The Making Of Split Fiction: Behind-The-Scenes Of Character Design & Animation

When did you both start working on the game? Before Split Fiction took shape, were there any ideas for your next game that you considered?

Felix Strandberg & Jacob Larsson: We started right after It Takes Two wrapped in February/March 2021. At that time, there were naturally several early discussions about what our next project could be, but once the initial pitch and story outline for Split Fiction started taking shape, it quickly became clear that this was the direction we wanted to pursue. 

Early on, we were presented with a broad overview of the story and characters, but much of what you see in the final version of Split Fiction actually grew and evolved organically over time.

Hazelight has gained recognition for its innovative gameplay and unique approach to co-op experiences. Have you incorporated or improved upon any particular mechanics from your previous game?

Felix Strandberg, Senior Animator at Hazelight: Personally, I spent a lot of time on the core movement system of the characters, one of the few things that stays consistent throughout the game. We had a good base with It Takes Two, but we explored many directions on how to fit the kind of game Split Fiction was becoming. Requests from different departments, such as design, art, and narrative, dictate what kind of movement we can have.

More realistic characters, in one way, meant fewer crazy moves than the dolls of It Takes Two, but we didn't want the players to feel like they were held back. At the same time, characters who can move fast and jump far need bigger levels, which increases pressure on level design and environment art.

Making the players move slower and making their jumps less vertical meant we could build tighter levels than in It Takes Two, making the decrease in the distances you can cover less noticeable. For example, in Split Fiction, your run speed is 5 m/s, whereas in It Takes Two, it's 8.

To differentiate the two games further and to give movement options for covering larger distances and segmenting gameplay areas, we created a more in-depth system for swinging and grappling than It Takes Two had.

What made you choose Unreal Engine 5? Can you provide some examples of your custom tools and systems? From a technical perspective, what are the biggest lessons you learned between developing It Takes Two and Split Fiction? Have you faced any particular challenges while working on split-screen co-op?

Jacob Larsson, Character Art Director at Hazelight: One of the main reasons we chose Unreal Engine 5 was the flexibility and powerful tools it offers, which allowed us to push the visual fidelity of our characters without sacrificing performance, something that's especially challenging in a split-screen setup. Compared to our previous project, It Takes Two, we wanted to raise the bar in terms of character quality and storytelling, and Unreal Engine 5's feature set made that possible.

A big part of our work was focused on developing a robust character pipeline. Our tech team built several custom tools and systems on top of Unreal's existing functionality, which made our work in the character art team much more efficient. For example, they created a data-driven setup that allowed us to manage multiple variations of skeletal meshes and materials in a structured way, which was crucial when dealing with cinematic close-ups and gameplay readability at the same time. We also relied heavily on Unreal's automated LOD generation system; it was straightforward to adjust parameters and settings to get the visual results we were aiming for, which made a significant difference in maintaining clarity and performance in split-screen scenarios.

One of the biggest lessons we took with us from It Takes Two into Split Fiction was how important it is to think about character readability and performance early in the process. Split-screen forces you to make tough decisions about where to spend your visual budget, and we learned how to balance detail and optimization without compromising the emotional impact of our characters. Another key takeaway was how valuable it is to have technical tools and systems in place that empower the art team to iterate quickly. Being able to control aspects like LODs and mesh variations directly in-engine allowed us to focus more on creative problem-solving rather than technical bottlenecks.

Overall, Unreal Engine 5 gave us the technical foundation to deliver on the visual ambition we had for our characters, while the custom tools developed by our tech team ensured that we could do so efficiently in a split-screen environment.

Felix: Since we had built many pipelines and tools for Unreal Engine 4, it was natural for us to go on with Unreal Engine 5. It was early enough in development, less than a year in, so upgrading was not too difficult a process.

One huge deal for us is the use of AngelScript with a custom open-source plug-in for integrating with Unreal Engine. This is easier to get into than C++ and more optimized than Blueprints.

One thing about working on a split-screen is that everything in the game is rendered twice, which eats up a lot of performance. Also, having one character able to see the other from a different angle means that you can cheat less than in a single-camera game. You can't teleport the character around, hiding with camera cuts, and you need to be more mindful about things like foot sliding, etc. In a single-camera game, you can easily hide the feet by just placing the camera above the hips, but in general, for most departments at Hazelight, having multiple perspectives means fewer shortcuts.

The visual style and atmosphere in your games vary greatly from project to project. Split Fiction blends different visuals to match its story but ultimately aims for a more realistic look. Are there any specific inspirations behind the visual style of the game?

