DEV Community

dev game
dev game

Posted on

Why Disco Elysium Bet on Unity: The Boring Book-Press Behind a 5-Million-Copy Detective Novel

Short answer: Unity, on purpose for words — not for graphics

Disco Elysium was built in Unity. The infobox lists Engine: Unity, and that choice shipped from Windows on 15 October 2019 to macOS on 27 April 2020, to The Final Cut on 30 March 2021 for PC/PS4/PS5, to Switch/Xbox on 12 October 2021, and even to Android on 5 August 2025 — on a minimum spec as low as a Core 2 Duo with 2 GB RAM and a DirectX 11 card Steam.

That spread is the clue. ZA/UM did not pick Unity to win a lighting benchmark. It picked Unity because Disco Elysium is not really an open world with dialogue in it. It is a 1-million-word novel with a walkable map attached.

Most explainers say: “indie studio, no money, so Unity.” That is true and boring. My different take after playing the loop closely:

Disco treats Unity as a publishing house, not a game engine. The valuable Unity features were cheap isometric cameras, boring 2D UI canvases, and fast C# iteration on text — not physics, not Nanite-style density, not frame-rate heroics.

If you want the one-sentence steal for your own project: when your hardest rendering problem is reading, rent the most boring book-press you can.

What ZA/UM actually built, in one paragraph

ZA/UM started as an Estonian cultural collective in 2009, registered as a game company in 2015, prototyped in a Tallinn squat from around 2016, then moved key staff to London and Brighton for voice and development capacity. Founders Robert Kurvitz and Kaur Kender came from novels — the 2013 novel Sacred and Terrible Air sold about a thousand copies — and from years of pen-and-paper sessions that became Elysium ZA/UM.

Disco Elysium is a single-player, isometric, open-world RPG with almost no traditional combat. Events resolve through skill checks and dialogue trees across 4 attributes — Intellect, Psyche, Physique, Motorics — and 24 secondary skills, rolled on 2d6 with clothing modifiers that can help or hurt Wikipedia. It won Best Independent, Best Narrative, Best RPG and Fresh Indie at The Game Awards 2019, passed 5 million copies, and sits on several greatest-of-all-time lists Wikipedia.

Steam’s Final Cut pitch makes the scale concrete: 24 skills, 80+ clothing items, 14 tools, 6 psychoactive substances, 60 thoughts to think in the Thought Cabinet, and full voiceover Steam.

Why Unity fit a novel that walks

1. Isometric + painterly is a budget strategy, not a style accident

Disco is presented isometrically. The player drags a detective around Revachol, clicks NPCs and highlighted objects, and reads. That presentation lets a small team fake a city with oil-painting backdrops instead of modeling one.

Unity is unusually good at this unglamorous job: cross-platform 2D/3D with a mature 2D UI stack, cheap cameras, and C# scripting that non-engine programmers can iterate on quickly Unity. You do not need Chaos destruction when your explosion is a paragraph.

My read, labeled as analysis: Disco uses distance as compression. The isometric pull-back hides missing animations, while painterly art turns low geometry into intention. It is the opposite of the modern AAA instinct to push the camera into pores. Pull back, write better, spend the saved triangles on more words.

2. 24 voices in your head need UI, not AI

The famous system is not just “24 skills.” It is 24 narrators. Inland Empire, Drama, Electrochemistry and the rest interrupt, argue, lie to you, and offer different actions depending on investment Wikipedia.

In Unity terms, that is mostly: dialogue trees + white/orange text + portrait flashes + dice UI. No navmesh breakthrough required. What is hard is writing and tuning — thousands of branches that must stay coherent when you fail.

This is why Unity’s reputation as “easy for beginners” Unity actually helped a team of novelists. C# iteration on dialogue logic is forgiving. A novelist can break a build at 2 p.m. and fix the condition at 2:20 p.m. without waiting for a custom tools team. The engine got out of the way of prose.

Try the feeling in miniature: Detective IQ: Brain Games asks you to master ten detective skills one teaser at a time. It is far simpler than Disco, but the loop is the same — read, infer, commit to a check, live with the result. Use it as a 5-minute warm-up for Disco’s 2d6 anxiety.

3. Failure as content only works if checks are cheap

Disco’s most radical rule comes straight from Kurvitz’s D&D years and the flop of Sacred and Terrible Air: embrace failure ZA/UM. High Drama makes you a brilliant liar and a paranoid mess. High Electrochemistry protects you from drugs while pushing you toward them Wikipedia. Red failed checks do not just block you — they open different jokes, injuries, and story branches.

My contrarian point: most engines punish failure with reloads; Unity let ZA/UM make failure cheap to author.

Because a skill check is just a dice roll + UI + flag, designers could afford to write the “you humiliate yourself” path. If every check required bespoke animation or physics, the budget would force binary pass/fail. Cheap checks fund generous failure. That is a production insight, not a lore insight, and it is why Disco has replay value without new maps.

4. Thought Cabinet is a slow install bar — and Unity loves slow UI

The Thought Cabinet lets you internalize ideas over in-game hours for bonuses with trade-offs. The textbook example is Hobocop: ponder living on the streets to save money, lose composure while you think, then earn double from trash once completed. You can later spend a point to forget a thought to free space Wikipedia.

