An Unusual Way to Fix This Common Problem in FPS Games
Вставка
- Опубліковано 4 лис 2022
- Wishlist on Steam: store.steampowered.com/app/22...
Support on Patreon: / obrooks
Go and subscribe to @garbaj! Do you know of any games that use this technique?
I first learnt about it from the Unite LA talk about the then-new FPS sample. Apparently the new Doom games use it, but I haven't been able to find a definite answer. The way they would allow for interacting with the environment and enemies could be that the projected viewmodel arms leave the screen briefly, then re-enter the frame with a normal shader that doesn't have any FOV magic on it. Then swap back once the animation/cutscene is complete. It's a possibility! One that I may end up implementing.
A Unity implementation can be found in the FPS sample project, released a few years ago: github.com/Unity-Technologies...
Link to the Unity LA talk that includes details about this method: • Deep dive into graphic...
Still working on the game. Some eagle-eyed viewers may have seen a greybox structure in the background. I'll talk about it in the next devlog!
#gamedev #unity
Social media things:
Instagram: / obrooksdev_
Twitter: / obrooks_
Music: / prodriddiman - Наука та технологія
Do you know of any games that use this technique?
A Unity implementation can be found in the FPS sample project, released a few years ago: github.com/Unity-Technologies/FPSSample
Link to the Unity LA talk that includes details about this method: ua-cam.com/video/8chy1aVpiiw/v-deo.html
I use it, I messed around for I think a week in UE4 trying to solve it trying all the different methods. Shading issues, performance issues, weird artifacts... Everyone on different forum threads said 'just shrink the model down' as I continued to resist. So yeah, it's literally the best solution... lol.
This was a very informative video! I have never heard of this method before. My solution to this is changing the properties on the materials. All you do is change the materials to transparent instead of opaque and change the depth test to always. Lighting would still work with the materials too, so that is also a plus.
The original halo games
T-Rex sized arms used in dead island? so the arms looked good in first person and allowed smooth jumps through windows.
Bro Make WW1 Fps In Android Pls
The main drawback of this method is the lack of scale if the game ever ends up ported to VR.
The method I've seen commonly used is to use a renderqueue to render the gun last, which makes it appear in front of any other object in the scene.
It's inexpensive compared to every other method, it keeps the model's lighting within the world, and it keeps its scale.
True! I've been meaning to look into that, it sounds like it will be easier to work with
do you know of a tutorial for this?
Well, to be honest, VR games would be entirely different anyways. You prevent clipping by adding colliders to the gun and hands that physically prevent collision. None of the methods discussed here would work in VR.
@@SashaMasonVR VR does not mean physical simulation.
It just means you've got motion controllers and a VR headset, and you're in the game.
Borderlands VR and Skyrim VR are great examples of this.
Hell, motion controllers aren't always used, or seldomly in some circles.
@@GeekyGami Usually with roomscale and motion controllers, it kinda does.
You can't use camera tricks or the likes on anything you hold in VR. Anything you hold you can move and rotate in any way, see from any angle. It's not like in desktop games where the gun and animations are in a very specific angle, without being able to see it from any other direction, without being able to hold something up close to a wall and then look at the wall from the side.
This doesn't even just go for weapons, it goes for absolutely anything you can hold as there is no difference between holding a gun and picking up something else.
Also, games like Borderlands VR and Skyrim VR that weren't built for VR are terrible examples. I've played a lot of VR games, some with swords, most with guns, and in all of them, anything you hold is "physical". You can't use a second camera to render something in front of something else. Everything you hold collides with the world and other objects. Blade and Sorcery would simply not work if wasn't like that. Besides that it would be incredibly disappointing holding something and it wouldn't collide.
Either it needs to collide with the wall (Like it does in most games) or it will simply clip (VRChat)
BTW I think your voiceover is too quiet relative to the music and both are too quiet compared to most other videos.
I think you're right - I'll keep an eye on the audio levels for the next vids
@@obrooks Nooo, I love it. It’s perfect. Almost ASMR-like.
Your voice and the music is incredibly relaxing. The only advice I have is to very slightly lower the gain on the music.
