After INVASION had been released and seen by the world, Bang! was kind enough to give us an interview about the work he did on the INVASION update. The following is his own words about how he put the 2fort map together, Other possible maps that could have been redesigned and his work on the Trailer.
Lets talk about the artpass of teufort. How many people were on the Team? How did you assign roles?
Ronin and I built the alien props set prior to the artpass being suggested by Valve, then I did the concept art for the artpass and Ronin modelled the the props for the blu laboratory, I made the red laboratory, destruction/crash sites and quarantine props, Freeflow, Void and Retro made posters which are all around the map, specific posters for the blu labs and some of which were used the teaser prior to release and Chaofanatic made the wonderful particles for 2fort.
This isn't the first time an artpass has been part of an update. Were you expecting Valve to offer it this time?
We weren't expecting to have to do an artpass as we expected Valve to pick maps from the mapping contest, but instead they asked to first do one or two map artpasses, we took advantage of this to introduce the story of the invasion visually. Although I wouldn't call this an artpass since we were not building from a blank canvas (blank map layout) nor using existing textures and props pre-made by Valve which define the TF2 artstyle, instead the focus was on designing new visuals that should blend and fit with the original map and artstyle/artpass. The new models would have to fit with the original's map's visuals while being distinguishable, we planned to make models to go over the original's map geometry, it wasn't about placing modular props or applying pre-made textures since the map already had that.
2Fort was picked because it was varied in its environment, it's an iconic and a classic map, in terms of gameplay it's different as it has versatility which you don't see as much in more recent maps, in my personal opinion it's by far one of the best and most detailed art passes in the game which defines the game's original artstyle best, everyone knows 2fort. The map also offers exterior and interiors, has great contrast with the opposition of the teams clearly defined by the bases facing one another and the objective of stealing one another intelligence briefcase.
The idea stemming from "how would each team react to the beginning of the alien invasion", 2fort is a great setting for this, as there is already a clear conflict between two teams and bases/locations, introducing a new conflict and telling a story visually of how each team would react was the focus for this map, as it’s an introduction for the invasion theme.
The new visuals introduced would have to fit the team's style, Blu team uses greys, blue and yellow colors, metallic materials, cylindrical and sharp shapes, concrete and metal, its a cold aesthetic and an industrial one everything is tidy in BLU's base, the reaction from the BLUto a foreign component (the aliens) would be systematic and clinical, so they would go in quarantine as it keeps the cold and clean aesthetic and a more serious tone.
The quarantine needed to feel a bit intrusive and improvised as an emergency, its messy but there's signs of Blu trying to clean up the mess, a lot of plastic materials were used to denote the "sterile" environment from the quarantine props to further accentuate the BLU's industrial and clinical character. The yellow color is quite predominant on 2fort invasion on Blu, as it happens to be the hazard/quarantine color but also to the complimentary color for the Blu's team (orange for the red team).
On the other hand the Red team uses warm colors, organic materials such as wood in their buildings, metal too, but the building shape tend to be more loose and imperfect, barns and farms, with a bit of nature, RED's team environments also tend to be a little more messy, hence accentuating this with RED not really doing much about alien saucers crashing into their base and leaving crash sites as they are, most of their base being made out of wood in 2fort, they feel more humane but also more fragile than Blu(using concrete buildings), so leaving the base destroyed suits RED's style better and contrasts with BLU's industrial and clinical approach.
We wanted to have big visual changes that blend with the original map scenery for the destruction and quarantine, we also wanted to turn the map to night or at least dawn time, and since the map was already art passed, it was a very different approach to how we could add to it without causing too much frame-rate loss in terms of performance and making it different while still keeping the original elements, there were also quite a few challenges adapting the map to a dark setting as some of the original models and texture that just didn't work nor look good in a night setting.
Could you tell us a little bit about the technical challenges you faced when giving this iconic map a fresh new look?
TF2 maps are composed of "brush" geometry which is directly created in the map editor Hammer and detailed with prop models which are created in a modelling software, exported and then compiled to the engine format so they can be loaded in hammer and placed there.
Usually the detail models are placed and duplicated in hammer, however recently, as demonstrated by Valve more and more large scale models are created directly in the modelling software to detail bigger parts of the maps than it was done before instead of using modular prop models, as seen in rd_asteroid. This approach is a lot more flexible and was great for what we were doing here since we were taking an already artpassed map and adding/modifying it, in order to be able to achieve continuity in the destruction details between brushes and props we would need to be able to first get that into the modelling software, convert brushes+models as a new single model, create the destruction and export it back and remove the old brushes/models to be replace by this, at the same exact location.
To import the map into the modelling software(Softimage aka XSI), I actually already created a VMF (valve map format) importer/exporter for Softimage I coded for earlier projects. With this I could import the map geometry (brushes) with their textures, the displacement (terrain geometry) and the prop models, as well as all the other entities, as they were in the VMF and the hammer editor.
