Question to NextGen Maps/Models

[3XIX]

Hello Cubers,

im working on updating the game mostly Models/Maps, this leaves alot of important questions.
the game seem to be pretty much community made atleast what i could find out within the last days, i´ve checked who developed Cube Engine and the Cube 1 game and took a look at its code, i never played Assault Cube before but i became in love with the simplicity of it. I´ve also been looking for games which doing good in performance per watt. Right now i would like to know what PC-Hardware you´re playing the game on, this would allow me to know if i should spend more time into creating models with higher quality, or in general giving the game a newer look. It also could open up a new page on how Maps will be created in future, i pretty much googled last weeks to find anything close to what i´m doing or if that´s been done already. And it doesnt seem to be. Im creating on a Low End system myself (would have been High End at AC release) with an Intel Core i5-3360m, Intel HD 4000 graphics. I´ve already created a little "Test-Map" to check how the engine performing, maximum verticles which could be used at different frames per second steps 30,60,120,200 or above. The following picture shows the "Test-Map" with over 1.6 Million Verticles (placed alot of bench seats there around 100 with highest possible verticle count, Anti Aliasing x16, shadows active, insane graphic settings.
without Anti Aliasing or shadows it run with 100-200 FPS depending on view angle. Problem at all this...im unsure how it will work server/client, since Maps will require the specific Models (4k Textures can be done that way).

would be happy if you could tell me what you think about it and your specs
thanks

[Marti]

9 fps isnt very realistic is it?

I suggest creating maps that are properly suited towards AC's gameplay, opposed to visually pleasing.

[Pi_Halo]

See the thing is, the older cube engine and the Assault Cube engine have a tough time rendering a high poly count model. As far as I know though textures aren't handled in the same way and high res textures aren't nearly as a strain on the engine as a high quality model. The other thing is, maps already in Assault Cube have to be of a certain style to prevent a type of "jittery" feeling. For a few examples, too much water, too many heightfields, too many ents, too much open space, are all ways the engine could be bogged down just by a map. I am aware most people have "good" machines lately and can run them for the most part. But it still creates a weird sort of half lag half not type of jittery feeling. Its quite limited in what should be done vs what can be done.

[3XIX]

Hey!

To the FPS part, the picture shows a Stress-Test. Its pretty much caused due Anti-Aliasing which isnt really supported by Intel HD 4000 and does not need to be used.

In general i made the post to ask if some people would enjoy to see the game in better graphics, while keeping a pretty solid FPS due its engine. If i play Valorant on lowest Settings with reduced Renderscale it looks worse and will end in FPS around 50, without Anti Aliasing etc. on Intel HD 4000. On top comes the much higher power use and a CPU going into 100%, while Cube Engine could max go up to 20%.

With all that said im sure that this Engine does a much better job and it will be more "greeny" as latest stuff developed. And the "glitches" you describe pretty much caused due shadows, which do not need to be used. Also a similar Map design can be done, its only limited to our fantasy.
I´ve made a little Pickups, Pistol mod package for you to check how it performs on your maschines. If possible create a new map and place multiple of the pickups, model shadows currently turned off.
https://workupload.com/file/6FDKF84erKS

[YesWeCamp]

This is on the homepage: With efficient bandwidth usage, it's low-latency and can even run over a 56 Kbps connection. It's tiny too, weighing in at a lightweight about 50 MB package available for Windows, Mac and Linux. On the correct settings, it can even run on old hardware (Pentium III and above).
I think this gives some insight in what kind of pc's people are using when playing AC, and why they (started) play(ing) it. I myself have played on really crappy pc's/notebooks. I wouldn't invest time in creating high poly count models. In this case I think less is more. And I'm saying that with a little nostalgia, so can't tell what the new players want/like.

[3XIX]

First of all im not trying to destroy something for anyone. Its meant to be additional or maybe standalone, so that people still able to play what they enjoy. Thats what the overall question of this thread is about, finding out what should be aimed not if it should not be done. And i´ve stated that im on a slow notebook myself from 2010 with a graphics card which isnt meant for gaming, but since the engine is this lightweight and the general stuff is already done, it might be a good chance to mod the last bit out of it instead of creating something with a new engine. Less code is the question to performance, overall when im done it will run on my Intel HD 4000 with 60 fps, 20% CPU usuage, 1 GB RAM but maybe full dedicated VRAM due 4k Textures but its worth it if you see what i mean. And dont get it wrong im aiming even higher.
Maybe im trying to find people having the same vision as i do.
https://i.imgur.com/cDzl7D3.jpg

[Mr.Floppy]

Hi 3XEX and welcome to the community.