Jacob: I'd like to start by saying that craftsmanship is at the heart of everything we do at Hazelight. We don't just make art, we build it with intention, care, and pride. Every character, environment, and asset was created with meaning, and we hope that sense of purpose and passion is reflected in every detail.

That philosophy was very much the guiding principle for Split Fiction as well. From the start, we knew we didn't want anything to feel generated or scanned. Instead, we aimed for a semi-realistic style where the artistic touch behind every asset would shine through. We wanted the game to feel grounded and believable while still maintaining a sense of stylization and personality.

In terms of visual inspiration, we pulled from a wide variety of sources. Everything from contemporary sci-fi and fantasy films to classic adventure games. We also spent a lot of time researching different environments, costume designs, and visual motifs that could help reinforce the narrative themes of the game. However, much like the story itself, the visual style evolved organically throughout production. Early on, we explored many different directions, some more stylized and others more grounded, before eventually landing on a look that felt like a natural extension of the game’s world and characters.

This project has also been a massive step forward for us, both in terms of scale and complexity. Not only did we create more character assets than ever before, but each one came with higher demands in terms of design, style, and technical execution. From humans to creatures to mechs, big and small, we did it all. And to make that possible, we assembled an incredibly talented team of artists. It's thanks to their passion, creativity, and collaboration that the vision behind Split Fiction's visual identity came to life. I couldn't be prouder of the team and how fearlessly they tackled every challenge along the way.

The lead characters, Mio and Zoe, are full of personality. Could you share about earlier versions of their designs? How do their various looks reflect their character traits, and does animation take that into consideration?

Jacob: Once the personalities of Mio and Zoe had fully taken shape, we felt quite confident in the direction of their designs. But it was a process to get there. In the early stages of development, both characters went through several iterations, not just in terms of clothing and visual style but also in age, personality, and tone. For a long time, we explored a variety of looks, some more stylized or exaggerated, before eventually landing on the grounded and relatable versions you see in the final game.

Their clothing styles were also carefully crafted to reflect who they are as characters, something you can already pick up on in the opening minutes of the game when they appear in their real-world outfits. Mio's style leans towards jeans and leather, punk rock, edgy, and slim-fit, while Zoe's is more playful and casual, with a pop-inspired aesthetic and looser, baggier fits. These personality traits and stylistic choices were something we tried to carry through all of their costumes, whether the theme was sci-fi or fantasy. Mio's outfits often feature a more fitted silhouette with sharper, angular lines, while Zoe's are typically more relaxed, with softer and rounder shapes.

This contrast in their personalities and visual style also ties back to the geometric language that represents them throughout the game, with Mio symbolized by a triangle and Zoe by a circle.

Felix: We tried to incorporate their personalities into their animations as much as possible. Most of the time, they have unique mechanics, but even when they share actions, like the core movement, which is identical, we wanted them to behave like themselves. So, Mio is, in general, a bit more controlled and precise in her movements, while Zoe is a bit more loose. Someone early in the project said, "like Jet Li and Jackie Chan," which I guess might apply.

Given that the world of Split Fiction is packed with distinctive worlds and mechanics, what was the animation process like? How did you keep the moves from feeling repetitive, and how did you reuse them without losing variety? And how did you keep the characters true to themselves, even when they became non-human? Could you share more about your workflows and the tools you used?

Felix: The animation team at Hazelight is quite small, with 7 animators, and there is no division between a gameplay animator and a cutscene animator. The amount of variation in gameplay, coupled with around 2,5 hours of cinematics, creates a huge amount of animation work, just counting frames.

To manage this, we, of course, use some external partners to varying degrees in cinematics, which is a bit of a closed environment. We can set up a scene with a first pass and then pass it along for detail passes. Then it goes back and forth with feedback, with varying degrees of internal passes as well. For something like gameplay, which has more dependencies and might need to be iterated on and tweaked to fit a lot more, we resolved to do it all ourselves. Since the animators are so involved on the technical side (see previous answer), it is often necessary to jump between animating and scripting, making it quite time-efficient if the same person can attack an issue from two fronts, so to speak. Also, sometimes gameplay animations can look strange in the animation software without input, etc., affecting the character in-game, so it is often easier not to worry about trying to communicate all of that to someone external.

We animate exclusively in MotionBuilder, which I suppose is quite rare. It is nice to have IK/FK switching built into the core functionality, and the Story function is nice for working with motion capture data. Working with multiple animation layers is more integrated and less resource-heavy than what I've experienced in Maya, and using MotionBuilder's takes to separate animations helps with keeping a work scene structured and manageable even when it contains all animations for an entire system.

