Spotlight
May 5, 2026
Crafting stylized digital humans on Daughter of the Inner Stars with MetaHuman and UE5
Daughter of the Inner Stars is a 50-minute original symphonic orchestral experience that blends live narration, animation, and a cinematic score. Created by Nicholas Buc and Melanie Hiluta and written and directed by James Vinson, the show follows the journey of Fia, a courageous young girl searching for her lost father Burto in the crossfire of a futile cosmic war.
What makes this story particularly compelling is not just the final result—which is undoubtedly spellbinding—but how it was achieved.
With a small team and limited resources, the creators relied on MetaHuman and Unreal Engine to push beyond what they had previously thought possible. That opened the door to a new kind of workflow—one that combines stylized visuals, real-time production, and high-fidelity performance capture into a cohesive pipeline for modern storytelling.
What makes this story particularly compelling is not just the final result—which is undoubtedly spellbinding—but how it was achieved.
With a small team and limited resources, the creators relied on MetaHuman and Unreal Engine to push beyond what they had previously thought possible. That opened the door to a new kind of workflow—one that combines stylized visuals, real-time production, and high-fidelity performance capture into a cohesive pipeline for modern storytelling.
Blending photorealism with a painterly visual style
One of the most distinctive aspects of Daughter of the Inner Stars is its visual identity. Rather than leaning fully into photorealism, the team at Inferstudio—who were tasked with building the visual world of the experience—aimed for a more tactile and expressive aesthetic.
“Something that spoke more to the matte physicality of an illustrated children’s book and less to the conventions of the glossy screen-based media,” explains Nathan Su, Production Designer, Director of Photography, and Visual Effects Supervisor.
At the same time, Inferstudio had already decided that real-time workflows in Unreal Engine would be the only feasible approach for them, given their small team size and respective skillsets.
In the earliest tests for the show’s visuals, the team created painterly background environments and composited in 2.5D illustrated characters with limited animation. “We thought of the scenes as frozen portraits or immersive tableaus,” says Su.
When MetaHuman Animator was released, they tested translating the characters into full MetaHumans—just for fun—and did a basic facial mocap test.
“It was at this point that James, the director, got really excited about the potential to bring real human performance into the show, so we quickly pivoted from the idea of frozen tableaus to a more animated approach,” explains Su.
MetaHuman’s fidelity and accuracy were a huge step forward for the project, because they enabled the team to explore the nuance and subtleties of the characters complex emotions. However, this also created a tension. While MetaHuman’s strength lies in realism, that pushed the visuals in an unintended direction.
“The extremely high resolution of the characters meant our scenes started to feel like video game cutscenes,” says Su.
They needed to re-find a more painterly aesthetic, and embarked on a new round of experimentation. That led to the development of a hybrid pipeline combining Unreal Engine renders with AI-assisted stylization.
“We basically rendered everything out of Unreal at production quality, then fed every single final frame of the show into a custom AnimateDiff script we developed,” explains Su.
The Stable Diffusion AI pass was deliberately light, and didn’t fundamentally change the image, but instead removed minor details that registered as distinctly digital—such as aliasing, blurring artefacts, and texture seams—as well as adding some variation and softness to edges and intersections on areas like hair.
The key wasn’t to overhaul the image but to subtly reshape it. By layering outputs together, they introduced diffusion and painterly qualities without losing control over lighting and composition.
“Most of the time, this was a fairly simple opacity operation,” says Su. “We would blend about 30-50% of the Stable Diffusion image over the render.”
The result is a visual style that sits somewhere between illustration and cinematic realism, demonstrating how flexible real-time pipelines can be when combined with post-processing techniques.
Images courtesy of Inferstudio.
Customizing MetaHumans to match stylized character design
Achieving this stylized look required more than just post-processing. The team also modified MetaHuman assets directly, especially for the main character, Fia. “She had a custom hair groom that we created in Blender, and upscaled eyes, as well as some custom skin textures,” says Su.
For her eyes, the team wanted proportions that went beyond what MetaHuman Creator (MHC) could achieve. To nail the look, they first created a character as close to the desired outcome as they could in MHC, then exported the head mesh straight out of Unreal Engine for editing in Blender.
“It was just really the eye meshes themselves that needed to be scaled up, so it was a fairly straightforward modelling process,” notes Su.
Scaling up features like the eyes beyond standard limits ultimately enabled the team to push the design closer to a 2D illustration style. “We were honestly blown away by how achievable it all was—especially since in many cases, we were learning the tools as we were using them,” says Su.
For the other humanoids, the team used base MetaHuman faces and bodies, with customizations to hair, costumes, and faces that resulted in a more stylized and hand-painted aesthetic.
“The logic of the MetaHuman Blueprint made it really easy to just swap out textures and body meshes so that we could create characters that felt really original,” says Su.
While MetaHumans provided a flexible foundation for humanoid characters, the approach shifted significantly when tackling non-humanoid creatures like the Leviathan. This required entirely custom rigs and animation logic tailored to each creature’s anatomy.
Unlike MetaHumans, which benefit from standardized tools and workflows, these characters relied on more experimental techniques—from procedural rigs to spline-based animation—highlighting how pipelines must adapt when moving beyond human forms.
Images courtesy of Inferstudio.
