Clair Obscur Expedition 33: development, design, and what the project reveals about modern JRPG production

Clair Obscur Expedition 33 as a game development case study

Sandfall Interactive’s Clair Obscur Expedition 33 arrived in 2025 as one of the most closely watched role-playing launches of the year, and the reasons it matters extend well beyond its combat system or its soundtrack. The project is a useful case study in how a small French studio used Unreal Engine 5, a deliberately old-school turn-based combat model, and a tightly scoped pre-production plan to ship a roughly thirty-hour single-player RPG in a genre that most publishers have written off as commercially risky. For developers, technical artists, designers, and producers, the title is less interesting as a single game than as a working example of a particular production strategy, and that is the angle this article is built around.

This page is not a review, a tier list, or a walkthrough. It is an editorial breakdown of how Clair Obscur Expedition 33 was made, how the team structured its combat and content pipelines, where the project clearly took inspiration from established JRPGs, and where it diverges in ways that have practical consequences for anyone trying to build or commission a similar title. If you are evaluating the game as a player, the second half of the article also explains the systems you will actually interact with, but the framing throughout is production-first.

Studio background and the production context behind Clair Obscur Expedition 33

Clair Obscur Expedition 33 was developed by Sandfall Interactive, a studio based in Montpellier, France. The team was founded by Guillaume Broche and several former Ubisoft Montpellier staff, which matters for the article because the project inherits a production culture shaped by large-studio console work, not by mobile or live-service pipelines. The studio has been transparent about its headcount during the marketing cycle, and the figures that have been reported by the studio and by outlets covering the launch describe a team in the rough range of around thirty to forty people at peak, which is small for a project of this visual and audio ambition.

For a game development audience, three structural facts about the studio are worth keeping in mind before reading the rest of the article:

• The team is geographically concentrated in a single city, which simplifies review, mocap scheduling, and audio direction, but concentrates key-person risk.
• The studio is independent and self-funded for the bulk of pre-production, with Kepler Interactive handling publishing, which changes the milestone structure compared to a first-party deal.
• The project is built on Unreal Engine 5, so most of the rendering, lighting, and animation tooling discussion later in the article is engine-specific rather than engine-agnostic.

These three constraints shape almost every decision the rest of the article describes, from the combat grid to the way cinematics are staged.

Why this title matters for modern RPG production

The wider context is that the turn-based JRPG market has been treated as a niche for the better part of two decades, with a few major exceptions in any given console generation. The release of Clair Obscur Expedition 33 sits inside a small but real wave of Western-developed, Japanese-influenced RPGs that includes recent entries from smaller publishers and a handful of high-profile indie projects. What makes the Sandfall project noteworthy from a development standpoint is not the genre choice itself but the cost structure around it. The studio has publicly discussed a budget that is orders of magnitude below the cost of comparable open-world action RPGs, while still shipping a title with full performance capture, orchestral scoring, and a thirty-plus-hour main campaign.

For a production lead or studio director, the useful question is not “is turn-based back” but “what does this project tell us about building a content-heavy single-player RPG at a mid-budget scale on Unreal Engine 5?” The rest of the article works through that question section by section.

Pre-production, scope, and the decision to stay single-player

One of the more interesting production decisions behind Clair Obscur Expedition 33 is what the team chose not to do. There is no multiplayer mode, no live-service layer, no seasonal content roadmap, and no microtransaction store. The game is a finite, self-contained single-player experience with credits and a clear endpoint. From a production standpoint, this is not a stylistic choice but a scope choice, and it has direct consequences for headcount, content planning, and post-launch support.