@@prod.royalsg1630 I could barely hear him
@@s.e.c.u.8340 you probably have bad audio then cause I had my volume at halfway and I could hear him perfectly
@@Kameron2335 Only this video i have hard time hearing a person talk
The Metro games use a similar technique. The fp arms and gun are quite small, but they also have a 'move gun when close to wall' system too for realism which works in an odd way by scaling the arms & gun smaller towards the camera as you get closer to the wall, which from a first person perspective looks like you're slowly lowering it. This also helps prevent clipping as you'd imagine
DOOM actually does do it like this too but swaps the arms out when doing any interacting with the world (like ripping apart demons or interacting with a button).
What's also super important is to consider the pro's and con's for each game.
Like for my game I want the weapon to collide with the environment and perhaps even have bullets come out of the barrel even when in ADS.
So having it in a 1:1 scale is beneficial for helping the player figure out where the weapon is relative to the world space.
But why is that behaviour a pro. What problem does it solve or what interesting dynamics does it bring to the game?
@@joel6376 I see someone hasn't played tactical shooters :P
A weapon tends to be more powerful the longer it is, but that power cannot be utilized if you're fighting in spaces too tight for it.
For example: in most games pistols just tend to be a weak backup weapons to use when your primary runs out of ammo at an inopportune time (queue that one quote from CoD4).
But when the player has to take into account the dimensions of the weapon and fighting in an enclosed space, suddenly the pistol is more deadly by the merit that it can be maneuvered with far more ease.
@@Killicon93 That sounds insanely boring but good luck making it interesting
The easiest method is now using URP render objects.
Renders last, recieves world lighting, custom FOV supported out of the box.
I love how short and yet helpful this video was. Ive messed with a couple different techniques all that felt like there should be a better solution and this seems like it could do the trick. Also your game actually looks really cool! Keep up the good work
i love your placeholder enemies
Downside: To render a small weapon close to the camera you're near clip plane has to be even lower than it would normally have to be. And the ratio of far clip / near clip is critical for floating point accuracy of the depth buffer. You're effectively limiting your draw distance when you choose this technique, or you're inviting nasty artifacts like z-fighting in shadows. If you're making Silent Hill First Person Edition it's great, if you're trying to make a Battlefield type game you'll run into issues and end up having to use multiple cameras any way.
Lights can be configured to affect different layers, so the two cameras option doesn't have any lighting issues that I'm aware of, in Unity at least.
The style and even the thumbnail everything felt like Garbaj. Good job!
This sounds funny, make it like you don't know who Garbaj is
Awesome job! I love your unique approach to things.
Cool solution, nice find.
You know I've done plenty of work with this sort of thing but I genuinely have never heard of or considered this method of shrinking the weapon and then blowing it back up with a shader. That's pretty brilliant.
You mention that this method doesn't work so well when interacting with the environment. That may be true, but if you take a close look at how a game like DOOM 2016 does it, a different mesh is swapped in every time the player's hands need to be positioned with other objects. For example, when you clamber, or when you start one of the first person cutscenes (scanning your hand on a keypad, say) your viewmodel arms are quickly lowered down in the same way as when you swap weapons, and the different arms in worldspace are raised up quickly.
Now this does require you to lower and raise the arms every time you change between these viewmodels, but this really isn't an issue I think! Even when I was trying to figure out what they were doing by looking at the game, I couldn't immediately tell they were swapping out the meshes. That is to say, it's an illusion that works really well even if you know it exists, which is great. I took that method and have been using it for my own project within the first person cinematics, and it works fantastically.
So I wanted to add this as a way to 'double dip' between the advantages of your method and being able to interact with the environment. Personally, I'm probably going to try this approach myself now. :)
What's a shader ?
Brilliant idea, this solved a lot of issues for me!
Very interesting, I've never seen this method. Thanks for sharing
I love your videos please don't stop producing. I really wanna see how this games turns out
Chill video, love your commentary style.
The simplest solutions are usually the best ones. Thanks.
Wow, really cool solution. Please keep making more vids.
Would love to see you model a zombie and multiple zombie types, like rushers, crawlers, alien type zombies and so on.
Keep it up. Don’t give up like you did last time!
yes i love this project..i hope he don't give up!!
I not only saw that video from Garbaj but also saw this one ;) nice video!
Cant wait for the next devlog
Best thing i seen so far is from Totally Accurate Battlegrounds. the gun you see is actually on the world with physics on, it seems silly and it is but I really liked it.
