Mask of the Rose

Failbetter Games has spent over a decade cultivating one of the strangest fictional properties in games. Fallen London — the browser-based story engine, then Sunless Sea, then Sunless Skies — established a house style: Victorian gothic filtered through absurdist comedy, prose that punches above the medium’s usual weight class, and a matter-of-fact queerness that never announces itself as a feature. Mask of the Rose is the studio’s first attempt at a proper visual novel, and it’s the closest they’ve come to making a game where the writing is the mechanic rather than the reward for surviving one.

The setup: it’s 1862, weeks after bats stole London and dragged it into a cavern miles beneath the earth’s surface. Death has become an administrative inconvenience. Rats hold opinions. Hell maintains an embassy, staffed by devils who are unfailingly polite about it. You play a newcomer, which is the correct vantage point — everyone else has already normalized the apocalypse, and their casualness is the joke.

What makes this more than a lore tour is the creative direction. Emily Short — a foundational figure in interactive fiction, and the mind behind the Versu social simulation experiments — ran point on this project, and her fingerprints are everywhere. The relationship system here isn’t a dialogue tree with affection points. It’s a simulation that reads what you’re wearing, what you know, and who you’ve been gossiping about, then decides who finds you interesting. That’s a research lineage showing up in a commercial product, and on paper it mostly works.

Commercially, it’s a modest release: $19.99, released June 8, 2023 on Steam and Nintendo Switch, with a launch build buggy enough that Failbetter publicly owned the misfire before patching it. The audience it found loves it. The audience it lost bounced off the runtime and the mystery.

• Emergent attraction modeling: Romance and friendship aren’t menu selections. Your appearance, clothing choices, character-creation background, and the rumours you’ve accumulated feed a system that determines who gravitates toward you — and who quietly doesn’t.
• Physical story assembly: You collect motives, evidence, and rumours as discrete objects, then slot them into a sentence-shaped frame to build a testable theory. You then take that theory to a suspect and watch it survive or collapse.
• Three seasonal acts, built for replay: A single run lasts 3–5 hours. Branching is heavy enough that the design assumes you’ll play four or five times, chasing different threads — the murder, a specific relationship, or a full career grinding out census work.
• Romance is optional, and the game means it: Platonic and non-human options exist alongside the conventional ones, and opting out entirely isn’t a punishment path. The census-taker route is a real ending, not a consolation prize.
• A murder with a living victim: The corpse is up and walking. Your friend is still framed for it. The investigation is therefore about narrative, not forensics — which is the whole thematic point.

The Relationship Engine Is the Real Product

Strip away the mystery and Mask of the Rose is a social simulator, and a startlingly ambitious one. Most games that promise “emergent relationships” deliver a hidden integer that increments when you pick the nice option. This does something harder. Characters notice things. They notice what you wore to the meeting. They notice you know a piece of gossip you shouldn’t. They form an impression of you that you did not explicitly author, and then they act on it.

The payoff is a specific and rare feeling: being perceived incorrectly and having to live with it. Dress for practicality and someone may read it as severity. Mention a rumour to seem informed and you become the person who traffics in rumours. The system is built to generate social consequences that feel like consequences rather than gates. When it works as designed, it produces the kind of character moment scripted dialogue trees can only fake.

It also creates the best argument for replaying. Your second run isn’t “the same story with different choices.” It’s the same city meeting a different person.

The Deduction System Is Where It Falls Apart

Now the problem. The story-crafting mechanic is conceptually gorgeous: you assemble a theory from physical fragments — this person killed that person because of this motive, evidenced by this thing — and then test it on a suspect. Thematically it’s perfect. In a city where death is temporary and truth is negotiable, the murder investigation is explicitly about which narrative you can make stick.

Mechanically, it’s opaque. The game gives almost no feedback loop for evaluating whether a theory is close. You build a story, present it, and get a reaction that rarely tells you which component was wrong. There’s no cold/warm signal, no partial credit, no legible failure state. So the honest player behavior becomes permutation — swapping cards until something clicks. A deduction system that can’t tell you why you’re wrong isn’t a deduction system; it’s a combination lock.