Production decision	What the team did	What it avoided	Why it matters for similar projects
Game shape	Finite single-player RPG with defined ending	Live-service, seasonal, or multiplayer layers	Reduces long-tail content cost and ops headcount
Combat system	Turn-based with timing-based and free-aim layers	Real-time action combat and netcode	Removes netcode, server, and matchmaking scope
Engine choice	Unreal Engine 5 with Lumen and Nanite where useful	Custom engine or Unity-specific workflows	Leverages existing tooling and contractor pool
Open world	Hub-and-section structure instead of a seamless open map	Large streaming open world	Cuts streaming, LOD, and navigation work significantly
Capture pipeline	Targeted performance capture for key scenes	Full cinematic mocap for every scene	Compresses animation budget without losing key moments

For readers building a pitch around a similar project, the practical takeaway is that Clair Obscur Expedition 33 uses every one of these decisions to keep the total scope inside what a sub-forty-person team can deliver. None of them are unique in isolation, but their combination is what made the project financeable.

Unreal Engine 5 as a technical foundation

Clair Obscur Expedition 33 runs on Unreal Engine 5, and the studio has been open about which features the project leans on. Lumen is used for global illumination in many of the interior environments, Nanite is applied to high-density meshes where it makes sense, and the world is built using a combination of World Partition and traditional level streaming for the more cinematic sections. For a technical artist reading the article, the interesting part is not that the team used these features but how selectively they used them.

Several observations from the studio’s published talks and post-launch interviews are useful as patterns rather than as prescriptions:

• Large outdoor vistas rely on baked lighting and traditional level streaming rather than full real-time global illumination, because the cost-per-frame of Lumen at long view distances was not justified by the visual gain.
• Nanite is concentrated on architecture and large set pieces, while foliage and characters use standard mesh pipelines with carefully tuned LODs, which keeps the cost of foliage density manageable.
• World Partition is used for exploration zones, while hub areas are loaded as discrete maps so that art direction can control lighting transitions more strictly.
• Post-processing, including the painterly color grading the game is known for, is handled through a layered material and post-process volume system rather than custom engine code, which makes iteration faster for the art team.

The general principle worth carrying into a similar project is that the engine’s headline features are most useful when they are applied surgically to the scenes where they are visible, rather than turned on globally. Clair Obscur Expedition 33 reads visually as a high-end Unreal project, but the cost profile is closer to a careful mid-budget build than to an open-world epic.

Art direction and the painterly Unreal pipeline

The visual identity of the game is built on a Belle Époque and art-deco-inspired palette, with desaturated environments and saturated character and effect work layered on top. From a production standpoint, this is a smart decision because it compresses a lot of art direction into a small number of rules: a constrained hue range for environments, a higher saturation budget for characters and VFX, and a clear separation between readable gameplay silhouettes and decorative background detail.

For technical artists and art leads, the practical pipeline looks roughly like this:

1. Environment artists build meshes at a high level of geometric detail, with Nanite enabled on the majority of architectural assets, knowing that the painterly material pass will hide fine surface noise.
2. Materials are built in layers, with a base color and roughness layer driven by vertex color so the art team can repaint surfaces in-engine without rebuilding assets.
3. VFX and character shaders sit on top of the environment grade with a separate, brighter pass, so readability in combat is preserved even in low-light or fog-heavy zones.
4. Post-process volumes are authored per zone, with a global “look” applied at the project level and per-area overrides for storytelling beats.

This kind of layered approach is not unique to Clair Obscur Expedition 33, but the project is a clean example of how a small team can make a constrained palette feel intentional rather than limited. The constraint is doing work, not just decorating the game.

Combat system design: turn-based with real-time layers

The combat system is the part of Clair Obscur Expedition 33 that has attracted the most design discussion, and it is worth treating carefully because it is also the part that is most often oversimplified. The system is turn-based at its core, but it is not a strict “menu vs. menu” JRPG in the classical sense. The team has layered two real-time elements on top of the turn structure: a timing-based hit input that rewards well-timed attacks, and a free-aim shooting layer for certain character abilities that turns a subset of actions into something closer to a third-person aim-and-fire problem.

