BeamNG.drive VFX & World Overhaul Making BeamNG better in every way This thread is intended as a central place to track AI-assisted BeamNG.drive mods and experiments that are currently in development. The idea is not only to post my own projects here, but also to let other developers share what they are working on. One important reason for this is coordination. We are already starting to see different people independently developing very similar systems. If we can see what everyone is working on, developers can communicate, exchange ideas, cooperate where it makes sense, and avoid spending months creating duplicate projects. Users are welcome too. If you find an interesting active BeamNG project somewhere — GitHub, Reddit, the BeamNG forums, YouTube, Discord, etc. — feel free to post it here so we can keep track of what is being developed. My own mods and experiments are planned to remain free. What I am currently interested in improving My goal is not one giant mod, but a collection of systems that can gradually make BeamNG's world feel much more alive, interactive and visually modern. Interactive world / environmental physics Add physics and interaction to parts of the currently static world. Objects such as barriers, fences, props and other scenery could react to vehicles, fall, move, collide and potentially be destroyed instead of always behaving like completely static map geometry. Water overhaul Improve BeamNG's water visually and interactively: better waves and moving surfaces; flowing rivers; vehicle wakes; foam around objects and vehicles; splashes and interaction with wheels; underwater effects; camera water effects; cleaner or muddy water variations. Vehicle / environment interaction Make the car react more visibly to the environment: mud and dirt accumulation; wet vehicles and wheels; dust; water interaction; better surface-dependent effects; improved contact effects between the vehicle and the world. Dynamic terrain Experiment with terrain that can actually react to vehicles. Examples include mud deformation, ruts, displaced soil, compacted ground and more convincing interaction between tires and soft terrain. Rain and wet-weather overhaul Improve rain far beyond simple particles: water on vehicle bodies; windshield rain; moving water streaks; wet roads; puddle/ripple effects; tire spray; camera droplets; better interaction between rain, vehicles and the environment. Characters Add proper characters to BeamNG with skeletal animation, different animation states and interaction with vehicles/world objects. Pedestrians and ambient life Eventually expand characters into people that can walk around maps and make cities and environments feel less empty. This could later include basic pedestrian AI, reactions to vehicles and simple world interactions. Trains I recently started looking more closely at the existing train projects in BeamNG and ended up spending quite a bit of time actually playing with them. BeamNG already handles trains surprisingly well and they are genuinely fun, but there is still a lot that could be improved. Possible areas include better railway physics and collisions, improved wagons and couplers, switches, railway interaction, derailment behaviour, effects, sounds and eventually more complete train gameplay. Particle and VFX improvements Improve how particles interact with vehicles and the environment. For example, smoke, dust, grass, rain and other particles should not simply pass through the vehicle cabin or body. Effects should understand where the vehicle actually is and react accordingly. Environmental destruction Push static-world interaction further into actual destruction. Not everything needs full soft-body physics. A lightweight rigid-body/destruction system could make fences, signs, barriers, props and selected map objects behave much more like modern open-world games while keeping performance reasonable. Surface overhaul Different surfaces should look and behave differently. Mud, sand, gravel, asphalt, grass, snow, water and other materials could influence: particles; tire appearance; deformation; grip effects; vehicle dirt; sound; water behaviour; visual feedback. Better world reactions One of the larger goals is making different systems communicate with each other. For example: Rain → wet road → wet tires → tire spray → dirty body → puddle interaction or: Vehicle → soft ground → terrain deformation → displaced mud → dirty tires/body Instead of isolated visual effects, the goal is to make the world behave like one connected system. This thread can also act as a project index. If you are developing something similar, post it here — even if it is still an early experiment. If several developers are working on related ideas, it may be much more useful to share research and solutions than to unknowingly solve exactly the same problems separately. And if you are simply a BeamNG player and discover an interesting project in development, feel free to post it here as well. The overall goal is simple: make BeamNG's world more alive, interactive, dynamic and visually convincing.