Additionally in order to import each prop model and the instances over the map, each model first needed to be decompiled into a format that can be read and imported back into the modelling software. So I also modified a model decompiler of my own(also made prior to invasion) which would read the VMF get the list of each unique prop model, decompile the models into SMD format as well as look for the materials in the game's files and find the texture dependencies in VTF format which were converted to TGA so those could also be read and imported into Softimage.
I could then import all the map geometry and entities into Softimage in just a few clicks.
Other data was also imported (entities) so the map could be exported directly from softimage and compiled. Now that the map was imported into the modelling software, this was basically a map editor where you could also model, rig and animate props directly it in with the precision and tools of a modelling software.
The props were modeled in place, exported and compiled as a model and automatically re-centered as well as added into the VMF as a prop_static model so the position and rotation of the prop would perfectly map the map's geometry without the need to manually place it. Having the map imported into softimage also allowed to easily get the total polycount of selected props for a room in order to make sure the same polycount was maintained for the different areas of the map and new props to keep the map optimized.
The plastic covers on Blu base have actually some lighting baked into them based on the lights of the maps so its phong specularity behaves based on the ingame light sources, the plastic was also made so you would get high specular from far, but it would still allow to see-through upclose so the players still have a point of reference when moving around those areas.
Some of the original buildings on the red side are a combination of prop models + brushes, for instance the red "Happy Farmers" tower original was made up of several prop models and brushes, also decals, which had to be converted into a single model with the decals projected back into the final model, plus destruction.
In some maps the exterior of the level is actually extracted from the playable area and converted into a 3D skybox, this is an optimization, as the 3D skybox is 16 times smaller and is projected into the 2D skybox, around the playable area, in Hammer its placed in a separate"box" scaled down 16 times and in a different position than the playable area.
Having the map in softimage allowed to have 3D skybox scaled back to its original size and placed at the center, so it would match with the playable area, and when it was exported it was scaled down 16 times, as well as the models, and placed back in its original position(as seen in hammer, top left image). This way I could model and animate objects in the skybox 3D while seeing it from the playable area perspective to scale in real time and also allowed to easily create the saucers patrolling around the map and coming in and out of the mother ship, you can see their curve paths in blue and red on the bottom left image.
Why did you decide on 2fort to artpass instead of Upward? I mean we could have had a Sarlacc pit!
I actually worked on an Upward artpass along with 2fort, but we scrapped it for various reasons, one of them being, for what we wanted to achieve visually we needed to add more geometry to the map, upward is an open area map so the player's sight line is constantly filled with lots of geometry and separate objects to render, so adding to it was having a really bad impact in terms of performance, loosing a lot of FPS just for adding a bit more geometry, 2fort is mostly interiors and just the middle area being open but still simple, so it wasn't so much of an issue there.
This artpass would add on top of the original map artpass, as we're having aliens invade and destruction models, I actually ran quite a few tests early on to see if the map could have more geometry, I added some high poly spheres around the map to see how it impacted the frame rate, and unfortunately it was dropping the frame rate by twice or more than the original map. I spent more time trying to replace original models with lower detail version, or simply remove some of the original details, such as rocks in the 3D skybox which would be mostly covered by a mist layer anyways, to compensate for the new props we were adding, unfortunately this didn't make much of a difference, the way the map is and was originally built didn't allow to add much more geometry.
We wanted to have saucers flying around, the mothership, building debris floating over the map, the problem with this is that not only does it add geometry to the map, but it has to be constantly rendering since it's in the 3D skybox, it cannot have levels of details as you move further away from it, since it's actually in the skybox AND covering the whole map, only 3 levels of detail for your graphical setting, low, medium, high model detail but that doesn't cut it either. There were other issue specific to how the map is originally built, the skybox on upward cuts through the center buildings, it would have made it impossible to have the mothership cast light on the buildings unless we used a bunch of hacks which weren't performance friendly nor work on lower specs (DX8 I am looking at you).
Ultimately it was just going to be a reskin and shouldn't be at the cost of the map's performance when you can just play the original map and also for what we wanted to achieve visually, it would have meant sacrificing more of the fancy visuals for performance than it was worth, it wouldn't have made the map different and unique enough if we cut down on the visuals that much, if we were building a map from scratch, it wouldn't have been a problem to achieve what we wanted to do visually, but that's not what we were doing here.
Originally we wanted to have one artpass to introduce the event and set the tone, 2fort is meant to be the start of the invasion, after that it would take place in other map(s) where the mothership is destroyed, as a finale, Valve showed interest in other maps near the end of the project it just made more sense to put the highlight on the other new maps where the mother ship is destroyed, as much as we love upward and considerable time was spent on it, sometimes you have to be realistic and decide what it's best to the project.
