Design lessons from the procedural puzzle pieces of Remnant II

Gunfire Games founder and Remnant II creative director David Adams has a personal problem with some games: he can't replay them. He considers himself an "explorer" of a player, and wants to see what's just over the horizon in a game. For him, a game like The Legend of Zelda: Tears of the Kingdom only feels memorable that first time he plays it.

So when Adams and his colleagues set out to make Remnant: From the Ashes, the quest to make a replayable game was personal—and didn't originally include an emphasis on multiplayer. "It just happened that [the game] also had really cool multiplayer," he recalled. "One of the cooler aspects of the series is that you can jump into your friend's game, or jump into an online game, and you're seeing stuff that you have never seen in your game, even if you completely beat it."

Remnant II beat the performance of its predecessor, selling over a million units in just a week. So what's so special about the game's procedural generation? What are the advantages of designing set pieces some players will never see—and how do you seed these levels with encounters that won't feel boring?

Adams said the key is balance—some pieces of Remnant II need to be blazingly unique and distinct from everything else in the game, and some parts need to be so unmemorable that players won't realize they're walking on familiar ground.

How does Remnant II's procedural level design work?

When players start a run in Remnant II, the game generates a mix of hand-crafted levels that are essential to the game's story. But after that, everything's off the table: as players venture between different biomes, they may realize that the biomes don't spawn in the same order. The levels inside them don't have the same puzzles and bosses, and if they pay close attention, the corridors they're wandering through don't spawn the same way each time.

Those corridors are the ones Adams doesn't want players to pay attention to. "I always tell the designers, 'In a weird way, there can be nothing distinct about these things, because you don't want people to remember them.' You don't want them to have repetitive elements, but you also want them to be fun."

Laughing, he called that a "total contradiction," because those sections still need to be "good," but if they're too good, the player may think, "I remember this exact thing that happened to me already."

By contrast, he says the main biomes of Remnant II need to rely on very strong visual storytelling, so players can associate levels with their defining features. The jungle area of the game is laden with Indiana Jones-style traps. Another dungeon has players fighting on top of a train, and another one is an underground maze where walls go up and down.

He gave one example from an area called N'Edru, where the locals discovered "a supermassive black hole that may hold the mysteries of creation. We have this one siege where there's these giant claws that come down and pull things out of the wall," he said. "It's visually interesting and memorable. That's how we try to get people to have these cool moments within our game."

Events in Remnant II are built around a classification tag that sorts content into events and quests. At the highest level of the game, there are story quests that govern the storyline of a single biome. Then there are what Gunfire Games calls "side dungeons," which can be "pretty much anything," because they're structured as one-off adventures.

Events take many forms, but Adams called one type called "injectable events," which are small vignettes injected into a single "tile." The game will take all of these randomly selected variables, mash them together into a level, throw in a few unique mini-bosses and bosses, and boom, a unique player campaign is born.

Adams said that the storylines selected for the biome generate the major variables that help the game select which side quests are appear. Some can then be anchored in the story of the main quest, while others can be standalone and make the world feel bigger and more expansive.

For instance, in the world of Losomn, there's a chance that a specific clock tower will spawn in the map. "If that generates, the game will generate another zone, and the clock-keeper's house in that zone. There's ways for quests to inject things into other quests or different areas of the game."

What's interesting is that all of Remnant II isn't a mush of interlocking systems that try to gauge each other's potential and optimally trigger every time. Adams said that the game's side quests are not tied into each other to create an ideal flow. "They're designed to be isolated one-off events," he said. Side quests can tie into each other indirectly through secret quest items or other easter eggs, but they're still designed to be self-contained events.

The system Gunfire Games developed might sound like a dream to designers who want to create infinitely replayable content. But Adams is the first to admit it's not a perfect system—and that there are drawbacks that shape how the Remnant series has come together.

The challenges of Gunfire Games' procedural generation methods

Adams told a story from before the launch of Remnant: From the Ashes. In the weeks leading up to launch, the Gunfire team worried about reviewers giving the game a low score because they encountered a bad mix of procedurally generated zones. "What if some reviewers just get all the worst stuff?" was a question being asked around the studio. Adams admitted that while he hoped the game is an A-grade experience, there are some B and C-grade moments lurking in the background.

"You could absolutely have someone who got the least cool quest in every single biome," he admitted. Even so, the studio rejected the idea of tweaking the formula for game reviewers so they'd see the best zones. "We didn't put anything in there to protect against that. We were like 'nah, we're committed. It's completely random, deal with it.'"

Giving up control over how levels spawn also means throwing away many tools that game designers and level designers use to define fun experiences. Adams, a veteran of the Darksiders series, talked about how he used to be able design levels knowing what the big climax would look like. Knowing how a level would end let designers tease the climax as players progressed through corridors fighting enemies, and tune the combat so the big fight would feel like a natural conclusion to the area.

When creating similar encounters in Remnant II, Gunfire's designers have to act like they can't know what the player has done up to that point, or what they'll be doing in the future. "You have less control over where the players are coming and where they're going," he noted. All they can do is tee up potential entrances and exits, using the few hand-scripted story missions as touchpoints to understand the flow of the narrative.

Speaking of narrative, Adams was rather frank about the strength of the Remnant series' storytelling. "It's hard to make the high-level story super interesting because at the end of the day, Remnant II is a game about individual little micro-stories," he said. Because of the layers of standalone-storytelling that prop up each world, it's difficult to build a big, satisfying arc off of that.

That makes the Remnant series more of a pair of short story anthologies than a big fantasy epic. In fact, Adams said the goal with Remnant: From the Ashes was that it wouldn't have a main storyline. "It was 100 percent just little vignettes and little story," he said. The problem? Everyone who tested the game said they needed a bigger reason to drive them through the adventure.

"For whatever reason, that just seemed to gel better for people if there was a main story thread, even if it wasn't that elaborate," he said. Gunfire's ultimate goal was to try and make a game that felt like Homer's Odyssey—an adventure that possesses a clear end goal but is defined by the encounters that happen along the way.

He says the studio wants to improve on that in future Remnant games, but it's still "bittersweet" to hear players praise the stories of the different game worlds while lambasting the overall narrative arc.

"People expect the main story to be the showpiece of a game, and it really isn't in our game. It's this thing where we need it to keep people through, but at the same time, we feel it's better to focus on the individual micro stories because that's the nuts and bolts of the game."

Can every video game be procedurally generated?

The dream of a certain subset of video game executive (and let's be fair, some designers) is that video games might one day be assembled not as handcrafted, repetitive experiences, but that developers will instead create increasingly randomizable tools that will perfectly generate stories, narrative arcs, and game worlds for players. Advances in generative AI technology have fed this particular fever dream.

With the procedural generation of Remnant II, Adams is one of the industry leaders who's seen that dream up close—and he's not convinced that dream can be realized anytime soon. "Randomization can only take you so far, and in the same way, I think AI can only take you so far," he said. "It can get you to a certain level, but currently you hit a plateau with what you can do with it."

He said that if developers want games to be "really good," they have to tailor handcrafted content somewhere in that process. That might be improving on procedurally generated environments, or it might be hand-crafting "point of interest" tiles like those in Remnant II. "If the game was completely randomized, the experience is just kind of 'all right,'" he said. "I think people will experience a similar thing with generative AI content. If you use it purely as is, [people] would just say 'yeah, it's all right.'"

Of course there's a flip side to what Adams is saying. If developers use generative AI tools as they do other procedural tools, there are opportunities to make replayable experiences. They'll just still need that human touch, that sense of wanting to show a player something unique and exciting, to really be successful.