Combining facial and body motion capture for realistic performance
Performance capture played a central role in bringing the humanoid characters to life. The team combined MetaHuman Animator for facial capture with Autodesk Flow Studio for body tracking, then merged the data by hand.
“Syncing the body data with the facial capture data was a manual process we had to go through for every animation, but once they synced, magic happened,” explains Nour Hassoun, Founder and CEO of IMPERSONAS—a digital human studio that worked on the project.
Because they were using a single camera-based capture system for both face and body, the team could have multiple performers interacting with each other at the same time.
“If we’d needed to have multiple suits or more high-end tracking systems, we certainly wouldn’t have been able to afford to do motion capture on this project, and the ability to capture every performer in a scene simultaneously was essential,” says Su.
The team also leveraged the power of Live Link Face to achieve performance capture on a budget, using a phone for facial capture.
“We put together a really low-tech facial capture rig using a bike helmet, a selfie stick, ankle weights, and a bit too much duct tape—and it surprisingly worked quite reliably,” says Su.
Images courtesy of Inferstudio.
“Performance capture was something I never would have dreamed a team of our size and skillset could tackle, even only a year before we took on this project.”
When it came to retargeting the captured actor performances onto their stylized characters, it was a case of trial and error. Too much eye movement would make Fia’s eyes clip, while certain body movements would result in her arms passing through her sleeves.
“There was nothing we couldn’t solve with texture masks, camera tricks, and careful animation cleanup,” says Hassoun.
The team found that a key differentiator between working with realistic and stylized characters is that the mocap cleanup isn’t just “cleanup”: there’s a whole process of creative additive animation on top, either to offset some of the quirks, or bring some more life into the exaggerated proportions.
“The workflow, despite these quirks, was still a lifesaver in comparison to traditional animation pipelines,” says Hassoun. “Especially since no one on the team is a character keyframe animator in the classical sense.”
All that animation clean-up took place directly in Unreal Engine. This was possible thanks to the high quality of the performance data the team got from MetaHuman Animator.
“The level of animation controls, the tools, the constraints, the switches…all of that on top of being able to render in real time—it’s just great,” says Hassoun.
“Especially when cleaning up facial capture with so many little subtleties. It helps a lot seeing the rendered result on screen immediately to adjust it, test it, and iterate on it. We’re a very small team—the luxury of exporting and importing data between tools is time we need to spend improving the film.”
When it came to retargeting the captured actor performances onto their stylized characters, it was a case of trial and error. Too much eye movement would make Fia’s eyes clip, while certain body movements would result in her arms passing through her sleeves.
“There was nothing we couldn’t solve with texture masks, camera tricks, and careful animation cleanup,” says Hassoun.
The team found that a key differentiator between working with realistic and stylized characters is that the mocap cleanup isn’t just “cleanup”: there’s a whole process of creative additive animation on top, either to offset some of the quirks, or bring some more life into the exaggerated proportions.
“The workflow, despite these quirks, was still a lifesaver in comparison to traditional animation pipelines,” says Hassoun. “Especially since no one on the team is a character keyframe animator in the classical sense.”
All that animation clean-up took place directly in Unreal Engine. This was possible thanks to the high quality of the performance data the team got from MetaHuman Animator.
“The level of animation controls, the tools, the constraints, the switches…all of that on top of being able to render in real time—it’s just great,” says Hassoun.
“Especially when cleaning up facial capture with so many little subtleties. It helps a lot seeing the rendered result on screen immediately to adjust it, test it, and iterate on it. We’re a very small team—the luxury of exporting and importing data between tools is time we need to spend improving the film.”
Images courtesy of Inferstudio.
Unreal Engine for real-time cinematic production
Unreal Engine served as the backbone of the entire production. Its real-time capabilities fundamentally changed what was possible for the team.
“A piece of this length and complexity would not have been possible for us using traditional PBR rendering workflows,” states Su. “The render times would have crippled us, let alone the project’s ambition to create sweeping long takes across alien landscapes and worlds dense with life and geometry.”
They realized that Unreal Engine’s core rendering technologies would enable them to do more with a smaller team.
“When we took on Daughter of the Inner Stars we assumed from fairly early on that we would be building and rendering it in Unreal Engine because of Lumen and Nanite,” says Su. “Access to Quixel Megascans assets and materials was also something we relied upon heavily.”
Beyond rendering, Unreal Engine enabled real-time collaboration. “We could collaborate with the Composer Nick and Director James via online meetings where we could make notes on environments or camera movements in real time via screenshare,” explains Su. “The real-time nature of the scenes we were building made it much easier for James to see and finesse the choreography within each shot.”
This enabled creative decisions to happen faster and more collaboratively than in traditional pipelines.
Images courtesy of Inferstudio.
Empowering small teams to tell big stories
Daughter of the Inner Stars is testament to how the combination of real-time rendering, accessible performance capture, and flexible character tools can empower small studios to create large-scale cinematic experiences.
But more than this, it demonstrates a broader shift in creative technology.
Tools like MetaHuman and Unreal Engine are not just improving workflows—they’re redefining what’s possible for independent creators. “It is no exaggeration to say that the release of MetaHuman Animator totally changed our ambitions for and approach to this project,” concludes Su.
Images courtesy of Inferstudio.