DEV Community

dev game
dev game

Posted on

Why Clair Obscur: Expedition 33 Bet on Unreal Engine 5: The Borrowed Blockbuster Tech Behind a 32-Person Turn-Based Miracle

Clair Obscur: Expedition 33 should not exist on paper. A 32-person debut team from Montpellier built a Belle Epoque turn-based RPG that beat every AAA finalist to Game of the Year at The Game Awards 2025 and sold over 8 million copies in its first year. The short answer to why it works is boring: Sandfall Interactive built it on Unreal Engine 5, after starting the prototype on Unreal Engine 4 and switching mid-production for rendering improvements.

That choice is the whole story. A team with no shipped title, no custom tech, and no 300-person art department rented Epic's blockbuster renderer and used it against type — not for a shooter, but for menus, timing windows, and oil-paint ruins.

The facts: who made what, on what, and when

Expedition 33 was developed by Sandfall Interactive and published by Kepler Interactive, releasing on PlayStation 5, Windows and Xbox Series X/S on 24 April 2025, day one on Xbox Game Pass.

The studio facts matter because they explain the engine:

Unreal Engine 5 itself launched 5 April 2022, is written in C++, and is source-available with a 5% royalty over $1 million. Its headline features are Nanite virtualized geometry for film-quality meshes without manual LODs, and Lumen dynamic global illumination and reflections. UE was originally a 1998 FPS engine that later spread to all genres through portability across desktop, console and mobile.

In other words: a Ubisoft exile learned UE as a hobby, hired 30 people, proved one slice, and got a publisher to pay for a UE5 JRPG that AAA publishers had declared dead.

Why UE5 and not Unity, custom tech, or staying on UE4?

Sandfall never seriously had another option, and that is a strength.

First, hiring. In Montpellier in 2021-2022, you cannot hire 12 Unreal generalists if you pick a niche stack. Over 200 interviews were conducted to assemble the first dozen artists and programmers. UE meant the tutorial, the marketplace, Quixel Megascans, and every junior who had touched Blueprints or C++ could be productive in week one. A custom C++ engine like the one behind Animal Well would have required senior engine programmers Sandfall could not afford.

Second, the prototype path. Starting on UE4 then moving to UE5 gave them continuity. Development began on UE4 and later switched to UE5 for rendering improvements. That is not trivia — it means gameplay code, turn logic, and tools survived while lighting and geometry got a free upgrade. If they had stayed on UE4 they would have hand-built LODs for every floating island. With Nanite automatically managing LODs by scaling meshes dynamically, a 5-person environment team could ship vistas that read as 200-person work.

Third, publisher risk. Kepler funded the game after the vertical slice. A publisher can evaluate a UE5 slice: it runs on PS5 devkits, it has a known certification path, it ports to Xbox Series X/S with the same codebase. A custom engine slice proves art but not shippability.

What UE5 actually bought: paint, not polygons

Most coverage says Expedition 33 looks AAA because of UE5. That is half true. My read — opinion, not Epic documentation — is that Sandfall used UE5 as a costume department, not a polygon pump.

  1. Lumen as Belle Epoque filter. Lumen provides dynamic global illumination without baking lightmaps. For a team without a lighting farm, that means dusk in the Flying Waters or candlelight in the Manor just works when you move a light. The game's signature look — chalk-white faces, saturated red trees, foggy stone — is largely color grading plus bounce light. UE5 let three lighting artists iterate like painters instead of waiting overnight for bakes.

  2. Nanite as anti-crunch. Nanite partitions meshes into 128-triangle meshlets and scales detail by draw distance and resolution. For crumbling Parisian rooftops and gestral villages, Sandfall could import high-detail sculpts and ZBrush pieces directly. My interpretation: this is how 30 people fake scale. You do not model low-poly background cities; you scatter scanned stone and let Nanite cull it.

  3. Sequencer and cinematics as turn-based glue. Turn-based games live or die on camera punch-ins, not frame rate. UE's cinematic tools — built for Fortnite live events and Matrix demos — give free impact: skill cutaways, gradient attacks, camp conversations. Sandfall's combat is third-person turn-based where you aim ranged attacks like a third-person shooter, complete QTEs for bonus damage, and dodge, parry or jump in real time on enemy turns. That hybrid only feels good if the camera snaps fast and the timing window is readable. UE already solved that for shooters.

What it cost: the tax everyone under-reports

UE5 is not free, even when the license is.

Performance is the obvious bill. Expedition 33 launched with stutter complaints on PC, especially in dense zones — the classic Nanite + Lumen + virtual shadow maps trade. A custom turn-based engine could lock 60 fps on a toaster because nothing moves most of the time. Sandfall paid shooter-engine overhead for a game where characters stand still thinking.