For a designer reading this, the underlying logic is that the turn structure gives combat a clear decision rhythm and keeps encounter design tractable for a small team, while the real-time layers add a skill ceiling that makes repeated encounters feel less repetitive. From a production standpoint, this is a deliberate trade-off:

Combat layer	Player input model	Engineering cost	Design benefit	Risk
Core turn loop	Menu selection	Low	Encounter design is approachable	Can feel static if not paced well
Timing-based hit	Real-time button press window	Medium	Raises skill ceiling on standard actions	Accessibility concerns on reaction-time gating
Free-aim abilities	Real-time aim and trigger	High	Adds variety to a subset of actions	Camera, control, and aim-assist tuning cost
Status and break system	Menu selection with reactive feedback	Medium	Enables build variety	Balance drift if shipped without telemetry

The risk column in the table is the part that production readers should pay attention to. Each of these layers adds work, and the higher-cost layers are the ones that interact with hardware, accessibility, and platform certification. The studio’s ability to ship a clean launch depends in part on how cleanly those layers were scoped, not just on how clever the design is.

Encounter and boss design for a small team

Encounter design in Clair Obscur Expedition 33 follows a pattern that should be familiar to anyone who has shipped a turn-based RPG: trash encounters are tuned to be solvable in a small number of rounds with basic resources, minibosses introduce one new mechanic, and bosses combine several previously introduced mechanics under heavier resource pressure. The interesting part is how the real-time layers are introduced, because they are gated behind specific encounters rather than available from the start.

For a designer or producer, the practical checklist for shipping a similar title in this style looks like this:

• Define a small number of combat verbs, then build every encounter as a combination of those verbs rather than inventing new abilities per fight.
• Introduce the real-time layers one at a time, with encounters that explicitly teach the input window before any encounter that requires it.
• Tune bosses around a target round count, not a target damage-per-second, so that balance is expressed in player experience terms rather than only in numbers.
• Keep status effect and break interactions in a small, well-documented table so the QA team can validate edge cases without consulting design every time.
• Plan post-launch balance passes assuming the encounter table will be revisited, especially for any boss that tests show players are retrying more than two or three times on average.

None of this is novel advice, but Clair Obscur Expedition 33 is a good example of a team that appears to have followed it under time pressure.

Animation, cinematics, and the performance capture pipeline

Animation in Clair Obscur Expedition 33 mixes hand-keyed combat animation with targeted performance capture for cinematics. The split is deliberate. Combat animation needs to be tightly tied to gameplay timing and hit reactions, which favors hand-keyed work in Maya or Blender with the engine’s animation tools. Cinematics need to carry the emotional weight of the story, which favors real actors, real facial capture, and real voice direction.

The capture pipeline, as the studio has described it in public talks, is roughly as follows:

1. Writers and the narrative lead lock the scene’s beats and dialogue, including the emotional intent of each line.
2. Voice actors record a reference pass that the audio team uses as a timing target for the final capture.
3. On capture day, the team records body, face, and voice simultaneously, using a head-mounted camera rig for facial performance and a standard stage for body work.
4. The animation team cleans up the captured data, rebuilds any interaction with props that the capture rig cannot resolve, and re-targets the result into the engine’s skeleton.
5. Lighting and camera are blocked in-engine on top of the cleaned animation, so the cinematic team can iterate on framing without going back to the capture stage.

This is a fairly conventional capture pipeline for a mid-budget Unreal project. The reason it is worth describing is that the team has been explicit that not every scene receives full capture. Dialogue-heavy or emotionally central scenes get the full treatment, while transitional scenes are assembled from in-engine camera work and pre-existing animation libraries. That triage is what makes the budget work for a small team.

Audio direction and the orchestral approach

The music for Clair Obscur Expedition 33 is a hybrid of orchestral recording and in-engine adaptive music, with the score recorded live rather than synthesized sample by sample. Lorien Testard is credited as composer, and the production approach has been discussed in interviews as a combination of studio sessions and a tightly integrated adaptive system inside the engine.

For an audio lead reading this, the pattern is worth noting because it is the one most often mishandled by smaller studios:

• Record the score to picture for the major story beats, so the emotional intent of each cue is locked at the recording stage rather than at the mix stage.
• Build the adaptive music system around a small number of stems per track, so the engine can crossfade between states without producing audible gaps.
• Keep the combat mix on a separate bus, with a limiter and a clearly defined headroom budget, so the mix engineer can tune combat loudness without touching the explorational music.
• Run the final mix in the engine, not in the DAW, because the only way to know how the adaptive layers actually sound is to hear them in the same environment the player will.