Me: drops phone while watching this
The video: makes Minecraft OG oof sound in the perfect timing
THIS IS WHAT I'VE BEEN LOOKING FOOOOR
There is also another way similar or the exact same as what Geekygami explained. It's called Custom Render Pass, it will allow you to render the gun before anything in the scene and I'm pretty sure games like Call of Duty use something similar to this since it's a feature in every Graphics API. Custom Render Pass doesn't hold you back on things like scale compared to the method you're using. I would look into it but it may not be available in every Template (3D, URP, HDRP).
Im guessing by "before" you mean on top of everything else rather than the first thing to render, right?
@@lengors7327 Basically yes but it still renders the viewmodel (gun) before the scene.
Coming at this from UE4's perspective, the double-camera solution seems to be ah not exactly as intuitive as Unity's hee hee so we gotta go about it differently. Ran into a vid of a guy who made a material that not only scales the size of the item held, but also causes itself to be always displayed 'on top' of everything. Put it on a ball and walked around a wall and you could still see it, in a almost inverted normal sort of way. Would that be, pretty much, the type of shader you used in this technique?
A bit of a late reply, but I hope this helps:
I know the video you're refering to for render/shader-passing an object to appear "infront" of any object with the given shader. It does NOT scale the object like what is used in this video. Rather acts as a layer to pass any given material that is rendered from the camera. In the shader you can set an ID for the layer you want incase if you want to build multiple instances of the same material and just set each one to a different "height". The only drawback is when interacting with objects or the environment, you would most likely need to disable the material shader for those uses ie: ledge grabbing/mantle, door/draw interactions etc. Otherwise you would see your model "clip" through the object/environment you're interacting with.
I've used it for my player model for my game and it works very well, especially for wall running and wall climbing.
It's a really odd topic considering not a lot of people talk about it.
@@GibbonDev Haha the more nuanced the topic, the more odd the details, Mr. Gibbon! Always fun to watch random tip vids and see if it's something I'm not aware of, but mainly to keep it in the 'back pocket' to help fix an issue I run into in the future. Thank-ya for the details reply n' best of luck on your game!
I have developed a publicly available method using shaders that allows the "separate rendering" method to receive shadows but also removes the overhead of an "extra camera" (because well, it's rendered in the same universe). That method is actually the "classic" way used by big companies : it pretty much clears the depth buffer where the weapon is then redraw it.
My solution is only meant for built-in forward rendering though. There is however a method using URP and HDRP. I'll try to find it and link it here ;)
Ah there we go found it : ua-cam.com/video/5AmI2yOx0Nc/v-deo.html
You can solve the issue with custom passes in script able render pipelines ;) Their example forgets to turn off shadow casting for the weapon so you end up with a weird floating shadow but that's an easy fix. You get shadows and a custom FOV as well thanks to this solution ;)
Also iirc your solution was actually the most common one in UE4. That was the one recommended to me as the "standard" when I dabbled in it years ago. I'm even pretty sure Unreal Tournament "4" (the free sequel Epic was making before shutting it down in favor of Fornite 😣 ...) was using this solution :)
Nice! I've not used unreal too much so hadn't seen it, and looks like quite a few games have used it in the past after all!
I'll take a look at the custom passes, a few people have mentioned it, and it seems like that could be a more manageable way to do it. Thanks for the info and links!
Barony does this. They had to downscale the weapon model about 4 times lower than it's normal size because of how big it looked in the screen being so close to the player.
Something what can also be done is to compress the matrix on the depth axis making the weapon very short. This trick also makes sure the weapons does not look like its stretched out in a higher FoV
Yooo this method is actually improving the performance of my game! nice
The layered camera techinique, with lighting is super easy in Godot.
your intro is spot on, mr zeitgeist
you can just renderer it with a special render pass, painting last. this benefit does wall clipping, lighting
Would love to see the shader magic you whipped up to achieve this FPS gun clipping issue!
I can't take credit for any shader magic! Just came across it in the Unity FPS sample - here is the specific shadergraph file github.com/Unity-Technologies/FPSSample/blob/master/Assets/Shaders/Graphs/FirstPerson_Projection/Firstperson_Projection.ShaderGraph
@@obrooks 🤩
I actually tried using this in one of my cod zombie "clone" projects and it worked fairly well at the start of development. The only drawback for me was animating at a smaller scale and having the character interact with in-game elements during animations.