This is a design failure, not a difficulty setting. Compare it to Return of the Obra Dinn, which validates in groups of three and thereby teaches you to reason in clusters. Or Her Story, which lets you fail productively because every wrong search still surfaces information. Here, the wrong theory teaches very little, and the interface’s tactility — the physicality of slotting evidence into a sentence — makes the emptiness behind it worse. Form is doing a lot of work that function hasn’t earned.

Onboarding Friction and the Three-Act Squeeze

The other structural problem is density. You are introduced to a character creator, a social system, an evidence system, a story-assembly system, a gossip economy, a job, and a cast of a dozen odd Londoners — inside a first act that’s maybe ninety minutes long. The game’s own structure works against comprehension. By the time the systems cohere, you’re in Act Three and the ending is arriving.

That abruptness is the runtime problem in disguise. Three-to-five hours is not inherently too short — it’s a defensible length for a replay-oriented design, and a tight four hours beats a padded fifteen. But the game asks you to internalize a lot before it pays out, and a first run frequently ends before the machinery clicks. Players who push through to run two and three are the ones who will love it. Players who stop at one are the ones asking where their twenty dollars went. Both are being reasonable.

The Writing

It’s excellent, and it’s the reason to be here. Failbetter’s prose is precise, funny, and comfortable with tonal whiplash — a joke about rat bureaucracy landing three lines before something quietly melancholy about grief in a city that no longer permits it. The inclusivity is structural rather than performative: pronouns, presentation, and orientation are character-creation inputs the simulation actually reads, not cosmetic toggles. That’s the correct way to do it, and more studios should copy the implementation.

The art direction is the game’s most conservative choice. Static character portraits over painted backdrops, in a muted palette of ink, bruise-purple, and gaslight amber. It’s handsome and it’s period-appropriate, but it’s not doing anything the genre hasn’t done. Compared to the moody, oppressive atmosphere Failbetter conjured in Sunless Sea with far cruder tools, this feels safe.

The UI is the more interesting critique. The story-assembly board — the drag-and-drop evidence frame — is by design the best screen in the game, tactile and immediately legible. The rest of the interface is dense to the point of hostility: overlapping panels for relationships, gossip, and inventory that never establish a visual hierarchy. Expect to hunt for information you know you have.

Technically, launch was rough. Failbetter acknowledged as much publicly, which is more accountability than most studios offer, and patched accordingly. But a text-heavy visual novel shipping broken is a bad look, because there’s no engine complexity to hide behind.

Steam (PC): This is the definitive version, largely because the story-assembly mechanic is a mouse mechanic. Dragging evidence cards into a sentence frame wants a cursor, and precision matters when panels are stacked this tightly. The dense UI also benefits enormously from a large monitor — on a desktop display, the information architecture is merely busy rather than actively confusing. On a Steam Deck’s 7-inch screen, expect squinting.

Nintendo Switch: The handheld case is strong, and the touchscreen is why. In handheld mode, physically pushing evidence into place with a fingertip should be more satisfying than mousing it — the interaction was practically designed for direct manipulation. A 3-to-5-hour runtime also fits handheld play patterns perfectly; this is a two-commute game, or a couple of evenings on the couch.

The cost is legibility. This is a text-dense game with small type and cluttered menus, and the Switch’s screen makes the relationship and gossip panels harder to parse than they already are. Docked, you get the readability back but lose the touch input and fall back to stick-driven cursor control, which is the worst of the three configurations for card manipulation. Play it handheld or play it on PC — docked Switch is the compromise nobody needs.

Mask of the Rose is a fascinating near-miss. The relationship simulation is a real advance — a system that models being seen rather than being liked, and one we want other studios to steal immediately. The prose is the best in its genre. The decision to make romance fully skippable, with friendship and career paths that carry equal weight, is quietly one of the smartest structural choices in a modern visual novel.

But the murder plot is the spine of the thing, and the mechanic holding it up doesn’t work. A deduction system that can’t communicate partial progress trains players to brute-force it, and brute-forcing a story you’re supposed to be authoring is the exact opposite of the intended experience. Pair that with a runtime that ends before the systems fully register, and you get a game that splits its audience down the middle — brilliant for those who commit to multiple runs, frustrating for anyone who gives it one.

We can’t recommend it for most people at full price. If the premise hooks you, go in knowing you’re buying a brilliant social engine and a mystery that never quite earns its keep — and consider waiting for a sale if $20 for a first-run afternoon gives you pause.