Please don't be discourage, even though some of the feedback might come across somewhat dismissive. Actually people are just sharing their experience with the cube engine and telling you what obstacles are ahead of you. I guess everyone in here appreciates any efforts aiming at providing something new for the game.

Using higher quality mapmodels/skins is a valid approach and certainly something worth looking into, in my opinion.

Though, I'm afraid you're indeed aiming a little high with that vision, but then, I just don't quite know what your vision is yet.

In order to find people supporting your efforts, you probably start best by telling us a little more about your background in terms of modding experience and with Assault Cube in general.

[3XIX]

Thanks,

Im not trying to be discourage, im a person who looks at facts. Also im the one spending my time at creating something and nothing good happens without a debate, thats a difference.

To the Map Models and my vision, its pretty much focusing on creating graphics like Modern Warfare for low end systems, the shader loads which coming with the new games, they even create shaders for darkening these days...whats Gamma for? Or just manuelly do it with Textures. Ray-Tracing and all that useless stuff. Its like game companys do it for hardware vendors like Nvidia/AMD to sell their "newest" graphic card which wouldnt be needed if games wouldnt be stuffed with unnecessary sh*t. Its pionier thinking not just for now, for our and the future of the next generations. Look at the system requirements for Modern Warfare, you need a good PC for that and those running at power levels far away from intelligent or smart.

How i wanna archive this? Well i use real 3D scans of objects which im converting to fit assault cube, cant get more real in 2020. )Normaly those objects come around with 100 up to million verticles, that wouldnt be possible even due the md3 format + the performance impact). The first picture i posted was to show verticle count in combination with Anti Aliasing and shadows, add up 200 Mhz to this graphic core and it would have been 30 fps. Deactivate AA and Shadows and its over 100 fps with a GPU running at 8 watt. That shows how much room there is since you can easily vert a object 1/10 in count. With that said i can stack up the Map with about 1 Million verts in view angle (degree depends on fov) and beeing above 60 fps. Take that times 4x thats around the size you will get to see the full Map, which makes 4 Million verts placed on a Map by having a fps staying around 60. Activate AA/Shadows and it will be around 30. Power? Well lets just assume 300 Watt for a average modern PC dunno how much when gaming but compared to the 10 Watt it will take me for 30 fps AA/Shadows enabled. And probably just 6-7 Watt without at 4 Mill verts. There is no way you could say that the Cube engine has problems with alot verts.

How i want it to look...depends not really a Map designer thats why more people have to try otherwise it stays at 1 Map...since this is alot of work. First Map will be a desert type map...will see have to create alot of things first to add more to the scene.
picture is just a start with AA enabled dont forget that! the other sides of the map with objects placed aren´t within the view angle therefore no fps drop.


experience? well ive hacked in past, created mods for games, create models etc, tried to create a FPS multiplayer shooter with gdscript in GodotEngine etc....which doesnt really work out since ive no time for coding and its problems. And Godot doesnt support anything below OpenGL 2.0 while cube runs on 1.1
And AssaultCube has a cool gameplay...why wasting months in copying that which also wouldnt come close to the fps without reducing the renderscale. And a "blurry" game isnt really good graphics.

Edit:
Whats needed are basicly a few people which could take away parts, like character animation,guns, skymaps etc. anything helps or even sharing sources like https://hdrihaven.com/
I just fucked up a character model, texture is off...will stop for today

 
Piece

[3XIX]

Evening,

i tought about what you all have written and i´ve found a solution.
You´re right the more verts the more performance it will cost, we can´t avoid that but i´ve scratched my head to find a way of still using 3D scans and shrinking them down to the absolute minimum verts.
Cause sadly the most detail you only get out of real scans, since you´re unable to use AO, Metallic, Roughness or Normal textures well some normal map texture detail could be stored within the alpha channel of the diffuse texture but that won´t give a realistic result on selfmade textures, all that requires effects which would just decrease overall performance.

the benefit with those scans are the textures, with them we can shrink the model mesh and let it be just a texture "holder" instead of trying to gain detail with the model. Right now i will replace all mapmodels with that method with optimizing models to 500 verts each max. That might still be to much for some PCs if many models stacked but cant get more out of pretty much nothing. With that i will put the map project aside for now, i think this might be better for more people

[Mr.Floppy]

Sounds like you're on the right track. ;)

I've got two little hints which might be of help.