The second cost is platform exclusion. PS5, Windows and Xbox Series X/S only — no Switch, no PS4. UE5 supports Switch and last-gen, but Nanite and Lumen do not scale down gracefully. Cutting 130 million PS4s and the Switch install base is a real sales sacrifice for an indie. Sandfall bet that Game Pass day-one discovery would outweigh it. With 8 million sold, they were right, but it was a gamble.

The third cost is sameness risk. UE5 games share a specular sheen. Sandfall fights it with art direction, not tech — mime-white skin, theatrical staging, accordion score. Without that, Lumen limestone looks like every other UE5 ruin.

My contrarian take: turn-based is UE5 camouflage

Here is where I diverge from most engine postmortems, and I want to label this clearly as analysis, not reported fact.

Most people say Sandfall overcame UE5's shooter bias. I think they exploited it.

A 32-person team cannot win a real-time action war on UE5. Combat designers would drown in animation trees, netcode, and 60-fps balancing. By choosing turn-based with real-time dodge/parry inputs, they turned UE5's weakness — heavy, cinematic, slightly late frames — into tension. When you have 800ms to parry, a 50ms hitch feels dramatic, not broken. Souls-like timing masks engine lateness the way film grain masks low resolution.

Similarly, the 210 Pictos items, with three equipped at a time and mastery after four wins are, in my view, a staffing hack. A tiny balance team cannot hand-tune six characters across 40 hours. So they shipped a compiler: let players combine Pictos into broken builds, watch YouTube find the meta, then patch the outliers. It outsources QA to the community. Games with small teams that ship fixed classes die on one bad tuning patch; games that ship combinatorics live forever on Reddit builds.

And the Gommage — the yearly erasure that kills everyone above a decreasing age — is, to me, a loading screen in narrative drag. It justifies a linear chain of gorgeous, disposable zones instead of a true open world that would expose UE5 streaming and memory limits on a small team. You never notice you cannot go back, because lore says the world is disappearing anyway. That is elegant constraint design.

In short: boring engine + unfashionable genre = moat. Unity could have done the same cheaper, but Unity does not give you Lumen out of the box, and Lumen is what makes screenshots sell a turn-based game to shooter-trained scrollers.

What indie devs can steal this week

  • Prototype combat in Blueprints, ship it in C++. Broche's hobby UE prototype became a funded slice because designers could touch it. Do not start with engine architecture; start with one parry window that feels good.
  • Use one vertical slice to buy trust. Sandfall's GDC March 2022 slice unlocked Kepler funding. One polished 10-minute dungeon beats 20 greyboxed systems.
  • Let lighting do staffing. If you have under five artists, pick the engine with dynamic GI. Baked lighting requires a lightmapper you do not have.
  • Ship combinatorics, not balance. 210 small modifiers beat 20 perfect skills. Players forgive an overpowered combo; they do not forgive three identical mages.
  • Pick a dead genre on a live engine. AAA abandoned turn-based AAA. That left YouTube, TikTok build guides, and Game Pass curiosity with nowhere to go until Expedition 33 arrived.

Train the Expedition reflexes in your browser

You cannot practice Expedition 33's parry timing without owning it, but you can train the underlying muscles — reading turn order, committing to builds, and staying calm when the enemy animation lies.

If Pictos teach anything, it is that squad composition beats reflexes. AutoRPG Arena is a turn-based auto-battler RPG where every item choice influences the fight and no two encounters play the same. Use it as a five-minute lab for the Expedition mindset: draft, lose, re-draft, and notice how one swapped item changes the whole tempo — just like swapping a Pictos set before a boss.

If you love the theorycraft side — mixing traits to craft an unkillable hero — try AFK Dungeon: Idle Action RPG, where progress comes from smart combinations of items, skills and traits. It captures the quiet half of Expedition 33: the camp menu where runs are actually won.

And if your parries keep coming out late, the problem is usually commitment, not speed. Battle Chess forces you to think ahead because the enemy mirrors your moves. Play a few rounds focusing only on not rushing: pick a line, assume punishment, and execute. That patience transfers directly to Expedition 33's dodge-versus-parry decision, where greed for Ability Points gets parties wiped.

None of these replace the real game's timing, but together they isolate the three skills Sandfall tests: plan the build, respect the enemy tell, and press once with confidence.

Bottom line

Expedition 33 did not succeed despite UE5. It succeeded because a 32-person team understood what UE5 is for in 2025: not code, but hiring, lighting, and publisher confidence. They paid the performance and platform tax, hid the engine's heaviness inside turn stops and parry pauses, and let players do the balancing work through Pictos.

That is a repeatable playbook — if you are willing to be unfashionable about genre while being utterly conventional about tech.

Top comments (0)