The team has also been clear that the recorded sessions were a significant line item, which is the part that often gets cut on smaller projects. Clair Obscur Expedition 33 is a useful counter-example to the idea that an RPG score has to be either fully sample-based or fully linear.

Narrative structure, localization, and writing pipeline

The writing in Clair Obscur Expedition 33 is structured around a small main cast and a clearly bounded set of locations, which is one of the reasons the studio was able to keep the writing team small while still shipping a thirty-plus-hour campaign. From a production standpoint, the narrative pipeline looks like a classic “lock the spine, then fill the branches” model.

The practical workflow, as described in interviews and inferred from the shipped game, follows these steps:

1. The narrative lead and the game director lock the main story spine, including the ending, before any side content is written.
2. Side content is written against a template that specifies the location, the cast involved, the emotional beat, and the mechanical reward.
3. Dialogue is recorded in batches, with the voice director working from a written transcript and the reference recording described in the previous section.
4. Localization is planned at the writing stage, not at the QA stage, which means the writers are aware of subtitle length and line count from the first draft.
5. Final review is done in-engine with audio, so the team can catch timing issues that are invisible on the page.

For a producer or narrative lead building a similar project, the takeaway is that the structure of the writing pipeline matters more than the size of the writing team, and that Clair Obscur Expedition 33 is a working example of a small team keeping a long-form narrative on schedule without sacrificing centralization.

QA, certification, and the platform reality

Clair Obscur Expedition 33 shipped on PC and on both major console families, which is a significant certification surface for a small team. Console certification is one of the areas where the project reveals its production discipline most clearly, because the cost of a failed submission is high in both calendar time and engineering rework.

The areas the team has indicated it focused on for certification and platform readiness include:

• Suspend and resume behavior, which is a frequent source of console certification issues, especially for titles that stream levels in and out.
• Achievement and trophy unlock conditions, which need to be testable and which the team had to validate against each platform’s rules.
• Save data integrity, including what happens when a player suspends the game mid-cutscene, which is a real failure mode for cinematic-heavy titles.
• Accessibility settings, which the platform holders increasingly expect to be present and which interact with combat timing in this title specifically.

For a producer, the useful part of the article is not the specific checklist above but the fact that the team treated these as production risks from the start, not as certification bugs to be fixed at the end. That distinction is the difference between a clean submission and a delayed one.

Accessibility, settings, and the timing-based combat layer

Accessibility is one of the more interesting tension points in Clair Obscur Expedition 33, because the real-time timing layer in combat interacts directly with reaction-time accessibility. The team has shipped accessibility options, but it has also been honest that the timing layer is a design choice rather than a bug, and that the options exist to give players more control over how that layer feels, not to remove it entirely.

For a designer reading this, the practical pattern is to separate two kinds of accessibility work:

Accessibility concern	Default game design	What the options should change	What the options should not change
Reaction-time gating on hits	Tight input window on standard attacks	Window length, visual cue size	Damage numbers or turn order
Camera sensitivity in aim layer	Standard third-person aim	Sensitivity, aim assist strength	Enemy behavior or projectile speed
Subtitle legibility	Default subtitle style	Size, background, color, dyslexia-friendly options	Dialogue content
Input remapping	Default controller and keyboard layouts	Remappable buttons, hold-vs-toggle for crouch and similar	Combat system itself

The point of separating the two columns is that accessibility is most credible when it changes the input the player has to make, not the game the developer has designed. Clair Obscur Expedition 33 is a reasonable working example of that distinction.

Player-side system guide for a new save

Switching briefly to the player’s perspective, the most useful way to onboard into Clair Obscur Expedition 33 is to treat the first hour as a tutorial that the game does not always label as one. The combat verbs, the status effects, and the timing windows are all introduced gradually, and trying to optimize a build before the second zone is a common waste of time because the resource economy changes significantly once the party expands.