That intro! 10/10
A game called Team Fortress 2 renders the viewmodel in the world, but makes it invisible to other players and render over everything. This allows it to have real-time shadows (atleast as much as the source engine allows for) and negates the problem of it clipping through world geometry
You could also change the render queue on the shader. It's like the best way
nice little video keep it up
nice i love garbaj, now it looks like i like ur channel too, great stuff!
Ah yes, the mighty Garbaj, I've used his videos a couple of times.
When I made my first fps I just made a gun in front of the camera and pointed to what the camera raycasts forward, it gave some funny effects when you approach a wall or when you shoot the bullet doesn't go straight forward from camera perspective but it goes from the gun to the objective.
That's a cool solution. Love to see unique strategies of small devs. My personal tactic was a 2d sprite. But obviously it wouldn't work with realistic art style
I love the kaiser capsules
Damn, that's a nice fix!
I love this game! It looks so cool. I really hope it works out for you. Also here are some suggestions for the game
A campaign would be really cool in this game
The blue aiming square could be upgraded and
You should apply a sprinting mechanic
Thanks for hearing these suggestions and good luck 😊
It's good that there are such talented developers! By the way, your game is similar to the game in Verdun!
how do you get the final position of the gun after projecting it? All the transforms will be mismatched? When you cast a ray from the gun it is mismatched because of the projection.
S4 League fixed it by giving weapons their own collision box.
So tricks could be utilized like holding block while walking against the wall to walk faster towards the goal since the ball removes stamina recovery.
Same thing happens when shooting and walking against the wall.
But I'm not sure if reloading also works.
damn. never thought of this way
There are 2 solutions I found in Unity.
1, use FPV addon from Asset Store. It's perfect to fix clipping problem on all pipelines.
2, Use HDRP custom pass. The principle is simple, draw the environment first, then use custom pass to delete the depth data on depth texture where your FP hands/weapons cover. Then draw the FP hands/weapon also to overwrite the depth on depth texture. So in the end the FP hands/weapons will always be on top, never clipping. And the depth of the FP hands/weapon will make it receive shadows from the environment correctly. Also because you use custom pass, you can use a standalone matrix to control the FOV of the FP hands/weapons. There's an offical example in HDRP custom pass about it, but it doesn't work correctly for practice. I modifies it to make it work correctly with post processing and transparent objects like optical sights. I posted a thread on unity forum about it, and I post the link in my last comment but it got nuked by youtube (probably can't post link)
Damn, this is really clever. The only thing that I think could be improved is that shadow on the wall that reveals the trick and makes it look weird. I love how this things are debated and expanded on by many creators, probably someone will come up with an even better solution in another video. Really cool
some engines, godot being one, you have different lighting layers. so its still recieves from all the lights if you tack off for them all the affect the layer the gun is on. and i dont remember how but having a seperate gun cam in unity never messed with my lighting either.
Oh my god I had never even considered this
I wish fps games actually had the 1st person view model go backwards when colliding with walls or objects. Whenever I see games, like CoD, never adjust the 1P view model, the sizing proportions are so unrealistic when you crouch or prone into a ragdoll body. Like the eye on the screen is 1 foot big because you're so close, but your gun remains the same and suddenly is now "smaller" than an eyeball.
Fair enough. Decent method...although personally I've found giving the gun itself a hitbox...works fine.
give it a hitbox, you mean a trigger? and then what, move it out of the way? that solution was covered in the video.
thats actually how Fnaf UCN achieves 50 animatronics, they shrunk them all down then bloated them up when called in. at least thats what I heard.
The Imperial German Beans cracked me up!
I wonder, then how some games have accurate reflections of the gun, the character holding it and interactions while gun not clipping in!?
while working with hdrp i used the "scaling it down till its inside the collider" method alot since it doesnt support camera stacking
nice seing someone else using the same weird trick lmao
The placement of the weapon inside the player-hitbox is my way out of this issue. very easy, no need for big gun either.
Thanks
The best method is to draw gun first and mark gun pixels with stencil buffer, and after when you drawing scene you set stencil render-states to not overdraw gun pixels and that's it. But you need to know programming to do this, or engine need to support this technique. Good article about when to draw gun (the first one) and when to draw skybox (the last one) is "Depth in Depth" from ATI.
