[ONG] Offroad Racer - Indie Game Dev Log
sublime tool
Latest Image:
Hi guys,
we (my brother and I) are working on a game project, which obviously will not be called offroad racer on release. We would like to share some insights and behind the scenes because basically, why not. There is no big company attached ... no nda´s and stuff. I learned so much reading this kind of topics on polycount that I feel like maybe this will help someone somehow or at least be entertaining for the real game developers who will probably laugh at our mistakes.
....well, who are you people?
We are 30 and 28 years old and have been working in military software development in germany for some time. I did a brief stint in game dev but sadly game companies in germany tend to like browser games very much.
....so?
We love games and we love developing games and the technology behind them even more. So we both quit our jobs and are working on this project fulltime now .... YAY!
....good for yeee! How is this supposed to work then?
We are planning on releasing our game on steam later this year. Since we are only two guys doing all the work and also developing our own engine this is a massive task.
....wait, you chumps are developing your own engine??
Developing the entire engine will be great, my brother said. It will be fun, he said. Yeah, but seriously it already starts paying off with the suspension animation, driving physics, wheel blur .... local splitscreen playing online and whatnot. But I dont want to bore you with details at this point.
....yeah, now show some pictures! And after reading your freakin life story they better be good -.-
Thanks for reading and keep in mind that we are still very much working on this so most is alpha, alpha. Like proper alpha!
High-poly renderings (more on artstation):
The car is pretty much modular. This is a bake check of the mainframe module (low-poly plus normal) baked with knald:
I purchased this awesome kit http://mvhaitsma.artstation.com/projects/Wdzqv and also the one with the cables.
And used it whenever possible:
I also purchased the "325 hard surface alphas". Unfortunately I couldnt use them as much as I had hoped, but still.
And here are some ingame shots (ignore terrible sky):
suspension at work (ignore terrible ground texture):
driving with oversteer through power and opposite lock:
suspension rigging:
car inside engine editor:
Will post more in the future!
we (my brother and I) are working on a game project, which obviously will not be called offroad racer on release. We would like to share some insights and behind the scenes because basically, why not. There is no big company attached ... no nda´s and stuff. I learned so much reading this kind of topics on polycount that I feel like maybe this will help someone somehow or at least be entertaining for the real game developers who will probably laugh at our mistakes.
....well, who are you people?
We are 30 and 28 years old and have been working in military software development in germany for some time. I did a brief stint in game dev but sadly game companies in germany tend to like browser games very much.
....so?
We love games and we love developing games and the technology behind them even more. So we both quit our jobs and are working on this project fulltime now .... YAY!
....good for yeee! How is this supposed to work then?
We are planning on releasing our game on steam later this year. Since we are only two guys doing all the work and also developing our own engine this is a massive task.
....wait, you chumps are developing your own engine??
Developing the entire engine will be great, my brother said. It will be fun, he said. Yeah, but seriously it already starts paying off with the suspension animation, driving physics, wheel blur .... local splitscreen playing online and whatnot. But I dont want to bore you with details at this point.
....yeah, now show some pictures! And after reading your freakin life story they better be good -.-
Thanks for reading and keep in mind that we are still very much working on this so most is alpha, alpha. Like proper alpha!
High-poly renderings (more on artstation):
The car is pretty much modular. This is a bake check of the mainframe module (low-poly plus normal) baked with knald:
I purchased this awesome kit http://mvhaitsma.artstation.com/projects/Wdzqv and also the one with the cables.
And used it whenever possible:
I also purchased the "325 hard surface alphas". Unfortunately I couldnt use them as much as I had hoped, but still.
And here are some ingame shots (ignore terrible sky):
suspension at work (ignore terrible ground texture):
driving with oversteer through power and opposite lock:
suspension rigging:
car inside engine editor:
Will post more in the future!
Replies
This is the mainframe module albedo texture. We use a color key method after dxt compression to pick basically any color for the colorable (green) parts without loading different textures.
I have no idea how to do these effects properly so I downloaded a video of a bonfire and extracted few frames and built a simple atlas texture. The texture is LDR so we use an emissive value.
The atlas is mapped to a simple poly cross (which looks ridiculously stupid btw). Will have to do something about this.
The car consists of five modules. Texturing was done using quixel. It took about 50-60 hours.
