's going to be slower because it's got to load all those objects into memory.   It's also going to be less flexible because each unique model you add to your level is going to increase memory consumption. So concessions will have to made if you do that.  There's also the issue of load times.  If you run a massive level, it will cause gigantic load times for game-guru.  You also *ONLY* will ever have 2.5GB of memory to work with.  That's a huge bummer if you're using high quality textures.

Fallout 3's map - the red spots are *mostly* a level portal to a new area as discussed.

Instead we're going to launch sophisticated areas (Schools, hospitals, houses, minizones) as their own zone and thus free up the memory we would normally use for the world and pour that into each individual zone.  For the sake of load times, we don't want to go too crazy, but even if each zone is 500MB in size it quickly would add up if made as part of a massive world map.

The difference is clear, at least in terms of memory usage.

This gives us the ability to rapidly distribute the load across multiple maps and bring up our total resources used.  It also gives us access to extra goodies in Game Guru such as differing skies or settings between maps.

To further speed things along we're going to use low poly models when available, low quality UVW (UV-Wraps, aka textures), and try to reuse models.  Every time we reuse a model, it further saves memory by simply requiring a pointer in memory vs a huge reload of the object into memory. 

Ground Up Design Processes

Now that you know *WHAT* we're going to be doing - it's time to do it!

The above is a simple demonstration of how to perform a multi-level configuration in Game Guru.

The first thing we want to do, as usual, is map out our levels.  We want to keep in mind that any main areas will be sectioned off.  Even the world map can benefit from being broken up into sections.  I prefer to use major areas or quadrants.

I like to draw on paper exactly how things will flow for the player.
But this isn't paper.  This is the internet. So you get a bad MS paint drawing.

It's terrible, but you get the idea, probably.
The next thing to do is to figure out the assets we're going to be using for these zones.  Each of these areas will have it's own specific feel and layout. So make sure you create a list of each area's similar assets so you can keep track as Game-Guru's asset management is pretty abysmal.

This is, in essence, the very rudimentary beginnings of a pre-production blueprint.  You should check this link on more about that:

Moving on.

So basically what we want to do is use similar, low cost entities when doing background pieces and higher cost/high poly models in foreground elements.  Remove anything extraneous and wherever possible duplicate items.

Picture sourced from's screenshot gallery by user Morphtactic

You can see above a good example of reusing elements.  The trick is primarily to use them assymetrically.  Giving an object that's visually similar the same orientation causes the brain to see it quicker.  Giving it a slight twist, say 10 degrees, can make all the difference in how two identical models are perceived.  I give more info on little tricks like that in this article:

So all that aside that's basically it!  Just break up the level into easily digestible chunks and it will keep things nice and low for you.  You can still have sprawling maps - just be smart and keep the load times down!

Finer Tuning

Once you've got your indoor levels fleshed out it's important to remember that indoor levels can basically be stripped of terrain and/or water using the stock script 'goindoors.lua'.  This will further help framerate on visually intensive locations.  Just make sure you account for the lack of terrain if you have transparent windows looking out into the world.

Another important element is to not load up your scenes with tons of enemies.  Unfortunately the AI right now really doesn't support group think very well so it tends to act in a very poor fashion when more than a handful of AI are active at a given time.  Done properly (use blind corners, traps, etc) you can still make the situation difficult for the player without killing their CPU in the process. 

Other things you can do are fine tune your occlusion settings, which are tricky to get functioning properly but once enabled will properly cull objects out of view to ensure that the maximum framerate is being used at a given time.  This however will be a topic for a future date, when I can devote more meaningful time to the discussion of occlusion.

After Christmas I'll put a post up showing more specifically how I used this and screenshots of the work involved so a better idea can be had about how it was utilized.

Wednesday, December 14, 2016

No, I haven't forgotten you.

Oh sweet, dear readers ... all ten of you...

I haven't forgotten about this blog.

No actually I've been exceptionally busy with the holidays!  There's a lot to do, after all.

I'm in the process of making a game, with my 6 y/o ... our first collaboration ;) for my wife, for Christmas.  In that vein, I've had to do some scripting for lighting transitions, collect quests, etc.

The actual game design work is coming along well.  Though there's still a significant amount to do.  Actually with such a clean cut end goal to run towards, it's really helped streamline my process.  I learned a lot along the way as well, which I will hopefully be adding here soon.

I'll try to get some reviews done.  All in all though, very good stuff.

One very interesting thing that's developed was this:

A forum user had asked me about some issues with fog effects not covering the sky sufficiently which I was able to duplicate:

It was forming a very clear line of demarcation on the skybox, which was making the fog seem box-shaped.

The solution was provided here:

Basically it's a replacement shader.  I'm still playing with the values suggested in the post, but overall the results are very good.