theres 2 out there at the moment but the one in repo hasnt been updated since april or since release pretty much. and the other is messed up atm from .39 i would say go for it. does look like ur taking much of a performance hit i like it.. u ever need another tester u can message me here or discord @clap99 i try to help give modders more feedback from another user.. so up to u and also when its ready lmk just throwing it out there
I’ve checked out all the existing BeamNG.drive mods, which is why I decided to make my own. Regarding that recent mod that lags and causes grass to flicker: it simply scripts the standard F11 editor brush to deform geometry in real time. While clever, it’s far too performance-heavy. Editor brushes are built for one-time level design, not runtime operations. The grass flickering happens on the editor level as well—you can see it yourself just by painting terrain with that brush. It’s simply the wrong tool for the job. The checkmark for mud physics is *Spintires* and *MudRunner*, but their tech is custom and complex. Instead, I’m using standard deformation techniques found in modern engines like Unreal or games like *Red Dead Redemption*. They are standard in theory, but implementing them from scratch in an engine with no native support is a real challenge. That’s where guiding the AI comes in: I supervise it closely to keep it on track. Testing is handled quickly and automatically. My main bottleneck is token limits—the \$100/month Pro tier is too steep, so I'm using the \$20 one. Without these limits, development would move much faster. Over the past four months of exploring BeamNG.drive, I've also experimented with other concepts: - A real dynamic rain system:** Spent two weeks on a prototype, but scrapped it due to performance issues; I'll rewrite it properly later. - A playable character:** A proper, fluid on-foot controller, unlike the clunky ones currently around super early nothing to show - Physics expansions:** Example video on my channel. For now, I want to stay focused on one feature and avoid scope creep. **P.S.** Of course, with unlimited compute and resources like Elon Musk have, you could build practically anything.
dude sounds great. def go for it man lmk if i can hepl in anyway with testing. I cant mod but i support ya ! love to be a help to you in some way.. cant wait to see ur progress!
Some super early testing with dirty wheels. I want to make some kind of system that paints the wheels when they drive over specific parts of the environment (wet, dirt, snow, sand, etc.), and maybe the vehicle body in the future (I've already spent some time on this, but have nothing to show for now ). *Want something like in forza 5-6
Honestly, this being developed by someone with actual game dev experience puts it a step above most vibe coded projects for BeamNG. As long as you monitor performance like the Radial Blur mod maker did, it shows you put in effort. Just don't charge for the mod or use some closed-source DLL injection, once it's done. Can you go into more detail on what the tech differences are between MudRunner and "standard deformation techniques"? All I know about MudRunner is from this blog explaining how the mud and water works, but it could be interesting to decompile the game if you own it to see what optimizations they made. Interesting that the mod currently seems to colour the sidewalls of the tires but not all of their tread. I would expect it to be the other way around.
I wasn’t really planning to charge for the mod. For me, the interesting part is simply managing to create something I’ve wanted in BeamNG for years. After BeamNG I can barely enjoy other racing games, but visually there are still many things I’ve been waiting years to see improved, so eventually I decided to try doing them myself. I might add a donation link though. I was affected by the recent industry layoffs, so right now I have the free time to work on this, but I’m living off my previous savings. Development also eats AI limits incredibly fast — I currently keep 1 GPT Plus + 2 Gemini Pro subscriptions just to continue working when one runs out. GPT Pro with higher limits would be ideal, but it’s quite expensive for Ukraine. About external libraries: it really depends on performance. My experimental physics system, for example, uses Box3D: Some things make sense in Lua + GPU shaders, while heavier calculations can benefit a lot from C++. I have GitHub repos for the different stages anyway, so eventually they can be made public and people can see exactly what the libraries contain. I’m also not really worried about people taking parts of the mod — the systems are becoming interconnected enough that it would probably make more sense to contribute to the project than rip one system out. Ground This is probably the most interesting part for me right now. BeamNG terrain already has something very useful: the actual physical ground surface and the visual terrain material don’t necessarily have to appear at exactly the same height. On some soft surfaces, when you drive over them, the wheels visually appear to sink into the material while the underlying vehicle/terrain physics still works normally. My current deformation prototype uses a POM-based material, but right now it’s basically just a test square placed in the scene. It proves that the deformation itself works, but it’s not yet clear how useful it can be until I integrate it into BeamNG’s actual terrain system. The real victory would be replacing BeamNG’s existing visual ground material — the material the wheels visually sink into — with my deformable material while keeping BeamNG’s existing terrain physics underneath. If that works, then we already have the important foundation for something approaching Spintires-style terrain