- cockpit
- mainframe
- suspension_front
- suspension_rear
- wheel
(AO rendered in Max from this scene. Thats why the inside of the wheels are missing. Didn´t want AO on the suspension parts)
Every module has its own texture and was textured seperately using Quixel.
Therefore I assembled a small library of custom materials inside Quixel so that every module looks similar.
(16 custom mats have been used for the entire car. Including dirt and sand)
Obviously I used mat ID´s to apply the base materials automatically
This is what the mat ID´s look like. I did a quick render to texture inside max with scanline renderer to translate them into input maps to use with quixel.
UVs were auto packed using maya. After months of retopology I REALLY did not want to do this manually too.
I´m just throwing stuff in that might be interesting. We have tons more. Let me know what you think or what you would like to see. C&C Welcome !!
Doing actually working offroad suspension was pretty difficult. All the pivot points and length have to match. I did tons of research on this topic. Watching videos about camber change on suspension travel and whatnot.
Also the camber should increase slightly as the wheel travels up as seen below. At first we encountered a lot of negative camber on wheel travel which looked really goofy because the upper and lower control arms were not parallel.
The animation is split inside the engine and then blended according to steering input and physics (spring compression)
With an open wheel car like this we decided that it would be mandatory to have a proper radial motion blur for the wheels. It turned out really great IMHO. We are using a custom blurcage mesh that fully encloses the wheel to achieve this.
The blurcage could use some polycount fix up
c&c welcome as always!
I'm in the same boat as you, I've built my own graphics and physics engines for my first mobile game. I agree it sounds great at first, and simple enough, but the work is just relentless - even once it's working well, you realise you have to implement every extra little thing you want
How does the blurcage work btw? This is the first time I've heard the term.
Here is a small glance at the first race track level. My personal best time is 1:22 min so far
thank you!
My brother does all the programming. I´m just the artist. I will ask him to answer your question about the blurcage. We use the bullet physics engine but he heavily modified it.
These are the exposed parameters to setup the car handling:
just want to show you something thats barely visible in-game but quite some time went into this.
screenshot from 3do while quixeling this up:
We saw this concept around christmas 2013. We immediately thought about how cool it would be to have a racing game with this kind of badass cars. So I started blocking something out that looked like it.
artwork from station02: http://www.3d-ring.de/3d/3d-galerie.php?img=3749
The first blockout looked like this (March 2014):
We decided to have a proper V8 engine and get rid of the round shaped bonnet (February 2015):
I didn´t know what to do with the strange wing on top and also wanted the exhaust to be more badass mad-max style. I bought few kitbash-sets and used parts to quickly come up with new shapes. We also decided to ditch the flat rear suspension and go with some chunky dampers (December 2015):
I spent a lot of time modeling the frame and the engine properly and fitting everything together (June 2016):
We decided the blockout tires are way to flat and have to go. I took two weeks off from work and really upped the pace on the high-Poly (June 2016):
New tire is finished (early August 2016):
The suspension high-poly is finished (end of August 2016):
I took another week off from work to finish the engine high-poly (early September 2016):
September to Dezember 2016 I spent retopologizing literally every free minute after work and every weekend.
Christmas 2016 the retopology is finished and I take another week off from work to unwrap this sucker and bake everything.
Since 01.01.2017 I´m working fulltime on this project. The texturing / rigging / animation / went down in January 2017.
I´m guessing a total of 6.000 working hours spent just on my end so far.
My brother spent at least 10.000 hours on the game engine.
Yes it has been a ridiculously stupid amount of work so far. I did 9-5 for the cash and 6-2AM of this after. Spent almost every weekend and used all regular vacation and four weeks of additional unpaid vacation 11-4AM every day.
Thats why we had to go fulltime. After some time its just impossible to carry on.
I will get back to the car but would like to show you construction of racetrack #1 for now.
There still is a lot of work to do but we got rid of the dynamic blue sky and added a nice HDRI skydome that added a lot of atmosphere already.
I started drawing a track layout with photoshop. Took it to 3dsmax and did a quick reconstruction using bend modifier on basically bunch of planes.
I sculpted in the height differences that I wanted and did a lot of play testing to make sure every corner is rewarding and a little bit challenging. Also to ensure the whole track has nice variations and is not to much of a challenge if that makes sense.
... rendered a heightmap of that.
Plugged that into Worldmachine. Did a simple overlay and painted the ridge and mountains in Worldmachine.
