obedient echo

a blog about echoes, and their obedience

[mahrib-fyp]

lighting

i’ll start off this post with the basic light bulb. there’s a light placed just below the mesh which is set to stationary, which gives it a dynamic aspect which lets it produce the really bright bloom effect

this light below is set to static which means it doesn’t produce the same bloom as the first one. you can see the spline mesh utilised here for the cable too, running down the rafter and down to the wall.

this was originally just created as a sort of test so i could throw some lights around the level, so you might notice this mesh in a lot of earlier posts. i originally intended to get rid of the light  bulbs in place of torches. the actual lights have been changed significantly however. i decided to keep this mesh in the end because i really liked the contrast of the torches and the light bulbs and how they gave this really harsh light. 

when you’re walking into a cell or down a corridoor with this light, and it’s only half visible because of the echolocation mechanic, the really harsh shadows and jarring white light make it all the more spooky because something about it just feels weird

here is the main light sources throughout the level though, the flaming torches. they provide light which is a lot less harsh that the light bulbs, and also a lot warmer. i didn’t think it was worth making a video just to show the torches, but they look a lot better in motion. there’s an element which looks like heat distortion as it rises. it’s subtle, but i really like it.

the actual blueprint consists of two lights. one of them is smaller, and stationary, which lets it flicker. i kept this one smaller because dynamic lights take much more resources to render in game so having it smaller is more efficient. there is also a much larger entirely static light which provides a lot of the fill light too. 

here’s the particle effect. in the content example particle level there was an example of black body radiation. this lets you specify the temperature and the colours will reflect that automatically. it’s a bit awkward to set up with the dynamic node in the particle system but it gives a really nice effect. 

and here’s the tiki torch version you can see in the blueprint above the settings which i’ve implemented to convert it to a tiki torch, which involves removing the dynamic light. during tests with the gpu profiler, i deduced having dynamic lights interacting with all the individual leaves on the trees is just a horrible idea.

 this is one of the outside sections. this bridge is lit pretty much exclusively by one light, which represents the moonlight. as a result you can see it’s pretty dark. only having one light source also helps to produce the same super harsh shadows as in the previous section. by having no fill light to fill in the almost completely black shadows it not only blends better into the black background, but adds an element of unknown. the player is used to clapping to reveal areas by now, so by having these completely dark section they cant reveal, it creates an uneasy ‘should i step here’ kind of feeling

i wanted to put a screenshot in here just to demonstrate the difference between the editor and the in game views. this enormous disparity made creating lighting for the scene extremely difficult, and incredibly time consuming, especially towards the end when i had to wait upwards of 2 hours each time i changed the lighting for a lighting bake to complete.

the sewer section was again a completely different lighting theme. this time i went for a very bluey green hue to the pipes, which tries to reflect the watery yet mouldy look of the cliché sewer. 

you can also see here a bit better than before the little ‘preview’ all over everything. this happens before the lighting is baked out. BUT for some obscure reason they don’t go away even after a light bake, which means the ‘lighting must be built’ error message was perpetually on screen while the game is running, which obviously cannot be a thing. i had to go into the engine files and disable the error message because even right after a light bake it was telling me i had thousands of unbaked objects, which literally just isn’t true.

[mahrib-fyp]

the sonar pulse was looking a little boring in its current iteration. the smooth edge just wasn’t working for me. in journeyman last year i remembered using the if node to create the rough edge on the fire material i created, so i figured that might help here. i basically took the combined opacity mask and compared it to the cloud mask texture i’ve been using for literally everything.

also in the video, if you've a keen eye you might have noticed the music box no longer pulses in time to the music. this is an unfortunate limitation of the engine right now. since it’s still early in the development, some of the plugins are still just tests. although i didn’t realise at the time, this includes the sound visualisation plugin, which means it literally doesn’t let me compile to an exe while it’s in use. i changed it for just a simple sine wave. this actually doesn’t look that terrible, and in fact might be better in some cases than the old solution because it keeps more visible for longer.

i also changed the way in which the constant emitters interacted with the rest of the sound emitters. the way it worked previously involved checking multiple times per second the same variables the regular sound emitters used and switching branches as needed. this however sometimes meant there were leftover parts on other branches which just kind of hung around. it’s now set so that i manually assign each emitter a number and it knows to use that specific branch. it probably would be better to implement a system that does this automatically but quite honestly i don’t have the time and since there’s literally 2 music boxes, manually assigning them numbers isn’t the end of the world.

