More Source Engine Experiments

 Allright, some more Source and Source Filmmaker updates. Also a post with some more technical stuff for those interested. I've started experimenting with Source in early february, since I wanted to do more in Source Filmmaker, wanted to experiment with the Source Demo Smoother (basically the only way to create cinematic playbacks of demo's pre 2012, before SFM was officially released by Valve) and eventually I wanted to move on to making my own maps and models for machinima in Source.

Learning Source and understanding Hammer has been a very interesting process so far. And to be fair although I've experimented a lot I have not yet built a custom map of my own. For some reason that's just the way how I like to do things. Just spending hours on decompiling maps, reading books, watching tutorials, reading mapping forums/discords/blogs, loading game maps in Hammer and understanding how they consist of various brushes and in/outputs is more interesting to me and also in addition on how to utilize various other tools, such as VTFEdit (texture editing), GCFScape (exporting raw materials out of source games), and Crowbar (compiling/decompiling models). It brings me up to speed much more quickly by doing it the 'reverse engineering way'. Like I mentioned in the previous post, the SFM was very daunting in early 2012 when it was released. Now that I have some 3D experience, the SFM was already a lot easier to understand. However, the penny didn't fully drop until I started to learn Source and now I'm finally up to speed.

In the previous post I shared a short video which I would like to call 'Inferno'. A ten second short made as a test that was made in one of the left 4 dead levels, including some props, lights and particles from the game. The one thing missing were some zombies that would crawl around on the map. I managed to get every single map from Left 4 Dead and Left 4 Dead 2 working on the SFM, but it would crash every single time after trying to import a common infected model. Long story short, and trust me I have been facepalming myself for so long since the solution was so easy, I finally got it to work.

In the same zip file I downloaded with fixed maps, there were also some directories with fixed materials and models. I just ended up looking over them and trying to do everything myself the hard way, by decompiling and recompiling models and animations. Literally last friday I had a brainwave: "I'd better just check the fixes folder again, just to be sure". And there they were, all of the fixes I needed including a installation file. I feel stupid, like really stupid because I had spent a good portion of a week trying to fix it. Here are two poster renders on 4K resolution with the new infected models in the frame. I managed to lock a couple of glows to the eye sockets of the infected, which makes it look like their eyes have a glowing red color. Problem is that they weren't properly locked so sometimes it would create weird artifacts. The first photo contains some glowing eyes on the clown. I turned them off for the second poster. The renders are coming straight out of the SFM; they're not edited in any way.

The primary purpose of the experiment was to build something cinematic in Source Filmmaker while placing emphasis on lighting, models, and particles. I consider this experiment to be completed and a success. 

Now. To answer the question: "why Left 4 Dead 2?" I'm working on a project in both Left 4 Dead and Left 4 Dead 2 within the SFM. It revolves around a visual study of all of the maps found within the game in a highly eerie and cinematic manner. This experiment primarily served as one of two tests. I am currently still working on the second one. It revolves around the Garry's Mod map: zs_swimmingpool_V2 by AlexG. I had used it in one of my very early Garry's Mod Skits back in late 2009 and it is a personal favorite of mine. Underneath you can find a couple of gameplay screenshots of what the map looks like in Garry's Mod.

Since a couple of years, early wip builds of Source 2 are available for download, for instance in DOTA2 Source 2, or in SteamVR. With the release of Valve's VR game Half Life Alyx in 2020, a more up to date build was released. Source 2 is the successor to the original Source engine used for Half life 2, Portal and Left 4 Dead. The SFM (called S2FM) found in that engine has a lot of QoL improvements, such as 64bit support which allows SFM to finally stop crashing after using 4GB of ram, better lighting, better interfaces, scripts etc. 

My plan was to originally do the L4D and L4D2 visual study in Source 2 and the S2FM. This would require masive efforts as all of the maps, textures, models, particles and props from the orginal games had to be manually ported over by hand to Source 2. Before accepting such a daunting task, I wanted to try to port this map over to Source 2 to see whether the final result was that much better. Unfortunately, Source 2 is incredibly rough around the edges. Having been left to the SteamVR build (as I am not currently interested to spend 60$ on a game I'm unable to play just to see whether the SDK is better), I succesfully managed to port all of the models and textures over. Compiling also worked with an exception of water. The water inside the pool did not render properly for some reason and changing the shaders gave various errors in the material editor. The main dealbreaker was the S2F. The map wouldn't properly load at times unless I used a workaround. Recently i have even been unable to succesfully load the compiled map at all. Second issue with the S2FM is that the improved lighting barely works at all. You're left to do lighting in Hammer instead of S2FM. So if I were to adjust a light, I would have to remap the light and fully recompile the map again for a full compile. That would take 20-30 minutes again by itself. Although the lighting would be considered better in general, I decided to use the original Source engine. Not only is there a lot more documentation on offer in terms of troubleshooting, but simply because 'it just works'. In Source 1 I can light and adjust on the fly. I'm a bit afraid that working in Source 2 at this very moment will require much more time than I would be willing to put in. The Source 2 experiment did not go in vain. Hammer and S2FM in Source 2 are a dream. Especially Hammer considering it was rebuilt from the ground up by Valve. When Source 2 is finally stable and a proper SDK is released by Valve, I am sure to give this a proper go. Hammer 2 is much, much better then Hammer 1.

All of the ported models in the Source 2 asset browser

Overview of the Source 2 port in Hammer. Unfortunately the water didn't display properly in-game.

The map in Hammer 1. Note how the water and the truss under the roof are now fixed compared to Source 2.

Source 1 turned out to be the better choice in terms of using and adjusting maps for use within the SFM. Above is a picture of the map in the original Hammer. Now we're getting to the actual second experiment. I adjusted a couple of parameters in the map and recompiled it in order to make it a nighttime map. It succesfully loaded in the SFM. The second experiment is to make a 30second - 1 minute visual study of the map with lighting animations and particles in order to compliment the L4D/L4D2 visual study. Both projects will be done alongside each other. Although I just see this as a simple test I do wanna take some more time for this and to create something interesting on this map. Depending on the end result I might share the film on my channels. I'm still in the process of adjusting a couple of things, such as completely closing the roof, adding a couple of windows and getting rid of the lighting props, so that I am able to turn off the lights on the wall by will after adding them manually in  the SFM (since they're now baked/compiled into the map. Now they can only be on by default)

A few screenshots of a quick lighting test I did in the map. More to come. Cheers!