I got this gig when Derek Yu DMed me on Twitter! He said they originally intended to hire me for the announce trailer, but ran out of time and ended up doing it internally. He later remembered to contact me for the gameplay trailer after I tagged him on my recommendation for his Spelunky book.
I'm going to gush a little here; this was the rare gig where I immediately start pinching myself because Spelunky is one of my favorite games. Veteran players will scoff at these numbers, but between my Xbox 360, PS3, PS4 and Steam account I've played well over 200 hours. When I used Spelunky as an example in my GDC talk about game capture, I allowed a small part of me to indulge in the fantasy of working on a trailer for Spelunky 2.
Sufficient to say this was a dream gig!
I got my first build of the game in late June, but production didn't start in earnest until August. Derek sent me the game; I played in my spare time to get a feel for it, and started a list of what to put into the trailer. The game controls the same as the original, so all my muscle memory transferred over with no problems.
The original launch trailer Kert Gartner made for Spelunky HD is a branching trailer with a narrator speaking in verse; I knew this wasn't something I would be able to do without writing help, nor was I sure it should be done again. Instead I focused on what the experience of playing Spelunky meant to me.
Spelunky is known as a very hard game. It's a very precise, deliberate, and unpredictable game, but the suddenness of death makes it very funny. What encompasses Spelunky to me is when one unexpected thing on your hundredth or thousandth run causes a chain reaction ending in your death.
One mandate from Derek was to highlight the characters, because as of Spelunky 2 they all have names! In the game they're still functionally the same, but this montage of the characters provided a basic structure for the trailer. In addition to the characters, he wanted to show the story introduction, and mention the online multiplayer!
These key points are how I arrived at this structure:
Open with an exciting long take of gameplay that ends in a sudden death.
Story introduction of Ana landing on the moon
Brief gameplay of Ana
Introduce online multiplayer
Character specific montages
I made an outline version with placeholder text, rough timings, and music notes so Eirik Suhrkecould start making music for the trailer. I wouldn't be able to do final capture until August because Derek and Blitworks were working on the game for its debut at PAX West, but in the meantime I could play it, and plan out my shots. Since Spelunky's levels are randomly generated, the most I could block out per shot were the character, setting, actions, items and enemies. The level layouts were by luck of the draw.
I was most worried about capturing the opening sequence. I decided on a long extended sequence because I knew Spelunky fans would be chomping at the bit to see any gameplay; I knew they'd have no patience for a cutscene introduction. I didn't want the death at the end to look like I killed myself deliberately; I wanted it to look like a typical accidental Spelunky death that I just happened to capture. I initially assumed I'd stumble upon a moment like this simply by playing the game a lot. I used Nvidia Shadowplay's "Replay" feature which constantly records gameplay, but doesn't save a video file until you press a hotkey; you can set the interval of time you want it to buffer from 30 seconds up to a few minutes. This allowed me to save hard drive space, but never miss a moment.
My first approach to get the intro was to play in the first area of the game somewhat recklessly, but after many attempts I didn't get anything interesting, or the shot was too long because the actions I captured were spread out too far. I shifted gears and started each run with a jetpack/shotgun combination. In Spelunky HD that combination is pretty much the strongest loadout a player can have; I thought it would be funnier to show a powerful character die in a silly way. For a while I stayed in the first area of the game, but later changed the setting to the more hazardous and spacious volcano region.
One snag I hit in the volcano levels were falling platforms that impeded my progress through the level; these stay in place until something touches them on top, then after a couple seconds they fall. There are certain areas where they’d add 1-2 seconds of standing still, and breaking the momentum. I previously read Derek's book about the design of Spelunky which is nicely summarized in this video by Mark Brown so I took this knowledge to dig into the game's files to customize the level layout for the sections that might block my path. I replaced a few symbols in the files, and those falling platforms were gone! This didn't end up being necessary for the level in the final capture, but it was fun to customize the game to have a better chance of a good shot.
I played and played until I found a level which allowed me to take a lot of interesting actions in a quick and exciting manner. Some of the actions I chose to do in this opening were inspired by accidental things I did in other plays, like killing a caveman and stealing his mount. I can't put a precise number to how many hours I spent to capture the opening, but it was quite a lot. Fortunately there was a debug option to reload that level that allowed me to play it over and over again to get the perfect take.