I´m using the AMAZING TeTo texture toolbox by Iri Shinsoj for splatmap creation and a unique basemap for large scale color-variation.
https://www.artstation.com/artwork/teto-the-texture-toolbox
I took the terrain to ZBrush after because it lacked a lot of detail right next to the track. There were nice big mountains and stuff but no mid size details near the track. This is what it looks like after few hours of ZBrush treatment:
Now I´m retopologizing everything because it´s basically a heighfield and we dont want that for performance reasons. I strive for a nice transition between track and terrain a la GTA5. Already spent hours examining that
Most panels on the car are rigged so I hope we will be able to shed off parts and have visible damage soon.
The gorge needs a bridge that has not been modeled yet. Collision only for now.
... overcooked it while catching up
This looks really great so far and i love games like this.
I just wanted to explain a little more regarding the blurcage. The basic idea i got while reading "Shader X6 Chapter 3.4 Per-Pixel Motion Blur for Wheels by Damyan Pepper from Black Rock Studio". It describes the concept of a radial blur very well. They use a cylinder as proxy geometry for the blur direction, but for us this wasn't sufficient. So we turned up with our so called blurcage as shown above. The blurring itself is more modern and similar to "SIGGRAPH 2014 Next Generation Post Processing in Call of Duty: Advanced Warfare". We render the wheel like everything else in the deferred path while marking it with a special material id. Afterwards we copy just the wheels to a half resolution buffer. From there it gets blurred and blended over the scene in one go with the blurcage used as proxy geometry.
Regarding the over and understeer we tried many many things including first hand real world driving with an FR drive train and much power
Did you and your brother thought about doing a track editor directly in the engine so that you don't hassle too much if you need to do quick changes ?
Thank you
Yes we thought about a track editor. At this point it is possible to vertex-snap track modules together in engine.
We decided against a "full on track editor" because its a lot of work and we dont consider it that beneficial at this point. We would like to give an open track editor to the gamers but thats far down the road.
Our priorities are:
- gameplay (fun single- / multiplayer online game)
- stability and performance (linux / windows, 4k resolution)
- the best gfx we can achieve (need to stick out on steam)
- get it out on steam 2017
Everything beeing equally important we have to be careful with new features and cut down on quantity. We think / hope that this is more gamer focused and hopefully will be rewarded in the end.
have been working on foliage and environment lately. Also did some new screenshots today. More over on ArtStation https://www.artstation.com/artwork/A5GVy
I caught it up on ArtStation, will follow more on Polycount now.
ah damn, now I feel bad for not using Substance on this one ...
It is in fact at home under Linux. My brother did a windows port for the editor and game. I hope my brother will post a little bit more details about the engine.
i thought a while about what i could write regarding the engine.
It is designed as an in-house tool and neither planned nor built for public use.
Maybe we open it up one day. I don't know. I am a strong supporter of linux and open source myself.
So here are just some features and techniques to give you a rough idea.
It is written in C++11 and has a Python interface for scripting.
The complete editor is scripted in Python using Gtk+3 as the widget set.
It has a versioned document supporting full undo/redo as well as file version migration.
It supports full online gaming including late join and spectator mode.
We have a working matchmaking server under linux with clustered game servers.
It supports local and online split screen.
It runs under windows 7 and linux using OpenGL 4.5 core as render API right now but Vulkan/D3D is possible and planned.
It has a fully threaded renderer with hyper aggressive batching.
Uses bullet for the collision detection and rigid body dynamics with a unique vehicle physic.
Features full 3d sound with steaming audio and video recording and playback at 4k with audio/video sync
and more...
The renderer is a deferred renderer with a full forward path for transparency.
It supports:
* High dynamic range rendering
* Physical based shading
* Contact hardening soft shadows
* Screen space ambient occlusion
* Screen space reflections
* Clustered dynamic lights
* Image based lighting for global illumination
* Volume particles
* Temporal anti-alias
* Skinning with non linear animation blending
* Motion blur
* Animated foliage
* ...
I am currently working on the atmospherics (around three unsuccessful attempts so far and counting
The engine in numbers, just for the fun:
563 c++ engine files
163 shader files
69 c++ files for the server farm
14 python files for the editor
17 python files for the game
all 826 source files together have around 130,000 lines of code, not counting the 24 third party libs and the ~167 files for the win32 and linux build system (has to build the 3rd party libs too).