First, concerning the lighting of models. By default a model's vertices are all lit with the same intensity, which is alright for small models and those which don't stretch horizontally. For all other models you might want to use the following command on the model's md3.cfg file, in order to have all vertices lit individually in relation to all surrounding light sources:

[SELECT ALL] Code:

mdlvertexlight 1

Secondly, since you're going to replace a lot of content you probably want to separate that stuff into one place and even keep the original paths and file names, so everyone could just use them without the need to override the existing content, or changing a whole lot of mapconfigs. Assault Cube provides a mod-feature pretty much exactly for this case. If you haven't already, just check out docs/mods.

If you have questions, or need someone to test, just post in here.

[3XIX]

Thanks for the advice,

sadly i´ve bad news. Ive changed some models so far but found out that it doesnt really makes sense to make mapmodels alone more realistic while the map itself contains old textures...it basicly doesnt fit together.
Now i dont know if im having false info or not knowing how its done but i cant get 4096x4096 textures working in the game besides models. I´ve tought about creating quad mesh models as walls but that would destroy the map making system of the game. I´ve also checked if gun models could be replaced with mesh texture and simple alphablend to gain better quality by not having to use alot mesh. It also seems like the game doesnt care much about verts instead the triangles of a model, converted triangle models to quad models and performs even worse. Overall everything i would do to make it look better would require better hardware, since atleast 1 part won´t go with it. Just models without wall textures doesnt fit, lowest possible mesh with alpha blend...wouldnt impact performance on viewangle but when beeing close to it.
Then i´ve found the project where one of the old cube developers worked with called "Tesseract" where they already did a great job on better graphics but with cube 2 engine moded (better hardware required).

With all that im unsure if we shouldnt let assault cube be as it is, trying to find a method for replacing to higher res textures first instead of models which later doesnt fit within the map environment. Cause only models looks like a new lamborghini on a junkyard. Also for map creation you pretty much need individual models for your needs and vision

Now the question is how we archive better or atleast cleaner graphic (texture) without touching the performance to much...or is there a way replacing editor textures with high res ones?

[Mr.Floppy]

Well, actually you came to the same conclusion we have come to years ago, hence the first responses you received further up.

There were attempts in enabling high res textures for Assault Cube for exactly that reason. The code probably made it even into the next branch, but certainly can be found in the automapcfg branch.

Basically it did work, you would set a scale parameter on the texture command in the map config file (2 or 4, albeit 4 was only experimental at that point), which would scale the texture when rendered accordingly. 

However, I remember one issue that probably remained unsolved. The LOD setting somehow couldn't be changed properly and kept kicking in too early. That is, when stepping back two steps off a wall its texture already got rendered with standard resolution, so the floor texture effectively only ever was displayed in high res when crouching, which made the whole thing kind of pointless. This most likely could be fixed, I guess there is some value hard coded somewhere, which would need to be made variable, or something like that.

However, we've got nobody working on this, so we're probably never going to get that feature, unless some proficient coder comes along.

[3XIX]

Well thats a big problem, besides that my only ideas would be either creating walls via models (static) or creating a cube model where people could create their maps walls with, minecraft kinda method. Would take much longer to create maps but it would keep its style. Also the resolutions would be untouched of LOD. And other Modellers could create them their mapmodels which they would like to have since i dont think a coder could easily fix that LOD issue plus the scale by 2 or 4 which would still have a slight blurry look. I sadly dont know shit about this engine and its pretty much guessing what could work. Its like a math teacher has to teach physics, well he might understand things but still doesnt really know it or knowing windows for years and having to switch to macOS. So yeah we need the right teachers otherwise we could try the methods we have and understand by dealing with it not beeing the best solution. Btw its easier to ask as to searching for hours...what are the commands for placing new models "newent mapmodel" doesnt know where it should refer to like "newent mapmodel */packages/models/mapmodels/model" how is it done..how you rotate/resize/remove them if possible and in general would it work via the -mods method so that the client will get them when joining the server? Cant set up a server rn since my ISP blocked port forwarding...have to call them aswell. At moment i´ve to search a new job so ive not many time to study anything for hours. But ye the cube model idea is all i got now to allow* people to still create their own maps with high res texture.