[mahrib-fyp]

to help with the atmosphere, i wanted to add in some weather effects. as cliché as thunder and lightning and rain is in a spooky environment, clichés are clichés for a reason. i made the rain as a particle effect using gpu particles. this let me have thousands of particles simulated extremely efficiently. i wanted the rain to fall realistically, so i just added a constant downwards velocity which emulated gravity. 

this meant i could spawn the particles in an angle other than downward. in this case they spawn with an almost sideways initial velocity, gravity pulls them down, and the end effect is rain which looks like it’s being buffeted by the wind.

this is all placed inside a blueprint which also contains a cube that detects the player and can disable and enable to rain as and when the player is near, to save on performance.

i would have liked to have the video in this post after this image because y’know, spoiler alert, but unfortunately that’s not a thing.

but here’s the lightning i produced. it’s based off of the beam data example in the content examples particle effects level, but obviously switched up to fit my needs better. when added into a blueprint which adds the noise emitter and a thunder sound, as well as adding in a couple of clouds, it makes a pretty nice effect.

here's the clouds too if you were interested! it's basically just a cloudy texture which spawns super big. since it's at a huge distance from the camera, and it's only on screen for a couple of seconds, it doesn't have to look perfect

[mahrib-fyp]

LUT

as the level progressed, i felt like i should have a look at the colours before everything went to far and it looked weird. this helped me to kind of better understand what the final image will look like.

so i did some colour correction with a colour lookup table. this essentially consists of taking a screenshot and pasting it into photoshop along with the lookup table that can be found online (that crazy colourful squarey thing in the corner) and then using adjustment layers to make it look right. you then export the colourful squarey thing back out with all the colour adjustments and UE4 knows what the default lookup table looks like and it calculates the differences and uses that to colour the scene. its super cool stuff, really intuitive and user friendly too which is always nice. 

so you can see i made it look a lot less green, which makes it look a lot more realistic. the effect is vastly more intense in this gif that you could usually see, simply because the comparison is there. no one would have played the game before and thought ‘gee this is super green’, because theres nothing to compare it too in game, but at the same time, i feel like it helps on a subconscious level

[mahrib-fyp]

world machine

i finally got round to creating the mountain that was just being a perpetual eyesore in my level due to it being literally 8 times bigger than the entire rest of the level (so hard to miss) and also consisting of 3 cone BSP’s (so ugly as sin).

after failing repeatedly to produce anything with just the sculpting tools in zbrush, i tried a different approach. at first i looked for some mountain alphas i could just drag out in zbrush and then tweak to get the desired mountain.

when i was looking for said alphas though, i kept getting search results for the tools which generate these alphas in the first place, so i figured yeah i might as well go right to the source. i’ll get a higher quality image, and i’ll learn something new. 

enter world machine.

this is a piece of software which lets you generate entire infinite landscapes procedurally based on parameters you set. i started off with one of the example templates with the program as a base, and tweaked it until i got a landscape i was happy with.

when you generate a world in world machine, it basically creates an infinite random plane populated by these mountains and crevices and rivers and whatever else you specify. this means you need to explore this generated landscape and just find a section you like. so in this case i liked this square, it was perfect for what i needed (in terms of shape, not colour)

since it was suited more for a desert environment, it needed retexturing. i exported the mesh from world machine and imported it into zbrush. from there i could cut off the base of the mountain (since it brings in that entire cube from above) and repaint it. using masks base on crevices and a spray brush, i painted a snowy texture. since i knew my scene was going to be set at night, i made it a bit purple to kind of accentuate the lighting conditions.

after a simple decimation within zbrush, and a cleanup of the produced mesh in 3ds max, it was done!

i also used this technique for the courtyard ground, but world machine crashed just after i had exported the mesh, without saving, so because i had the finished mesh, it didn’t seem worth it to recreate the settings just for screenshots when it’s essentially the same as above.

[mahrib-fyp]

art bits

so i found a really cool feature in zbrush. i mean, i knew that there was a normal map looking material, but i never really thought about how it would be useful. turns out though if you sculpt on a plane (or in this case a cube but that’s effectively two planes) you can just apply this material and then capture a normal map directly from the document which, as it happens, is super useful!

i designed the corner tile based off of this image 

after exporting the normal map out from zbrush, i took the model into quixel to add some textures. you can see from the fact there are two parts to this diffuse here that the model is double sided, same as the steps, to make it a bit more diverse. there were numerous attempts are more colourful diffuses, but in the end i kept it pretty monotone looking to make it a little bit more realistic of the paints available in the sort of time zone it’s supposed to be set in.