A practical first-save checklist for a player new to the game looks like this:

1. Learn the timing input on the basic attack as soon as it is introduced, because every later combat system builds on it.
2. Resist the urge to spend attribute points before the second or third zone, because early choices can be respecced but the cost of respeccing grows with the number of points spent.
3. Treat the free-aim abilities as a secondary tool, not a primary one, until the party has a character whose kit explicitly supports them.
4. Save before any boss attempt, because the game’s save system is generous but not unlimited, and a single failed run should never cost an hour of progress.
5. Explore optional zones in the second act before finishing the third, because the difficulty curve assumes a specific level of optional completion.

This list is not a substitute for a full player guide, but it is the part of the system that the production design is actually built around, and understanding it makes the rest of the game more legible.

How the project compares to other mid-budget RPGs

For readers who have shipped or commissioned mid-budget RPGs, the most useful comparison set for Clair Obscur Expedition 33 is not the major JRPG franchises but a small group of recent Western-developed, Japanese-influenced RPGs at a similar budget tier. The table below is an editorial comparison, not a benchmark.

Project	Approximate team size	Engine	Combat model	World structure	Single-player or live
Clair Obscur Expedition 33	~30 to 40 at peak	Unreal Engine 5	Turn-based with real-time layers	Hub and sectioned zones	Single-player, finite
Recent mid-budget action RPG A	~80 to 120	Custom or in-house	Real-time action	Open zones with hub	Single-player with planned DLC
Recent mid-budget RPG B	~20 to 30	Unity	Turn-based	Linear with branches	Single-player
Recent AA JRPG-styled title	~40 to 60	Unreal Engine 4 or 5	Turn-based	Hub with semi-open zones	Single-player, often with DLC

The numbers above are editorial estimates based on the size of the public teams and the visual scope of the shipped projects, and they should be read as rough order-of-magnitude comparisons rather than precise figures. The point of the table is that Clair Obscur Expedition 33 sits at the upper end of the small-team tier in visual ambition, which is the part of the project that is genuinely interesting to learn from.

Risks, open questions, and what a post-launch follow-up could look like

No project of this kind ships without trade-offs, and the honest production read of Clair Obscur Expedition 33 has to acknowledge the open questions as well as the strengths. The areas that will most likely determine the long-term reputation of the project are the ones a careful reader should keep an eye on.

• Balance pass frequency: how quickly the team addresses the encounter or boss that the community identifies as the most unfair or the most over-tuned, and whether the post-launch patches stay within the scope of a small team.
• Platform parity: whether the console versions hold up to the PC version in frame pacing and load times over the lifetime of the product, especially as the team adds patches.
• Accessibility follow-through: whether the team continues to add accessibility options, or whether the launch set is treated as the final word on the subject.
• Documentation and tooling investment: whether the project produces enough internal documentation to support a future title, which is the part that determines whether the studio’s learnings compound.
• Long-tail content: whether the project expands into DLC or a follow-up, and if so, whether that expansion is a small, focused addition or a scope creep that strains the same team that shipped the base game.

These are the questions a producer or studio director would ask in a postmortem, and they are also the questions a player who is considering a long-term investment in the studio’s catalog would want answered over the next year.

Practical takeaways for developers and producers

Pulling the production analysis together, the patterns from Clair Obscur Expedition 33 that are most worth carrying into a similar project are these:

1. Scope is the single largest determinant of whether a mid-budget RPG ships on time, and the team’s willingness to say no to live-service, multiplayer, and open-world scope is what made this project financeable.
2. Engine feature use should be selective. Lumen, Nanite, and World Partition are most cost-effective when they are turned on for the scenes where they are visible and turned off everywhere else.
3. Combat systems benefit from a small number of clearly defined verbs, with real-time layers introduced one at a time rather than all at once.
4. Performance capture is a triage problem, not a coverage problem. Capture only the scenes that need the emotional lift, and build the rest in-engine.
5. Accessibility is a design problem, not a checklist. Separate what the options change from what the game design is, and be honest about the difference.