[Mr.Floppy]

Well, replacing geometry (the actual cubes) with models could work for certain areas, where additional details are needed, or one needs to get around the vertical limitations, like you can't have more than one opening in a wall. However, that's not going to work as well on a larger scale. The Cube engine is built around cubes and solid walls are the only occlusion mechanic (performance!) there is. Maybe even worse, it is also the only reliable visual barrier every client will always have in common. I don't want to get into too much detail here, but there's no validation whether all models on the client side are the right ones, or even if those are loaded at all. Therefore one might have a blank space where others have a model sitting in the view... Hope you get what I'm trying to tell you.

In order to place a model by command you need to supply (at least) the number of the mapmodel's slot according to the map's config file, e.g. '/newent mapmodel 3'. If the map has no separate config, the default_map_settings.cfg will be used, which can be found in the config folder.

Manipulating models isn't as easy as you may know from other engines, like Unreal.

You can resize only on the model's config file using the command 'mdlscale'. However, this is then always applied to the model when using it, you can not change that in the editor per map or per instance in a map.

The vertical orientation (yaw) of a model is set accordingly to your orientation when placing the model in the editor and can be changed afterwards. Move close to the model until its entity-sparklies shine blue (model is selected), keep pressing 1 and scroll up/down. There's also another method, where you can type the values, but I keep forgetting the command... Note, you can only change the yaw, not the roll or pitch of a model, that's why there usually are a lot of variants of one particular model, which only have different rotations applied.

If you want to move models you either delete and place it somewhere else, or you again 'select it' keep pressing M and scroll up/down. However, this does move the entity-sparklies, so you'll have to follow the model... Hard to explain, try it and you'll understand what I mean.

By the way, models need to be clipped since there is no mesh-collision in this game. For simple models this can be done on the maps config file, however the orientation of that clip is not going to change when the model is rotated and complex models are most likely required to be clipped manually every single time its used in a map.

The -mods method has little to do with all this. It is used to bundle content, especially if its going to replace default content which ships with the game. If you want other clients to get your custom content, without them manually downloading and extracting it into their game folder, the content needs to be made available through packages.cubers.net, which (currently) has no user interface, so the people in charge will have to upload it for you there. At least that's how it works now, this might change in the future. We actually used to have a dedicated upload platform for AC content, but this isn't running anymore.

If you want to try stuff on a server only by your own, you can simply launch a local server using the server wizard and connect to it via 'join a LAN server' on the multiplayer menu. In this scenario the server doesn't need to connect to the masterserver and therefore it doesn't matter whether port forwarding works or not.

I guess you figured by now, how it is not an easy task to enhance the visuals of Assault Cube. There absolutely is potential utilizing higher res and/or more realistic skins on mapmodels, like taken from actual scans. Though, not 4k, but more like 1.5 or 2 times the standard size. In the end, everybody got to decide on their own, if this is worth the efforts. ;)

[Felix-The-Ghost]

Not updates to models per se, but seven years ago (Jesus Christ) I hacked together some code to use models in different ways with 32 different yaws available instead of the current 24 as well as being able to assign independent shadow yaw for models/pickups and even being able to use different (I think they are called) roll and pitch of models with the same data footprint.

[3XIX]

Well with all the testing i´ve found out that cube models itself doesnt work out. The model size aint a big issue, you would just save the models with other XYZ coordinates, mdlscale just seem to be the fine tuning which makes sense. But there are some open spots to even "legaly/illegaly, idk" cheat due that, while i´ve found out at character model replacement with cube 2 character models, hitbox is as big as the character...sadly im way to newschool to understand how they did animations, i first tought they created a different model/position for each frame like run1 a model run2 a model. Well probably won´t really find out without 3DS software. At the end im much more confused about the system requirements of assault cube compared to quake 2 (1997) while graphic on minimum AC require 600 MHZ CPU, 256 MB RAM, 64MB VRAM/GPU Core 400 Mhz and Quake 2 require 90 MHZ CPU, 16 MB RAM, 32 MB VRAM....while both beeing graphically pretty much on the same level besides shadows (which you dont have activated on minimum) that somehow doesnt really make sense if pretty much 80% of the code is from Quake, well OpenGL could be the difference why there is that fucked up performance range between both since Quake 2 using OpenGL 0.2 and AC 1.1 or 1.2 i guess. Im really unsure if to enhance graphics we shouldnt besides from shaders and lighting effects recreate assault cube within quake 2 engine, which after that could be ported to quake 3 engine for those who want better quality, not the Nvidia Ray Tracing remaster of Quake 2 (thats exactly the type of modding i dont like, since ray tracing first wasnt even fully developed to that time and second still isnt. At the end its for a new generation like me funny that they started the best they could but with mostly every new "function" making performance worse without really having graphics increased, who cares about reflection/lighting if walls are blurry...i really dont get it. I think the only chance to stay in the performance area is switching to quake engine since we´ve no engine developer here i guess