i wanted to produce the floors here in a super modular fashion so that i could basically randomise it from just a few different pieces. i decided to go with a gutter to the side of the floor panel, but with a metal grate over so it was still walkable. in order to make this look right though, i needed to be able to specify the start and end, or i would just have a gutter ending into nothing. i switched up the array blueprint with this functionality, so i could set specific start and end pieces if i needed to, as well as flip them to keep the 2-sided thing i have going on with some pieces.

this is a blockout of the kind of floor i’m going for, and also a demonstration of the new array functionality. this entire floor section here is created using the array tool, with the start and end specified to the same reversible piece, and the end flipped to match up.

so basically. those old floor pieces looked like crap and i hated them. so, i made some new ones. i designed them base off of this image, which was actually from the same album as the corner one.

i essentially cut it in half to make different sections that could tile together to make an infinitely long corridor with different length ‘chunks’ and different colours.

i wanted to show off what is essentially the finished (for now) cell. you can see i’ve made a start on the wall decorations with the chain links, and then i’ve utilised the spline mesh blueprint with the chain itself. the cell is in a juxtaposition with the actual corridor in terms of textures and colours, in that it’s much more bland, its essentially all grey. i decided to do this because it’s a prison cell, and y’know, prison cells are boring and bland. so this equates to what is essentially just a concrete room. 

[mahrib-fyp]

knald

since i got accepted into expotees, i was looking at the past brochures from previous years to get a feel for the paragraph of text about my project i needed to write. instead of paying much attention to that, though, i noticed a couple of people had mentioned that they had used a program called 'knald' in their projects, so i thought i would check it out.

oh man, it's pretty amazing it has to be said. i found this (relatively low resolution) texture on CGTextures just to try it out, and the result ended up being so impressive I decided to use it in my level.

so on importing the texture into knald, you're given a bunch of options which let you adjust the amount of detail in the normal map, which it is generating on the fly. this thing is basically crazybump on steroids. it then generates the other maps i need, such as height/displacement and ambient occlusion. all while giving me a handy real time DX11 view of a 3D object with the texture applied where i can see things like displacement.

so then since my original texture was quite low resolution, i can take this newly generated normal map into Quixel suite with all the other maps it generated too in order to rebuild the albedo from scratch based on the topography

and this is the finished texture, from original source material to multi layered, multi mapped, tileable texture in literally about 5 minutes. 

and after setting up a material to use displacement in my scene, i can plug in all the exported maps from quixel and it runs in engine. i actually found out that i can plug in the sonar opacity to the tesselation multiplier, which means it only generates extra polygons for the pieces of the model which are in sight, which is pretty handy in terms of saving performance.

[mahrib-fyp]

optimisation

i noticed my level was running a little sluggishly. this might not be much of a concern for a static environment presentation piece, but for my purposes, i need a steady framerate of at least 60fps, preferably 75. so when i was getting under 30 i realised something was wrong.

thanks to the built in profiler in unreal engine, i was able to narrow down my search area to dynamic lighting. this was strange because i only have a few lights in my scene so far, and they are all static lights.

after having a look at the lightmap density view i noticed that all of my blueprints were yellow, and had no lightmap coordinates. this was troubling, since the staircase blueprint from the previous blog post generates a huge amount of geometry.

i did some more research online, and i found some obscure forum which said something about movable vs static objects.

BINGO!

turns out, when you create a static mesh in blueprints, by default it is set to moveable, which basically meant i had 100+ dynamic objects in my scene, all needing dynamic lighting.

quite the improvement.

as well as fixing the lightmap issue, since my view distance in the level is so shockinly low, it was a prime candidate for culling. i was actually planning on doing this if i had time, but the way i thought of was much more convoluted and less efficient and time consuming since i was going to detect distances and such within blueprints and kill off the actor and spawn it in at points like that. that meant i would have to have every object as a blueprint though.

but then i found the cull distance volume, which basically does everything i need and more, and much much much more efficiently than i could ever hope for.

this is the start of the level now, looking back from the courtyard in wireframe view. it's almost all gone, and the only reason those pieces are left is because they are still currently BSP brushes, once i switch them out the whole thing will be gone!

[mahrib-fyp]

asset specific blueprints

basically, i wanted to create some stairs, but i didn't want to go through the effort of placing each block manually. plus when i first made the stairs i didn't have all the blocks ready so i would have had to go through and manually switch out blocks and change rotations etc. in essence, all things i really could not be bothered to do.

so i made a blueprint to do it for me. it is essentially a recreation of the 'array' tool in 3ds max, but with editable parameters. it also lets me stick in a bunch of meshes and it will pick random ones from that. if i'm not happy with how random it is, i just play with the seed until i am.

when i created the steps for the spiral tower, i made them in such a way that if you flipped them over they would be a new type of step, so the second branch of the blueprint here does just that. since they are wedge shaped though, they don't translate well to straight stairs, so i left in the option to disable them.

finally, a teensy tiny blueprint i made for these doors, so i can click a checkbox to decide whether i want it open or closed.