This is the rare trailer where I didn't worry about teaching the game to the audience at all. Spelunky already communicates very well visually because it's about hitting and jumping on things, but more importantly I wanted to cater to the people out there I knew were dying to know what is new in Spelunky 2 after waiting nearly a year to see gameplay footage. No sense in being coy when you have an audience waiting for the opportunity to pick the trailer apart, and share it with their friends.
After playing quite a bit, I made an inventory of the new things in the game. Via the debug menus I had literal lists of the power up items, level biomes and some of the enemies. So I ranked them from most mundane to most exciting, and also cleared with Derek what was okay to show and what wasn't. I then decided how to form the waves of the montage. I didn't want the trailer to start with the most mundane and end with the most exciting; I wanted to it to rise and fall in excitement from beginning to end.
I thought of each section of the trailer like a mini teaser unto itself. For example, I didn't want all the clips in one section to be from one biome. To entice the audience to keep watching, I made sure the last clip in each section was set in a different biome. For example, the first montage ends in the jungle, but the beginning of the second montage one goes back to the first area. Hopefully the audience will think: "Wait, where'd the jungle go? Go back to the jungle!"
It wouldn't be as interesting had I structured the trailer mini-montages as:
This point by point structure wouldn't set the expectation that something new was coming, but by putting in one shot per montage in a new biome I feel like it had a better chance of hooking the audience. Essentially, I waved something shiny in front of their face, then yanked it back just as quick. Would that new area get shown again? They have to watch in order to find out.
In addition to new enemies and power ups there are some fundamental things about Spelunky 2 that make it different from the original that a new audience might not understand the implications of, but I knew Spelunky players would watch and think: "Wait, WHAT??? Go back to that, what just happened??" Namely liquid physics and equipment that offer new mobility options.
The next step was to make a cut with some first pass capture that showed what I wanted in each shot, just not captured well in terms of composition, performance or timing. From when I started to when I performed final capture some features like the moles were added in, so I incorporated them into the final cut.
For the coop scenes my friend Mark (who is also a HUGE Spelunky fan) happily helped me out over a few weekend afternoons. There was a debug option added later which hid the flag that is visible when playing in local multiplayer.
Once I figured out what was in each shot it was simply a matter of loading and reloading the levels until I ran into a level layout that I found clear, aesthetically pleasing, and gave me enough room to perform the actions I wanted. Derek showed me how to modify the probability of certain enemies showing up so I could save time. Some shots like Ana getting chased by the rolling enemy and the shopkeeper climbing the ladder I captured by sheer luck.
The shot of Margaret jumping on robots, the turkey riding caveman, then riding the turkey into the explosions in particular took a LOT of takes. I had some cheats to spawn those enemies, but jumping from one to the other, then to the turkey and jumping over the robots and "Yoshi jumping" off the turkey so it would explode AND to have the cooked turkey clearly visible was very difficult. In the end I made it a bit easier on myself by cutting together two separate takes.
Margaret being swept away by the water is another that took a while. I originally had a take that I really liked, but the swimming animation wasn't visible so I reloaded until I found a pool big enough with spikes below. I ended up bombing away the terrain below the pool to make the drop bigger because I really wanted to see the water flow.
The teleporter backpack shot with Colin is there explicitly because my friend Mark said the speed running community would go crazy over that because it means the ability to teleport while holding an item. I'm terrible at using the teleport, so I just did take after take, and hoped I didn't teleport into a wall and die. Derek contributed one bit of capture where Colin gets swarmed by monkeys. In that version they steal a bomb which then explodes and he falls on spikes, but sadly there wasn't enough time to show all of that.
After finishing the build for PAX, Derek made the character illustrations, and I used the game textures to make backgrounds for the graphics. I originally experimented with a look from this Video Copilot tutorial, but it contrasted too much next to the hand drawn art of the game. Instead I roughened the edges of the text a bit and added some paper textures. I made some light text animation, added in particles and they were good to go!
The trailer was well received, and gave plenty for die hard fans to dig into frame by frame. Towards the end I worried the trailer is too fast to follow, but I'm pretty confident the most important parts of the shot are clear, and everything else in the shots are details people can see by rewatching it. There are even a handful of features in the game I didn't realize were in there despite me going in frame by frame to make a list of what new features are in the trailer.
I met Derek and Erik at PAX West, and had a lot of fun talking about the game and the trailer. I don't know when the game will be out in 2019, but I can't wait to play the finished version!