[Mr.Floppy]

Not sure if I understand you correctly, but the Cube Engine has nothing to do with Quake's Id Engine code wise. As far as I know, that's two completely different approaches to a game engine in terms of geometry handling and rendering pipelines and probably a whole lot more. The creator(s) of the Cube Engine probably only decided to support the Quake model formats (md2, md3) in order to make use of Quake's vibrant modding scene at that point in time. Today they would most likely pick another format.

However, I think you start confusing things and losing focus.

Actually, the detailed and photorealistic looks of modern graphic renderings are not primarily created by 4k textures and high poly models, but mostly by per-pixel-shading. Bumpmapping, tessellation and realistic lighting (and everything that comes with that last one alone) is what makes modern games look 'nice'.

Porting Assault Cube to any iteration of Quake's engine doesn't seem to make any sense to me, to be honest. If at all, you'd go for something that has a huge and healthy community and which is being maintained, like Unreal or Unity or something like that. Anyway, in either case you'd end up rebuilding the entire game pretty much from ground up, so I guess it's more likely we'll find one programmer to work on improving/enhancing Assault Cube's code base. ;)

Btw, Felix how about your coding skills and did you go by another nickname 7 years ago?

[3XIX]

Those are facts man and sure cube engine developers havent reinvented the wheel. Id Tech´s engine basicly is where it all started with, so cube using same code language pretty much [Edit]: Turns out quake is written pure C* otherwise same renderer (opengl just other version) and its pretty much about the renderer how performance goes. Which renderer choice we got? Whats used everywhere? Directx, OpenGL, Vulkan, Metal and whatever those make the difference and its obviously that OpenGL 0.2 will do alot better as 1.1+ since its not stuffed with additional code for whatever.

And since sadly everyone thinks that good graphics is about lighting and reflection, normals etc you see why it all went to the point of having to buy new hardware every few years for thousands to play actual games. While they didnt understood that its nothing special to create good graphics if you dont care about efficiency. And yes a 4k texture would give a much higher advance in quality while pretty much just eating up more VRAM and most low end devices it doesnt matter due their shared memory, you wont need much higher CPU/GPU frequency´s due that. And compare cube or cube2 with quake/quake2 and tell me how big of a graphic difference those lighting, reflection and all other effects made? It simple feels like same for me. While the hardware difference is unreal. And im talking about empty maps, just the 3D engine itself. Just go and get Unreal Engine 3/4, Unity and whatever engine is out there and make a clean map and compare FPS/Watt performance, thats whats important. If they would do the same in the car industry...we couldnt pay for the fuel anymore.

Anyway at the end you´re right with the point that its not worth it, atleast if you´re not a team. Maybe someone will come across with what you want, may someone already did (AssaultCube Reloaded project) or Tesseract. At the end no render update will make anything better, it just make it worse for those with slow hardware, render downgrade in combination with better models...would like to see that, could look more realistic as seeing a old model reflected somewhere and having to think if thats "me" the player model.
For a better understanding quake 1 is id tech 1, quake 2 id tech 2 engine, after that id tech 3 which been using in call of duty, and the following call of duty games with infinity ward engine is id tech engine at its core, same as cube engine is: https://en.wikipedia.org/wiki/IW_(game_engine)

There you can see what got added each new engine version and then look up system requirements for each game, then we look up the hardware for each engine to play at 60 fps. For the last call of duty Intel Core i5-2500K its 4x3300Mhz together 13200 Mhz IW8.0 engine vs Id tech 2 with 90 MHz now lets see the features we got more is it worth it x146+ times? Pretty much the same is at other hardware aspects, sure lets count in the models and textures which require higher gpu core/vram and ram but did models increase by x146? Not really, did the textures increase by 146? nope. Did the lighting, shaders and other effects increase? yep. What else increased? Power consume by x650+(if you dont count in the architecture optimisations over the years to consume less power, otherwise it would be even higher) and now tell me that those effects are worth the hardware it requires.