[mahrib-fyp]

a tutorial, of sorts.

since the sonar concept is a fairly obscure one, and therefore somewhat difficult to grasp, i needed to provide direction to those players who don't know anything about the game, and those who don't immediately grasp the ideas behind the gameplay. in short, i needed a tutorial. 

but this isn't a tutorial in the normal sense. this is literally just a command, which hopefully when executed by the player will force them to understand what the left mouse button does. since the player will have already learned the right mouse to interact from the menu, this is the only remaining concept to teach.

so this screenshot here is essentially the entire tutorial. if after 3 seconds from the start, the player hasn't tried clapping, this will hopefully nudge them in the right direction. the text will differ depending on whether the player is using a keyboard and mouse, or a gamepad.

and this is all the blueprints required, all it basically does is calculate where the player is looking at the time, and puts a piece of text there, in a similar fashion to how the text on the notes work. once the player does clap, the text is destroyed.

[mahrib-fyp]

this time i wanted to develop a system which would allow me to really clearly show paths and guide the player around. i reasoned that just creating larger bubbles wouldn't be much help, since they're spherical so it offers nothing in terms of directional help. the only other alternative was to have multiple ambient sounds working in tandem to lead the way. 

i ended up updating the sound emitter blueprint so that i could turn it into a repeater. this allowed it to detect when another sonar pulse occurs in its vicinity and react by sending out a pulse of its own. this let me create chain reactions which could illuminate a relatively large area.

i needed to use a trace to make this happen. basically every time the pulse button is pressed, an invisible sphere is created around the player, which detects all the actors inside. if any of those turn out to be a sound emitter, it triggers an event which tells said emitter to pulse.

there's also an essentially identical trace inside the sound emitter pulse, so that it can pass on the sound.

choosing whether or not i want an emitter to be a repeater is as simple as checking a box, this means i can iterate very quickly and test different combinations of repeaters and none repeaters. the trace is set to be the same size as the sound radius, so that it will always match up.

since you've been good and read through all that, here's a special bonus. while the sonar material was really good for generic assets which just needed maps applying, or even just colours, it was getting ridiculous to add in parameters for each possible eventual possibility for everything i wanted to do with my materials. take this crystal for example, i wanted to use translucency, something which would have been awkward to do via parameters. 

so what i ended up doing was creating a material function out of the main sonar material. everything that makes the material work with the sound pulses is now contained within one handy node. this lets me develop much bigger, crazier, and more complex materials, i just need to add this one node in and it works. 

there is one more bonus too, but no picture this time. you can spot it in the video better than a picture could show it, but i have started working on some rudimentary particle effects. i want the outside to give off a night time feel, so i have added some blue ambient particles, which could just be dust, or fireflies or whatever. 

[mahrib-fyp]

I wasn't even actually going to give this it's own post, but you can only have one video per post with tumblr, so I had to.

ive basically just created a new blueprint which doesn't contain any visible components, but instead just has a sphere which checks for overlaps. once the player enters the area, it casts to an instance of sound_emitter and causes it to trigger using a special custom event. this bypasses the player near boolean set, so it will only fire once, and will not loop.

each trigger is manually linked to a sound emitter on a per instance basis. 

[mahrib-fyp]

implementing this was super simple, because it basically just required repurposing the system i already had in place for notes, but it now needs to run every frame to check if the player is looking at anything. so i use the same sphere trace as before, but this time i just hooked it up to a different custom event, and used a tick event to trigger it.

then to make it go back to the original material once the player stops looking, i just used a retriggerable delay which is reset at each frame the player is looking at it, so it can't complete until it stops reveiving triggers.

[mahrib-fyp]

layout

i spent quite a bit of time over the weekend developing a new layout for the level, and blocked the thing out over the course of the last couple days. it's going to be difficult to explain an entire new layout through images, but i'll give it a go.

it was suggested that i try and base the layout of the level off of the actual tower of london itself. now it would be crazy to think i could create the whole tower, so i chose a section of one of the corners, and even though it doesn't conform exactly to the layout of the tower, i'm using quite a few of the walls and forms that it dictates. i'll definitely be making a number of tweaks to this layout over the last month of the project, but this should hopefully be the final layout. 