without trying to replace BeamNG’s actual terrain physics. The final implementation could use deformation/height maps + POM, or possibly tessellation if that turns out to be practical. Current systems Wheels: universal dirt/mud accumulation. Dirt should follow the tread and gradually decrease toward the center, while wet mud can cover much more of the wheel. Ideally it should also have some visual depth instead of being just a recolor. The version I showed was only a proof of concept; I already made a better one, but it became too expensive performance-wise. Body: universal rain, mud and dust accumulation on vehicles. Weather / world wetness: rain shouldn’t affect only the car. Roads, terrain materials, decals and other surfaces should gradually become wet too. A lot of this could potentially be done by globally driving existing material parameters — mainly reducing roughness, increasing/specifying specular reflections, darkening the base color slightly, and where appropriate adjusting normal-map strength or other wet-surface properties. This could make the whole map react to weather without requiring completely new materials for every surface. Wipers: probably based partly on ideas from the Animated Wipers mod, with improvements shared back with its author. Glass: rain/water effects inspired by Assetto Corsa, Unreal and Unity implementations. Ground: Spintires-style visual deformation integrated into BeamNG’s existing terrain/material system. World physics: arcade-style destructible objects. BeamNG beam-based objects are too expensive in large quantities, while my Box3D test already handles 620+ objects. With pooling, potentially much more could be possible. Character: GTA-style entering/exiting vehicles. The current problem is that the 3D character is not synchronized with BeamNG’s vehicle physics. This may be related to the character being rendered as a regular 3D object while the vehicle is a BeamNG beam structure running through its own high-frequency physics simulation. If I can’t solve that synchronization problem, there probably isn’t much reason to continue this part. The biggest difficulty with AI development is debugging. AI can generate extremely complex code very quickly, but once the system becomes large, even the AI can lose track of what it changed and repeat the same mistake. So a lot of my time is basically: change something → launch BeamNG → drive/test → find a problem → fix it → repeat. That’s why I’ve now split the project into separate systems and work on each one independently. P.S. I’d also love to investigate BeamNG’s Particle Editor 2 eventually. Coming from VFX, I find it very limited compared to systems like Unity Shuriken. Something closer to modern Battlefield-style smoke/fire using motion vectors and modern particle shaders would be really interesting, although that might also require an external solution.
Right now, I can’t even really run everything together myself. I have two PCs, and each one has different parts and work-in-progress versions of different systems. So if I start working on something on one PC, I basically have to keep working on that same PC. At the moment, everything comes down to my ChatGPT subscription, because I already pay for Gemini — actually, I have two Gemini subscriptions. So Gemini is more reliable for me when it comes to usage limits and how quickly they reset, but I feel GPT is better for this kind of work. That’s why I’m doing the mud/terrain system with GPT. For example, the terrain system — something like the ground deformation in Spintires — has already been waiting for about a week until I can continue working on it. The limits of the regular subscription just aren’t enough, so sometimes I have to wait five days. If the limits were unlimited, maybe something would already be finished by now, because I’m ready to test things and I already have ideas and further improvements prepared — but the AI isn’t ready because I’ve hit the limit. I’m also doing everything gradually. You can’t just type one big prompt and expect it to work, because it can easily ruin everything. You have to be very careful with it, make small changes, and keep everything under control through Git. And I’m spending around 10 hours a day at the PC working on all of this, so it’s not really a lack of time or motivation on my side. The biggest bottleneck right now is simply the AI usage limits. As long as I’m not working, I’ll keep putting as much time as I can into these projects. Once I find a job, I probably won’t have nearly as much time to dig into all of this. If that happens and I can’t continue developing everything myself, I’ll just make the Git repositories public so other people can pick them up, experiment with them, and continue the work themselves. While I’m waiting for ChatGPT limits to reset, I switch to other AI tools and work on something else in parallel. But obviously, losing focus like that isn’t great. I might come back to the terrain system later and forget something, or start doing something differently because of the interruption. Once I set up a donation page, maybe people will throw in something toward a yearly top-tier subscription — Pro, because right now I only have Plus. We’ll see. But first I need to make a bit more progress so there’s actually something interesting enough for people to support. By the way, yesterday I made some test characters inside the car. It looks like I managed to get rid of the jittering. After that, another issue showed up with motion blur, but that can also be fixed simply by disabling it, so it’s not a big problem. The wheels, though, still aren’t going very well. It seemed like such a simple thing at first, but they keep glitching. And again, the main challenge is making the whole system universal...