[Felix-The-Ghost]

Btw, Felix how about your coding skills and did you go by another nickname 7 years ago?

No, always been Felix online on other communities as well.

Cooperating with at-the-time active devs I was able to commit several bugfixes and new features (and several unimplemented features) that are in the current version of AC

I'm the guy behind the mapping client and various prospective map-version updates that you may have seen if you're in the Proving Grounds (dead since 2017 tragically but at least it's all archived aside from dead links)



If it was ambiguous, the screenshot with the rotated models is a preview of dynamic orientation with the existing models in-game and no .cfg settings etc.

Because you only need so many unique values you can interpret the 0-255 bits for mapmodels etc in clever ways to allow extra parameters with identical data footprint (storage and network) because it's the same 0-255 value. Because this changes how the stored values are interpreted, such a change would require incrementing the internal map format so you can identify legacy maps and handle the data appropriately (which I was able to get working, but obviously would require servers to run that format as well)

IIRC the current development version has various neat mapping upgrades though I haven't seen them firsthand (I don't remember why it hasn't been released, I thought it was tested extensively)

[3XIX]

Found out some more interesting history and it really all started with id tech where some created Action Quake which later got added by the wish of id tech within Quake 2 Pro a package of most popular mods at that time. Some of those Action Quake creators later then left to Action Cube now Assault Cube and the most of them to Action Half Life which then became Counter Strike. Its funny since they all have a similiar style and Action Quake somehow really looks like the first counter strike, models look really like that. And also game files didnt really changed much, they still using bsp format in CSGO. With knowing all that...CSGO and Call of Duty been closer as we tought atleast you could find the red line which connected them or atleast the ideas id tech to action quake - action half life - counter strike and id tech to call of duty given to activision. For me its important to know, and found the core which seem to be quake 2 pro these days where my GPU runs at 60 fps with 0.2 Watt, and CPU usage of 0.5-1% imagine an actual architecture with that...i think i sadly have to say that i am going to mod quake2pro instead of AssaultCube since there is a difference of 1/10 usage by having shadows/dynamic lighting on both sides. Btw it runs at 700-800 FPS without FPS lock, while AC without FPS lock runs at 300-400 on this hardware, by still having 1/9 power consume. Pretty much same model vert counts, maps itself also pretty similar (action quake ones, the old ones)

And to Felix, that seem to be a cool feature but i think its not in, cause clips would need to rotate aswell, i think they only work X and Z axis. Still would be a cool feature to have the Y axis aswell, not that it couldn´t be done with blender these days. Simply convert old models to obj, 3ds whatever and rotate them in blender and convert them back to md2/md3. But idk maybe it can already be done and i or we just dont know it. Or its due the shadows which would doesnt have that axis, as you also see in your sreenshot (the palette where shadows would need to be mirrored) But yea was a try

[Felix-The-Ghost]

And to Felix, that seem to be a cool feature but i think its not in, cause clips would need to rotate aswell, i think they only work X and Z axis.

1) Clip entities can be placed manually in-game not just the centered auto-clips with models specified in the .cfg so you can often use multiple rectangular shaped clips to roughly represent the collision shape of an object

2) Models don't necessarily have to have clipping depending on the placement if they're somewhere the player/grenades couldn't reach or are flush enough to existing geometry

3) IIRC the current development version has rotatable and maybe even texture-able clips which is exciting though I haven't personally seen them in action (If I'm understanding them correctly)

[3XIX]

Alright, seems like i know shit about it...so adding that together with the time i dont have i shouldnt try creating a Map. But for those like me who don´t know, would be cool you could explain how you doing it so that if someone trying in future will be able to easier...maybe its already written somewhere so link would do aswell. In general i would still not use the Y degree axis to avoid non hitreg feelings if clipped, but thats just my opinion, btw isnt there any modeller anymore or someone who converts CC0 or CC by 4.0 license models? No one making money with it anyway so why creating every model yourself if millions did already, its pretty simple do to even for non experienced people. With tools like Blender and Milkshape 3D it probably is the fastest way to archive that. For those who dont know, since i cant do it rn