In unity you can actually render 2d games layer by layer so higher priority layer remain top of the other other layes.. So you can do this in to 3d games too.. View mode will get correct lighting information. Just set the view model as UI layer. And lower the sorting than the UI.
I've thought of this myself. When I was in college making an FPS with my friends, we recorded all the animations the player would see their hands doing, then saved them as 2D animations (basically) and put them in front of the player's camera. We turned off the player's model to their camera so they couldn't see their real selves. If we ever needed to do something different for whatever reason, we could just turn the player's model back on and off along with the 2D animation.
If you watch him daily that that means you have heard this word 6090 times
"Hey everyone garbaj here, Iam making a first person shooter"
you could just use a stencil buffer mask if you know how to use shaders
I would love to play your game. Can't wait to shoot Pickled Rick in a trench
is not better just to make the player collision box bigger?
you could just flatten the mesh in camera space to avoid the double camera overheadh by using a vertex displacement shader. Rainbow 6 does it.
hey, I love the idea, but how do you make the hand animations look real for the weapon? I mean the weapon is actually much smaller than the hands are.
If I understood you correctly, you would use the same shader for both the hands and the weapon, that way it would fit together and the hands wouldn't look out of place. Then if you needed some specific animations where the hands interacted with the environment (e.g. a cutscene where the player pulls a lever), you could swap out the arms for a model that is full size when the animation is playing, then swap back to the custom shader afterwards
@@obrooks ah yes makes sense, thanks sir.
Hey do you have a discord where i can follow this project more closely?
You're doing great! Have you decided on the name yet? I like "Ardennes Under Siege" (by zombies)
The answer is always true full body awareness. Say no to floating arms.
Hey this is cool, I used the same technique for Screencheat: Unplugged!
The 2 camera layer thing gave me correct lighting and shadows, I don’t know why you didn’t have it but that’s unity I guess
Why not just use the stencil buffer to draw ontop of everything else?
0:44 this is not possible overexaggeration, this is literally real possibility with rtx 4090
In my game I use the one with sepert layers and have listing and shadow as well.
My game is multiplayer game and have the option to play up to 4 player on the same computer so if want to make right I have to use different layers
The way I mange to make the layers are the following I have one layer for showdows and light's one I layer for the weapon I layer for how the enemy Player see you ECT
The on the camera able to see the weapon and showdown and light
So in that way wean the Player is inside I building for example the have shadow for the weapon same wean are on light
This system works perfectly for my case.
What I have done in the past is just make the weapon small enough to fit in the player collision, and then move it right up to the camera. This way it’s in the physical world but can’t clip in any way. I don’t mess around with any of the fov settings or shaders, and I instead reduced the far setting to its minimum. Clearly not the best solution but it was easy to set up, and it worked ok.
The yoda meme made me subscribe
nice video, but the sound is too quiet
Is it not possible to just render the world first, then turn off depth testing and render the gun?
Hey. Love your videos.
Another solution would be just to utilize a different depth buffer's range, e.g. [0, 0.1] for the weapon, (0.1, 1] for the world.
It won't clip through the walls, and it will also be in-scale relative to the world, so illumination data would be correct. I believe most of the games use this technique
turn up the sound of the microphone :)
btw nice idea , but i like to make the gun just point with the nozzle to the raycast of camera , so it just never get inside and it looks like some animation :)
Woooww, that’s smart!
Just go the hard way and model it properely in the world and have it collide (btw i know why don't ppl do this in the first place, but just do that instead of scratching your head trying to find a workaround)
call it "the cold grave"
Great video. Btw I could hear you fine I don’t know what’s wrong with peoples ears. If I can hear what you’re saying fine then the mixing with the music is subjective. Do what you think is cool! I love it
im a bit iffy about the double camera and small gun techniques because when you go to a wall it kinda makes it look 2d, and breaks immersion. Could be just me
Why are the German soldiers pickles?
Funniest shit I've ever seen.
Game idea name: The undead ( or dead if you want) men of Deutschland
best solution was in game "Jurassic Park: Trespasser" : )
Animations are probably the best route. I’m always thrown off by ultra realistic skyrim mods, that look great, but everything attached to the character clips through each other. No one seems to be bothered by it though
ooooohh....what a genius
I just render it last without depth test, nothing can be drawn on top of it anyway.
Giá như tôi học làm game FPS để remake game 7554 thời cổ sẽ trở lại:)
This method is great for like 90% of every FPS ever