That's actually awesome and I'll make sure to follow progress with great interest ! Can't wait to see more of this !
a little update on GTA style system, that i'm working in parallel I'm trying to build a GTA-like character system for BeamNG.drive — allowing a character to walk around, enter vehicles, and get back out. The characters are not the standard BeamNG Unicycle / JBeam “snowman”. They are external 3D character models. Currently working: • Basic debug character movement • Entering vehicles • A system for creating seat positions inside vehicles Ideally, each vehicle will need a small amount of setup to define its seats and entry positions, but this should be fairly easy to do. The next step is improving the character controls and camera system. Right now both are still pretty buggy and very much in the prototype stage. The long-term goal is to have a GTA-style system in BeamNG.drive where you can freely walk around, enter different vehicles, drive them, and get back out. Still very early development, but the basic system is already starting to work. PS oh so many limitation, now i'm noticed that character animation not play in cycle(loop), so now i must somehow create locomotion/state mashine by myself(((
some update on GTA style character system - add simple jump - gta style third camera PS also i saw on youtube that someone else work at rain system, so for now i will stop my studie in this vector becouse why spend time for the same
Some updates on dynamic MUD What new: add test materials add water in tracks add shadow that support haight of material add support many vehicles add support vehicles with two tires on one wheels(trucks) add support different size tires have a bug, that standart static texture of mud (in vanila game wheel go thru it) cut off my wheel, so need find way somehow fix it..
It's amazing ! Can't wait to try this mod out ! Are you planning on different terrain " textures " ? Like this kind of deformation but on snow , dirt , sand and so on ?
Yes, sand already has a different deformation pattern — more V-shaped — while mud is more U-shaped. You can't really see it in the video because the sand layer is too shallow, but the system itself already works. Right now, the main problem is figuring out how to properly adapt this to BeamNG's terrain materials. Ideally, this mod should simply replace the original game's "static" terrain textures underneath the actual physics/deformation layer. That would make the system much easier to use. Currently, BeamNG's Terrain Editor paints different materials on the terrain visually, but those materials are not just visual. They are also connected to different physical properties and terrain behavior. For example, when a vehicle drives through mud, it interacts with the physical mud volume/material underneath, like the standard BeamNG mud shown in the video. If I were a developer with access to the engine source code, I could probably remove or replace the original static terrain rendering much more easily. But as a modder, I now have to find some kind of workaround or hack to achieve the same result. Ideally, I would like to give this system to the BeamNG developers eventually. I think they could integrate it directly into the engine much more cleanly. If I'm being honest, both I and some of my friends feel that BeamNG development has historically been quite slow. Maybe AI will help speed things up now. There are many features I've been waiting for for more than ten years, and some of them still aren't in the game, so at this point I'm just trying to build them myself. Torque3D already has support for rigid bodies and characters — normal skeletal meshes with animation blending — instead of hacks like my current GTA-style character system. I don't really understand why some of these classes/components were removed or left unused in BeamNG. They wouldn't necessarily add a huge performance cost by themselves, but for modders they could provide the foundation for proper character systems, interactive environments, and physics-based worlds. Right now, we mostly have to rely on hacks. In my opinion, for BeamNG to really reach its full potential, there are several major milestones that could be achieved either by the developers or by modders. Complete mud / environment VFX system I think we eventually need something approaching the level of Spintires / MudRunner / SnowRunner. Water — I also want to try implementing this. I've already seen some prototypes, but some of them appear to be paid mods. I want to try making something like this for free. Mud deformation — already in progress by me. Rain on vehicle bodies and glass — something similar to Forza, Assetto Corsa, or DriveClub. I've seen a few prototypes on YouTube and they already look good, although some may be paid