i'll try to walk you through based on what you can see from this top down view, and hopefully the subsequent screenshots will make sense and fit into your imagination. so the player starts towards the top of the map, in the prison cell just to the left of the top tower, and moves through the corridors and up the stairs to reach said tower, passing a number of other cells on the way. (the red numbers indicate floor height) the player then climbs this first tower and exits onto the top of the wall, where there will be a couple of special events. on the first parapet as the player looks out over the landscape lightning will strike the mountains and the corresponding thunder will create a large bubble of vision.

entering the second parapet will activate the bell tower, which will also create a large bubble of vision in which the player can see a bit more of the tower.

moving onwards, the player drops down the second tower into a pool of water at the bottom, before exiting out into a courtyard like area. a tree lined path will guide the player towards the entrance to the sewer section which will double back on itself and go underneath the courtyard. a door at the end of the sewer will be the end of the area.

the prison cell starting area

mountains, revealed by an event triggered by standing on the first parapet.

you can see the bell tower, as well as the courtyard. you can also see the route the player takes down the second tower and to the entrance of the sewers.

the sewer doubles back on itself to avoid there being too long of a straight line.

the sewer section, which will be walled in. i am also planning on altering this slightly to add a bend in there. 

[mahrib-fyp]

now that i basically have everything crucial implemented as far as sonar related goodness goes, i needed to get on and do some art related things (this is an art course after all.) so since i'm doing a catacombs style environment, this means there's gonna have to be a LOT of stonework around; stone blocks, bricks, rubble etc. so i thought i might as well get started with some practice for that.

i wanted to make something that it doesn't really matter if it turns out a bit rough, since its a first try, so i made this gutter arch thing, which will always be fairly low down and out of the usual line of sight. i went through and sculpted each brick, added texture etc, and baked normal maps to kind of get used to that process.

as part of the sonar shader, the colour all fades out after a few minutes leaving the model grey, but the normal maps are still visible. because of this, i want to kind of exaggerate the normal maps more than i would normally, so that the detail still pops. it is intended to look similar to the 'detail lighting' view mode within unreal engine 4, which i always did think looked kinda cool.

i textured this using the quixel suite, which is quite frankly amazing. at the moment i am using some of the supplied textures, but the software makes it super easy for me to change that out later once i've had a chance to take my own photographs. the most useful part of quixel suite is its ability to make masks. it can generate procedural masks based on the topology of the model, or based on maps that you supply it, which you can then modify the settings of to make it perfect. so in this asset you can see i added a lighter texture around the edges (which will probably get toned down a bit) and some moss within the cracks and crevices where water would perhaps collect.

[mahrib-fyp]

so to go along with the sporadic ambient noises which create temporary vision bubbles, i wanted to create a system which would let me create sources of constant noise, toggleable by the player.

originally i anticipated having to go through and create curves which matched up to music manually, but while looking for a way to automate the curve making, i actually found out about a plugin included with UE4, but disabled by default. this gives me two new blueprints nodes, one for frequency analysis and, more importantly for me, one for amplitude analysis.

this lets me analyse sound files in real time during gameplay, saving me hours of tedious curve creation.

before i show you that i'll just quickly explain how it's triggered. basically is uses the same system as the note. you just look at it, and right click. i want to keep the controls as simple as possible. hopefully all i'll need is left click to pulse and right click to interact throughout the entire thing.

here's the fun part. so basically what it does is uses this get amplitude node, and i just set an amount which is approximately the maximum amplitude as the constant high variable. the actual amplitude is then divided by this to get a value between 0 and 1 (ish).

it then checks to see if the current amplitude value is higher than the one currently already in use by the blueprint, and if it is it will set it to the higher value. there is also a loop on the left there which constantly shrinks the bubble at a slow pace. this is for two reasons, firstly it means the bubble will eventually go away, clearing up a branch for other pulses, and secondly, it makes it seem more animated.

i decided to do it this way instead of just straight up setting it to the amplitude value because a lot of the sounds i will be using are music boxes, which have periods of silence between *tinks* during which the visual bubble would be tiny, which completely defeats the point of a constant sound source.

and of course then it needs to set the material parameters, with multiple redundant branches just in case. here's 3 of them but if you've looked at other posts you'll get the idea.

[mahrib-fyp]

so i fixed that problem that i was having wheeee! i the theoretical fix i proposed in that other post pretty much worked, except instead of multiplying the greyness by the sphere mask, i used a lerp

i also added in a bunch more customisation options for the material instances parented by this material. this lets you add in things like specular, roughness, normal maps etc. you can choose whether you want to use a single block colour (as well as what colour to use), or an image texture. super simple stuff, and easy to implement, but i thought i might as well finish this off before i stop faffing with this shader.