Monday, December 5, 2016

GameGuru games to watch out for.

Game-Guru is fairly synonymous on steam as being a source of 'total abject shit' games.  And I don't say this lightly as someone who's a frequent user of it.

This isn't really a bad thing, it's more a predictable thing - primarily because the game engine is marketed towards and genuinely appeals towards the extremely casual novice game maker.  You unfortunately end up with a lot of this:

And for some reason this guy thought to 'improve' the stock 3d rifle with a static 2d image set.

And that's just not good for your reputation, despite the fact that it's really easy for even a total neophyte to make a game with.

There are, however, some very skilled Game-Guru game makers.  People who use it both professionally and casually but have a well of experience that make their products and offerings worth looking at or keeping an eye out for.  Some of these projects are completed.  Some are in progress.  Bottom line is they make a good baseline for what kind of work you want to emulate if you're going to create and try to be successful, legitimately, with Game-Guru.

1) Ertlov's "Father's Island"

This is probably the most commercially successful game made for Game Guru as of the time of this writing.  It's a strange game with a lot of interesting elements.   Basically a puzzle thriller with some interesting secondary elements and a lot of personally done acting done by the creator (Ertlov).  It's been critically reviewed in Austria and has a small, loyal following.  The graphics are well done, the lighting superb.  The impression I get is it's mostly stock assets too, which is even more impressive.

2) Wolf's "Acythian" and "Shavra" games.
Wolf is well known in the Game Guru Community as a unique artist and freethinker with an eye for dramatic lighting in his games.   He's got two separate projects on the table:

The above is Shavra - Renaissance .  It's a high-fantasy adventure with tons of custom assets.  The scenery is gorgeous and should stretch the limits of the engine's power.  I love the tapestries and imagery.  His custom characters are pretty impressive too. Check out the progress thread here:

His other game, of comparable quality, is a Sci-Fi thriller called Acythian.  It's right up my alley with well paced action, a gritty futuristic vista, and fast action.

This game, in my opinion, is probably as close to AAA quality as you'll see from Game-Guru. You can find out more here:

3) Bugsy's Kshatryia File 002

It was incredibly hard to pick a scene for this game, too many good choices.
This is another really gritty cyberpunk style game, though it's far more in the vein of blade runner or more specifically Ghost in the Shell.  It's not done yet but my only complaint about it is that it will only be three levels.  You can, however, pick up most of his objects on the store if you want to build similar levels.  That said, what's available here should be complete very soon.  Once it's done, you will likely want a copy if for no other reason than to try to learn a few things from it.

Forum thread:

4) Savior v3
Savior's been in progress for a very long time.  So far, however, the results are promising.  It seems Tarkus runs into the same issues I do with life actually eating up available free time.

The positioning of the clutter items shows a fine degree of care and thought.
This is one of those games that just OOZES perfection.  He's obviously taking his time trying to make sure that this is exactly how he wants it to be, something I can respect given my own slow development tendencies.

On a side note, I made a custom sky for Tarkus and I'm hoping it finds its way into his game.  We'll see soon enough I suppose!

This is where you can learn more and see a huge volume of additional pictures:

5) Hunted: One Step Too far

Ok I wasn't going to include this one, but I have to admit it's promising.  He's taken a pretty basic concept (alone in the woods but not really) and made it into a very compelling visual thriller with a good combination of eye candy and suspense.   He's smartly using the dark to hide the flaws in Game Guru and as such it looks and acts very clean.

His only problem (and he's aware of this) are the mountains are a little bland.  I pointed him at my tutorial on building good mountains in Game Guru so hopefully he'll be able to bring that up to a level he's more satisfied with.  Overall though, a good looking game and also being greenlit on steam.  Nicely done!

The full details can be found here:

There are a few really superb ones I haven't seen get off the ground yet or seem stalled in their development as well:

Pretty much considered the high water mark.  Huge expectations here, but development seems DOA at this point.

Interestingly this game has a similar style as MLAW but that's not a bad thing.  The game looks good but I think the dev gave up and moved on.

My own project

I should mention that I hold myself to the same standards that I hold others to.  It's part of the reason my process is so excruciatingly long and torturous.  I take every scene and evaluate it against what I want, feel, and try to maintain a reasonable performance to boot.  It's difficult, at best, using this engine.  What you see is a very early work in progress.  I hope one day, it'll be more nearly what I consider acceptable.   This scene lacks clutter objects.  That object behind the glass is a ship.  A single, massive ship.  I'm looking forward to bringing this project to fruition.

Hopefully you'll take some hard looks at my blog and see some good lessons to draw from in developing your own projects.  I truly hope to raise the level of everyone's game development and now with 1.33 available, we might have a real door opening towards a brighter future with this engine.