and, as far as I can tell, they aren't universal systems yet. Dynamic dirt on wheels and vehicle bodies — I'm also trying to implement this, although my current version isn't working properly yet. I've seen someone here making a prototype as well, but I don't think their solution is universal or supports wheels and modded vehicles properly. Better particle effects — terrain deformation alone isn't enough. We also need proper mud, dust, smoke, water splash, and debris effects. I've seen someone working on volumetric effects on this forum, which is very promising. Features that would greatly improve immersion and make the world feel more alive GTA-style characters — already in progress by me. In theory, this could later be combined with animated hands, pedestrians, passengers, and characters inside buses. I think I can already make something like a bus full of characters, but I haven't tested it yet. The next important part is creating a proper system for characters entering and exiting vehicles. Driver hands like in other racing games — I know someone is already working on this. Environment physics — I experimented with Box2D/Box3D-style physics integration, but the current results are still very rough. It needs much more research, so I've paused this project for now. A proper train system — I already have some ideas for this. Train mods already exist and work, but creating and configuring them is painful. We need a proper editor/system where trains and tracks can be set up almost in one click. What I think we really need from the BeamNG developers Proper racing AI — BeamNG is also a racing/driving game, so good racing AI is very important. DLSS / better anti-aliasing — and OMG, DLSS is already coming in 0.4! I can't wait. Multiplayer — not just split-screen. Split-screen is useful, but it's a completely different thing from proper multiplayer. Continued improvements to vehicle physics — even what BeamNG has today is already a million times more interesting to me than most other driving games. BeamNG is one of the only games where you genuinely feel that hitting a single rock badly can completely change the outcome. We also know the developers are working on tire physics, which suggests they are taking more serious racing simulation very seriously. If tire physics becomes significantly better, it's going to be even harder for me to enjoy other racing games. Repairing broken vehicle parts/JBeam structures without resetting the entire vehicle — I haven't heard much news about this, but it would be extremely useful. More GPU-based physics processing — this is a huge one. Right now, much of BeamNG's physics workload is CPU-heavy. GPUs are designed to process huge numbers of operations in parallel. If more parts of BeamNG's physics could eventually be moved or adapted to GPU compute, it could potentially allow significantly more physics objects, more vehicles, or even much more of the environment to use node-based physics. I haven't heard about any concrete plans for this yet, but it would be very interesting. After all of that, BeamNG could become something almost like GTA for the racing/driving genre — an open vehicle sandbox with realistic physics, characters, environmental interaction, trains, weather, water, mud, destruction, and racing systems all working together. P.S. Other gameplay possibilities based on BeamNG And that's not even the end of it. BeamNG could also become a foundation for completely different types of games and gameplay modes, for example: Vigilante 8–style vehicular combat — weapons, rockets, machine guns, armor, shields, pickups, and combat arenas. Carmageddon-style gameplay — chaotic vehicle combat, destruction, environmental interaction, and crazy game modes. A Mad Max–style universe — heavily modified vehicles, weapons, convoys, hostile vehicles, large open environments, and physics-driven combat. BeamNG already has an incredible vehicle physics foundation. With the right systems built on top of it, it could become much more than just a driving simulator. I have so many ideas that I've been keeping in my head for years, and now I'm finally trying to turn some of them into real working systems. But if someone else builds any of these things before I do, I'll only be happy. The important thing isn't that I personally make every feature. The important thing is that we get the game we want much faster. So if anyone knows that one of the systems I've mentioned is already being developed — even as an experimental prototype — please post it here or send me a PM. I don't want to spend months reinventing something that somebody else has already figured out. I'd much rather cooperate, learn from existing work, or simply focus my time on another missing system. In the end, I just want to see how far we can push BeamNG.