There is more scrapped content, just like the trailer, things were changed or simply scrapped to make the style and story telling more homogeneous, we normally don't tell you much about it because it's a tease but it's actually pretty common to have quite a bit of unused content by the end of these kind of projects since we concept things, test them iterate and sometimes scraps that don't work or fit with the final content.
How much freedom where you give? Were you allowed to change around the mechanics of the map as Valve have done for halloween?
Valve didn't give any specifics, other than it had to be an themed artpass , its Valve's map so we didn't want to mess with the gameplay too much, some explosive robots were added in 2fort to alleviate for the infamous engineer sentry choke points, you can take down a sentry in one shot or two (shoot the robot then finish) with the explosive alien robots, this also damages nearby players, so the engineer eventually fixing the sentry, dies too.
Have you hidden any easter eggs?
Yup! most of them are in the red base, they are pretty small or subtle, you may have a hard time finding them.
How do you feel about 2fort now you've worked on it for so long?
I think for the most part we like how it turned out, personally as for anything I work on for a very long time, I tend to look at it under the microscope rather than the big picture and see every little flaw in it that others may not. I think when I work on something for a long time and once it's finished, I need to not look at it for a while before I can come back to it and not be so critical about it. That said it's actually great to finally see people playing in the map and reacting to it.
Lets talk about the SFM. How far into the film's production did you get involved?
Ronin and I started talking about ideas for the trailer months before we started the project, we wrote some scripts initially and decided to get in touch with Brent (aka CobaltGemini) to do the character animation for the trailer, Brent ended up writing a new script and doing most of the shots for the trailer as well as the character animation.
I made some concepts while Brent blocked out the character animations and some shots, once I got the characters blocked out from Brent I started building some basic models to blockout the scenes in various shots and get the scale for the barn right, meanwhile Ronin was modelling the barn and other assets such as the windmill, popcorn machines, project and later the barn. Ronin and I built the props for the trailer, I built all the scenery shots, modelling, rigging of the barn destruction, characters and handled all the compiling, importing/exporting.
While Brent was animating the characters i was working on other models and building the scenes and details for the barn and the b-movie. At this point, once most shots were locked, Hasan started working on the score and Harry on the sound fx, with sound it was really starting to come together.
Since we wanted to have the barn destruction and that would need physics simulation, I was processing all the models into my pipeline and compiling all the models and materials, so everything would be needed to be imported into the modelling software in order to rig and create the destruction simulation and have the objects at the exact same location between shots, before and after destruction and the pan-out to earth orbit.
Otherwise some detail props specific to certain shots were actually placed in SFM, the barn and objects inside and around it were all baked into a few models that were automatically placed at the center of the scene when exported and imported into SFM.
Once I got the barn model from Ronin, I then prepared it for the destruction simulation, sliced the structure beams and the planks into many pieces, every piece was given its own mass depending on the size and volume, the structure of the barn used physics constraints and I scripted it so the joints would break based on the tension they were under, once it passed a certain threshold they would detach from their neighbor.
The global gravity was set to zero and some additional forces were used to pull the barn up as well as a spiraling force, it took quite a bit of tweaking to get it to break up and lift gradually, while also getting the derbies to circle over and around the barn's center so when they wouldn't fly or fall over where the mercenaries are standing and also to show that the saucer's beam isn't trying to pull in the mercenaries up.
Once the physics were simulated, it was cached and converted to bone animation, every piece of debris had a bone attached to it and its position/rotation was keyed for every frame of the animation. Finally the barn was split into several models (structure, several wall pieces, roof, objects, rocks) because SFM has a limit of 256 bones per model and compiled as a model for SFM.
Once the scenery models were done and placed in SFM and I did the lighting after that Chaofanatic made the particles effects, Harry did the final render, montage and sound FX.
How close was the final film to the idea you originally had for it?
Pretty different, originally we were going to make something a lot shorter but we ended up making something more complex than we initially planned to, under one minute, but in retrospect putting the effort we did on the trailer was worth it, it got Valve's attention and support.
Is the "alien invasion" the sort of film that you usually watch? Did you pull direct inspiration from other films we may recognize?
I would definitely watch the type of B-Movie seen in the trailer that Brent came up with, there are also some references/inspirations, particularly the ending shot is reminiscent of Mars Attacks intro and Independence day with the mother ship.
As an SFM film maker are there other content creators you would like to work with? What about real world directors/cinematographers etc?
I love movies, it's a great medium that combines multiple disciplines and ways to convey emotion through visuals, sound and rhythm, I experimented doing animations for music videos a few years ago and considered to work on movies, but over time seeing things evolve, to me the ultimate medium for now seems to be videogames, since can be cinematographic and on top of that include interactivity, it's more immersive and that's what I love about these mediums immersing other people or yourself into a handcrafted imaginary world.
That said I have been wanting to make SFM shorts and post them on youtube, just for fun but so far I have always been too busy between projects, so it's hard to find the time, but I'll get to it.