We have a lot of internally developed tools and a solid internal pipeline set up by our great technical animation team of 3 people, whether we work on gameplay or cinematics, motion capture, or hand key. The definitions of what you can have MotionBuilder recognize as human are also quite loose if you think creatively. This makes it easier to have rigs use a similar template, thus making it quicker and easier to get into working with a new character.

What was your workflow like, and what tools did you use during character production? Did you utilize any custom ones? Which parts were the most difficult and required the most iteration, and which were the easiest? Could you tell us more about the visual design and functionality of the girls' costumes?

Jacob: We like to keep our workflow relatively simple and streamlined, relying on some of the most commonly used software in the industry. For us, it's important to minimize the amount of time spent navigating menus, adjusting parameters, and organizing files and instead maximize the time we can devote to the creative, artistic side of production. One thing we're especially proud of and grateful for is the incredible suite of Hazelight-specific tools developed by our technical department. Thanks to their efforts, implementing and iterating on characters, from Maya to Unreal, is just a few clicks away.

Developing characters is never a linear process, and it always requires a lot of iteration to get the results we want. Of course, we follow the classic production steps, high-poly sculpting, low-poly modeling, texturing, and so on, but because our games often take shape over several years, and so many elements like animation and lighting come together late in production, the polish phase typically involves multiple passes. In the final six months of development, we revisited all characters and character-related assets several times. We also played through the entire game regularly to evaluate what still needed improvement.

One of the biggest challenges we faced was Mio and Zoe's costumes. The decision to give them unique outfits for every level wasn't part of the original pitch, it came much later in production. This added a significant amount of work: we had to design all the concept art for their costumes and figure out how to produce them within the project's timeline. As Character Art Director, it was essential to me that every outfit felt cohesive in both design and quality and that they represented the very best of what our team could achieve.

Another area we put extra focus on was the overall art direction and readability of Mio and Zoe. Since their costumes change in every level, it was crucial that each look still reflected their unique personalities and that players could easily recognize their character. To support this, we incorporated clear visual language into their designs, each character is represented by a geometric shape and a specific color palette to reinforce their identity.

If you look closely at Mio and Zoe's costumes, you'll notice that Mio often features a triangle motif, while Zoe is associated with a circle. Sometimes, these shapes are integrated quite literally, and other times, they're more subtle. We also maintained a consistent color scheme: Mio’s outfits typically include accents of red, blue, and magenta, while Zoe's feature touches of green, yellow, and turquoise. Striking the right balance and making every costume feel fresh and unique while keeping them visually consistent and readable was a major challenge and something we spent a lot of time refining.

That said, not every part of the process was equally challenging. Designing and producing the girls' initial real-world outfits was relatively straightforward, as they were closely tied to their personalities and provided a clear starting point for their visual identities. However, some costumes proved far more demanding. In particular, the sci-fi-themed outfits pushed us outside of our comfort zone. Working with complex hard-surface materials and shapes, rather than the more organic fabrics and forms we were used to, required us to rethink and adapt our workflows in ways we hadn't done before. It was a different kind of challenge, but one that ultimately helped us grow as a team.

To achieve the level of quality we were aiming for, we also worked closely with external partners. I want to take a moment to thank the talented artists who collaborated with us every day, taking our direction and feedback and helping us shape each costume to perfection. We couldn't have done it without them.

What have been some of the community's first reactions since the release?

Felix: Just shipping a finished game is an accomplishment I think any developer should be proud of. Just getting so many different moving pieces to somewhat work together to form a game is still a wonder to me. That said, the positive response has been humbling and overwhelming. You lose perspective looking at something for so long, somehow overlaying all the iterations on top of one another all the way back from janky prototype to finished product. In that way, having someone just seeing the end result and reacting to it – positive or negative – really gives you perspective. Though, of course, the positive feedback is more fun to hear, it's all interesting.

Jacob: The reception has been fantastic, and we've been both humbled and excited to see how players around the world have connected with the game, especially its visual storytelling, artistic direction, and technical execution. As the art team at Hazelight, we're incredibly happy to see the art resonate so strongly with players and to know that it's helped elevate the overall experience.

Bringing a game like Split Fiction to life, with so much variety, while rendering two fully realized experiences in real-time, really pushed us to our creative and technical limits. None of it would have been possible without the incredible talent across the entire studio, from artists and animators to designers and programmers. It was truly a collective effort.

Felix Strandberg, Senior Animator at Hazelight

Jacob Larsson, Character Art Director at Hazelight

Interview conducted by Amber Rutherford