For a studio considering a similar project, the article’s main argument is that Clair Obscur Expedition 33 is a useful proof point for a particular kind of mid-budget RPG production. It is not a template, and the team is explicit that several of its choices were specific to this project. But the shape of the production is repeatable, and the rest of the industry is likely to be studying it for at least the next year.

Frequently asked questions

What engine does Clair Obscur Expedition 33 use, and how heavily does it lean on Unreal Engine 5 features?

The game runs on Unreal Engine 5, and the studio uses Lumen for global illumination in interior environments, Nanite for high-density architecture, and World Partition for the exploration zones. The team has been clear that these features are applied selectively rather than globally, with baked lighting and traditional streaming used in the outdoor vistas where the cost of full real-time global illumination is not justified by the visual gain.

How large is the team that made Clair Obscur Expedition 33?

Sandfall Interactive is a Montpellier-based studio that has described its headcount during the marketing cycle as being roughly in the range of around thirty to forty people at peak. That is small for a project of the visual and audio scope the game ships with, and it is one of the main reasons the production decisions described in this article lean toward scope control rather than feature breadth.

Is Clair Obscur Expedition 33 a live-service or single-player game?

It is a finite single-player RPG with a clear ending, no multiplayer mode, no seasonal content, and no microtransaction store. From a production standpoint, that choice removes a large amount of post-launch scope and is the single biggest reason the project was financeable for a small independent studio.

How does the combat system work, and is it purely turn-based?

The combat is turn-based at its core, but it is layered with two real-time elements. The first is a timing-based hit input that rewards well-timed attacks, and the second is a free-aim shooting layer for certain character abilities. The turn structure gives the design a clear rhythm and keeps encounter design tractable, while the real-time layers add a skill ceiling that keeps repeated encounters from feeling repetitive.

How long is the main campaign in Clair Obscur Expedition 33?

Main campaign length depends heavily on how much optional content a player chooses to complete, but a player focusing primarily on the main path will spend on the order of twenty-five to thirty-five hours with the story, and substantially more if they complete the optional zones and post-game content.

Does Clair Obscur Expedition 33 have accessibility options?

Yes. The game ships with accessibility settings including subtitle customization, input remapping, and options that affect the timing-based combat input. The team has been clear that the timing layer is a design choice rather than a bug, so the options are framed as giving players more control over how that layer feels rather than removing it entirely.

Is there a New Game Plus or post-game content in Clair Obscur Expedition 33?

The game includes post-game content in the form of higher-difficulty encounters and optional rewards, and the structure supports a second playthrough with carried-over progression. The team has not announced a large-scale DLC roadmap, which is consistent with the project’s scope as a finite single-player experience.

Who is the publisher behind Clair Obscur Expedition 33?

Kepler Interactive handled publishing for the title, while Sandfall Interactive retained development ownership. The publisher-developer split matters from a production standpoint because it changes the milestone structure and the post-launch support model compared to a first-party deal or a fully self-published release.

What should a developer take away from Clair Obscur Expedition 33?

The most useful patterns are scope control, selective use of engine features, layered combat design with a small number of clearly defined verbs, triage-based performance capture, and accessibility treated as a design problem rather than a checklist. None of these are unique to the project, but their combination is what made a thirty- to forty-person team able to ship a content-heavy single-player RPG on Unreal Engine 5 in a genre most publishers treat as risky.

What should a player know before starting Clair Obscur Expedition 33?

Learn the timing input on the basic attack as soon as it is introduced, avoid spending attribute points heavily before the second or third zone, treat the free-aim abilities as a secondary tool at first, save before any boss attempt, and explore the optional zones in the second act before finishing the third. The game is more legible to a player who treats the first hour as a tutorial the game does not always label as one.

Related reading

• Jujutsu Odyssey Codes: The Daily Grind on the Tokyo Streets
• Project XL Codes: The Pain of Grinding Low-Tier Mobs