1. take a model (mapmodel) like barrel copy the tris.md3 to desktop (easiest)
2. open milkshape, import md3 and export as .obj
3. open blender, press A and DEL button then import the exported .obj
4. get a model from https://sketchfab.com/ and check its lisence (dont forget to add a credits/license of the creator txt file into the modelfolder)
5. import that model from sketchfab in the same blender session, resize the new model to fit the old      (press N to get a define menu)
6. delete the old model from blender session
7. export the new model to .obj has to be max 8192 verts (md3) and 2048 verts (md2)
8. import .obj in milkshape3D, check material slot...rename it to "skin" or "1" as
    example how you will call the texture later in the md3.cfg
    example md3.cfg:
   "md3load barrel.md3
    md3skin 1 1.jpg or md3skin skin skin.jpg (always double name if custom name)
    mdlshadowdist 0" - depends if you wanna use shadows
9. Go in milkshape3D to "tools, quake III arena, generate control file" and just save it as barrel.qc
10. export the model as .md3 and save it as barrel.md3 (.qc and md3 need to have same name)
11. put it into the model folder of the game (can be mods/packages/models/mapmodels/makke/barrel)
12. move the model texture "diffuse, albedo however its named" into the same folder and rename it to the example "1.jpg" or other like tga or png if you wrote the same in the md3.cfg (texture can be up to 4096x4096, make sure to lower res depending on model size, if needed compress it with tools like GIMP, to 50% quality as example to avoid to high VRAM usage, you can check the usage with tools like GPU Z when game is running)
13. check if your md3.cfg has the correct model and texture/skin name
14. start the game and check if everything is working correct (same model size, you wont have to clip         them anymore then, make sure you dont make them different sizes or you walking against invisible "walls"
15. You can get Milkshape3D 1.8.4 here, newest version requires payment as far i know
http://www.milkshape3d.com/ms3d/download.html
sadly not allowed to share, if you use any serial like this
"Name: Milkshape
Code: J2Bd1-TCdcZ3-T27d11d" its your responsibility. If you´re not poor asf i always recommend buying a lisence, so you will get new versions and support.
16. Dont use FBX in milkshape and Blender, they have different formats.

Same can be done with Quake II Modeller, but requires older Blender version, i think 2.5-2.6 to import/export .3ds files (atleast the easiest way without having to add scripts)
If you dont have a low poly model, you can use the decimate modifier to shrink the mesh, also you should check in Edit Mode/Mesh/Clean Up/Merge by Distance and Delete Loose if there are not needed verts, both in Blender...

[Mr.Floppy]

If it was ambiguous, the screenshot with the rotated models is a preview of dynamic orientation with the existing models in-game and no .cfg settings etc.

Sorry for the late reply, but yeah, I see what you've done there. Actually, I still remember a discussion way back on IRC about exactly this approach, however, didn't know someone did put it to test.

That's pretty awesome, but at the same time somewhat depressing as I start realising how close we were to a great update and how far from materialising this all seems today.

[Felix-The-Ghost]

That's pretty awesome, but at the same time somewhat depressing as I start realising how close we were to a great update and how far from materialising this all seems today.

I mean you technically could just compile your own server to host yourself that can run new mapping features but yeah optimally that'd come officially in a new version for everyone

I'm still not sure what's holding back the release of the current development version but I've been out of the loop for a while

[3XIX]

Ive tested a few things and checked why cube engine performs worse than quake engine, since that difference was way to big.

Screenshot of wqd and wvt´s low as possible 4-16:


Screenshot of wqd and wvt´s viewpoint shrinked with minfog:


Screenshot of all wqd and wvt`s in viewpoint maxfog:


Overall env verts/triangles doesnt have that high impact on fps, which means the most fps loss is due the map creation itself, unless this aint getting fixed or there will be made a new method for maps, its useless to mod anything out of it. That´s why im right that walls etc should be done with model mesh, same as you do in new engines these days. Map been 256x256 and as you see wvt just increase by 800 and fps dropped by 300, to drop the fps by 300 on the other hand you would need over 100.000 evt´s. (obviously if you run same system)

If someone could get the game working with clips/meshes instead of those heightfields like this:


It would change it but otherwise...dont know a way of getting rid of those cubes. (to walk on clip and spawn, and then texturize clips via model mesh) Im pretty sure if that could be done we would be able to get better graphics, higher texture etc by even getting higher fps as now, depends on the quad mesh walls later on (1 wall, 9 quads to have best use 1024x1024 instead of 4096x4096 textures) so VRAM wouldnt be that much higher. At the cost of lower fps (more mesh lower fps, lower mesh higher textures more VRAM balance of both, or more to the VRAM side since thats mostly dedicated and makes use of RAM where i think everyone has 2GB these days)

[(aCKa)Deathstar]

wooow i'm very impressed by your work! I'm inactive at this time, but i think you are one of the best model maker on AC from all the game times! Theses rocks are pretty beautiful! the playermodel and all the other stuff too! I think people are right saying AC is very limited, all people who tried to upgrade AC stuff always meet problems, but with you it looks diferent: you looks like understand verywell all the stuff to perform the game better! I hope you will suceed in this AC upgrade! I will like to help you, but i only get good artistic sense, i'm a big bullshit for the rest! If you suceed doing nice improvement and new way i will do a new map with your models! it will be a big pleasure! don't give up, i trust in your work, i'm with you 300% It's nice to hear other people experience about the game engine, but i think sometime in this community some people don't trust in other people and want to discourage. Don't let tge embittered people kill your dream, try what you trust, and never give up, if you suceed you will do a big part of the community very happy!!!! Peace bro :D sry for my bad english, i'm french xD

[Felix-The-Ghost]

3XIX -- can't tell what you're looking at in your screenshots

While in edit mode type /fullbright 1 or place light entities so we can see

Are you correctly marking cubes as solids?

Solids and spaces
The reason we are covering solids and spaces, before moving cubes, is because solids block visibility and do not need to be rendered, so they are more efficient to use in maps. The use of solids in a map is essential to creating a good map. Without enough solids AssaultCube maps would lag too much for the average user and would be unplayable. Solids are also easy to create and can be used to make walls with a single keystroke!
Tip: When you've finished your map, change any unused space into solids.

• To make a cube into a solid wall, press F.
  
• To make a solid wall turn back into space, select it and press G.
  

[3XIX]

Happy new year everyone,

thanks [aCKa]Deathstar for all the kind words but aslong there is no development team it would take us years to update graphics (models,animations etc) since its pretty much creating a new environment by just keeping the gameplay part, let´s say you could effectivly create 2 models each day(converting those to cube needs that time), with bug fixes etc. since you wont want to retouch them later on...would require much more time, you will have at the end of the year around 600 new models, which isn´t much if you compare model amount in games like CSGO etc. it wouldn´t even fill half of a map. In general there have to be much more people working on it to get a cool experience. Just look at CSGO maps they basicly retouched them over the years, half of them already been in there since the starts which means most of the maps been developed for 10 years or more (Dust, Mirage, Cache etc) Rebuilding that with even increasing the amount of models/details will cost more as anyone could spend atleast if he got a job and not much time to spend on modding, sadly.

Felix-The-Ghost, its hard to explain what exactly i mean since im not a native english speaker, but the screenshots show the difference between how much cubes you see and with that the effect on FPS which you can see down right bottom. I basicly tried to explain with it that the solid blocks (the player area, when you create a new map) in this engine called "heightfield" is the problem and i´ve not found a method to remove heightfields (the basic map platform) which in other engines you mostly would use a mesh ground and scripts to determ the max range where i player would get killed passing the line.

In every engine you have a 3D Room (I call it the skymap cube, 4 sides top and bottom textures) in this area we build our map, where in engines like unity, unreal, godot whatever we create a static ground for the player to stand on, creating a ray collision to make the player move on it without falling thru (basic method) and in assault cube they use this solid block method (the area created by the heightfield) which in this case costs us the most performance, even if they aren´t solid, tried that aswell. So i tried removing the heightfield and using a clip as ground and quad mesh as ground texture (clip doesnt has texture slot and wouldnt work that way) but to make this work there has to be a new engine code (kinda ray detection thing) which allow the player to spawn on the clips, since you would float in the air without "heightfields" since the game doesnt know where to place your player "model" when its not within the heightfield area or if there is no heightfield. It´s the 3D Room you see when creating a new map (its basicly a room in a room in assault cube)

Hope i could explain what i mean, you could try flying out of the heightfield/map and look away then check your FPS that should be the room your build your maps in (Do it in a new map, so you understand better how much it costs by just having a room in a room, which with quad mesh/clips would cost much less FPS.