And all this just to produce this image in under 16 milliseconds at 4k. Behold i give you THE ENGINE 4.5L V8:
Worth it
Honestly? I have already written many programs in my life and many of them for the military and agencies. I don't know why humanity needed them. What kept me going with this project and programming as such has always been the fascination for art. Whenever i lost motivation i browsed the fantastic artwork you guys do and immediately felt the inspiration and enthusiasm to write a kick ass renderer again.
Feel free to ask any questions.
Excellent community spirit shown with you guys sharing your development endeavours.
As a long time developer, mainly in the racing genre, this is very inspiring to me personally and I am very envious of the fact you are working together as brothers. Perfect scenario.
Keep up the great work and the excellent posts to this thread.
Again all the best for final launch
I am Fahad from Artstation btw
Do you have a twitter account? I would love to be kept up to date with this project,
Grüsse aus Warthausen !
@Ckafee: Thank you!
I recorded a video "tutorial" about the suspension rigging & animation today. Sorry for my awful english
no, typically everything gets baked to keyframes when exporting as FBX file. The engine doesnt have to deal with IK or anything. On export there is one key placed for every frame with all bone tranformations as fixed values.
ehrm... lets make a very simplified example:
Front right wheel hits a rock. Physics "says" the front right wheel has traveled up 20%. Engine plays animation clip "spring_FR" to frame 15 which equals 20% up movement. Of course this happens live so the suspension matches to the wheel at all times. Now lets say the player is also steering to the left at the exact same moment the wheel is hitting that rock. Now we also need to play animation clip "steer" to frame 63 but since the wheel has also traveled up it needs to be blended with the animation clip "spring_FR".
So in other words the suspension is just reacting to the physics.
It is a lot more complicated than that but lets keep it simple for now.
So how do you get all this into the engine together with real physically moved and steered wheels?
The way it works is basically this:
The whole suspension with the spring and all gets set up and animated in 3ds max. Nothing of this is exported directly to the engine. Instead the full travel way for steering as well as suspension movement is baked as ready made key frame animation data together with the bones. This gets exported the exact same way as mocap data. It is a skeleton of bones with there associated vertex weights plus one list of key frames consisting of one transformation per bone per key frame. So you can simple play this back in the engine. Just hit play and you see the suspension moving from full up to full down and afterwards from full left to full right.
The individual clips get set up in the engine so the scripting knows from which key frame to which you have which animation. You can see this on one of the screenshots above.
Then comes the trick:
One special helper point is exported like a normal bone together with this data. This special bone is detected by the engine by its name. It follows the wheel throughout the animation. So it always represents the final position and orientation of the wheel axel for every key frame. The engine now creates a mapping from key frame to final axel transformation by playing back the animation once and recording the position and orientation of this helper bone. If you now inverse this mapping (if you want by searching through it for a specific result value) you know which key frame you have to use to get lets say 10% steering left. This way you can take the needed wheel position from the physic and set the fitting key frame from the animation data. Voila you have a traveling suspension.
Now comes the problem:
It gets tricky when you want both, traveling suspension, as well as steering at the same time on the same wheel. Then you have one key frame for the suspension to travel 20% in and another one for it to steer left 10%. You have to use them both at the same time now. This is no problem as long as the two animations affect a different set of bones like for a character when the run animation is played on everything below the waist and the shoot animation on every bone above. In our case this is unfortunately not a solution cause both animation frames affect exactly the same bones.
And the solution:
We export a special neutral position on key frame 0. This neutral position has no steering and mid suspension travel. It doesn't belong to any animation directly. It is just used as a reference. When we want to have an animation that does both, steering and travel, we do the following. We subtract the bone transformations of the reference pose from our steering and travel animation transformations. So we have two sets of bone transformations representing the delta between the reference pose to the wanted steering and travel pose each. Now it is just adding everything together. Reference pose plus steering delta, plus travel delta. Voila you have both at the same time.
This works well but not 100% exact. To not have the wheel look a bit loose on the axel but instead like a fully rigid thing we place it at the helper point and not at the "should be position" that was coming from the physics and was our original input to all of this. This has the nice side effect of giving us the correct camber and inward movement which is not provided by the physics engine.
Hope i didn't bore you with to much technical stuff here
@orontes Great explanation, didnt bore me at all! You guys earn respect:D