Mechanically, this is Fallout-style traits stretched over time. Technically, it is timers + slots + modifiers — classic Unity UI work.

My idea, labeled as mine: the Cabinet is Disco’s real leveling system, and it is deliberately boring. No flashy skill tree, just rumination. Unity’s unsexy Canvas is perfect for rumination. A fancier engine would tempt you to visualize thoughts as planets. ZA/UM kept them as index cards, and the restraint makes them hit harder.

If you like that slow-burn linking of clues across rooms, Escape Room: Strange Case 2 is a useful trainer. You collect notes, connect symbols across locked sections, and learn to take notes like Harry should — prove the code, don’t guess it. The asylum is not Revachol, but the habit is: observe, connect, revisit.

5. Portability as literary distribution

Look at the shipping list again: Windows, macOS, PS4, PS5, Stadia, Switch, Xbox One/Series, Android Wikipedia. Minimum PC: Core 2 Duo, 2 GB RAM Steam. Unity’s cross-platform pitch — desktop, mobile, console, even non-game industries Unity — is usually marketing. Here it is strategy.

Disco does not need a 1060. Recommended is a 1060, minimum is almost a toaster. That means students, laptop novel-readers, Switch handheld players and late Android ports can all enter Revachol. My take: ZA/UM chose reach over ray tracing because its bottleneck was literacy, not GPUs. A gorgeous game that only runs on expensive cards would have filtered out exactly the text-hungry audience that made it 5 million.

What Unity cost them

No honest engine choice is free.

1. The game looks still because it is still. Isometric Unity with painterly sprites will never give you Cyberpunk crowds or RAGE traffic. ZA/UM compensates with writing density — but if you come for systems-driven crowds, Revachol feels empty. That emptiness is intentional, but it is also a Unity-2D-budget tell.

2. Click-precision jank. Highlight hunting, pathing around Martinaise, mis-clicks on tiny interactables — classic Unity point-and-click friction. Disco mostly avoids combat so this never kills runs, but it does kill flow. The fix the team chose (bigger highlights, Kim as guide) is editorial, not technical.

3. No physics punch. There is no satisfying door kick because there is no door physics to kick. When Disco wants violence, it narrates it. If you need systemic slapstick, Unity can do it — but ZA/UM spent that budget on words. That is a trade, and you should copy it consciously.

4. Studio risk dwarfs engine risk. After success, Kurvitz, Kender and other leads left in acrimony, projects were cancelled, layoffs followed, and by 2026 ZA/UM shipped Zero Parades to weak sales while at least four ex-ZA/UM teams chased spiritual successors Wikipedia. Engine stability cannot save authorship instability. If you clone Disco’s stack without cloning its writing room, you get an empty UI.

Why not Unreal, Godot, or custom?

Common advice says Unreal for beauty, Godot for freedom, custom for control. For Disco, all three would have been wrong at the time:

  • Unreal (2016-2019 era): Heavier baseline, stronger 3D pipeline, weaker rapid 2D UI iteration for novelists. You would pay for lighting you do not need and fight for text tools you do need.
  • Godot (pre-4.0): Closer in spirit, but console porting and hiring in London in 2016-2019 favored Unity. Disco needed PS4/PS5/Switch/Xbox without a porting saga. Unity already solved that boring problem.
  • Custom C++: The Witness/Teardown path buys control and costs years. ZA/UM were novelists with no engine team. Custom would have turned a 3-year text sprint into a 7-year tools project. Unity’s C# scripting API over a C++ runtime Unity gave them just enough control without owning the whole stack.

My sharp version: Unreal renders worlds; Unity publishes them. Disco needed a publisher.

What indie devs should actually steal

  1. Make failure a branch, not a wall. If your check can only say yes/no, you wrote too little. Write the humiliation path. It is cheaper in Unity than any new level.

  2. Treat skills as unreliable narrators, not stat sticks. Give each skill a voice and an agenda. Players will remember “my Inland Empire lied to me” longer than “+2 Logic.”

  3. Use clothing as readable UI. Disco’s 80+ garments telegraph builds at a glance and modify checks Steam. A hat that makes you stupid but charming is better tutorial than a tooltip.

  4. Ship on toasters. If your game is text-first, test on the worst laptop you can find. Every extra million players who can run at 2 GB RAM is worth more than one extra reflection probe.

  5. Prototype dialogue in-engine, not in Docs. Unity’s fast C# iteration means writers can play their words the same day. Keep writers in the build.

Practice the observation muscle without the 60-hour commitment: Detective Holmes: Hidden Object trains the exact Disco skill of scanning a messy scene for the one small clue that reframes the case. Play it to warm up your eyes before you interrogate Measurehead.

FAQ for searchers

Was Disco Elysium made in Unity?
Yes. Engine: Unity Wikipedia. Language-side, Unity scripting is C# over a C++ runtime Unity.

When did it come out?
15 Oct 2019 Windows, 27 Apr 2020 macOS, Final Cut 30 Mar 2021 (PC/PS/Stadia) and 12 Oct 2021 (Switch/Xbox), Android 5 Aug 2025 Wikipedia.

Is there combat?
No traditional combat. Skill checks + dialogue trees resolve events, with 2d6 rolls Wikipedia.

What is the Thought Cabinet?
A secondary system where you internalize thoughts over hours for lasting bonuses and occasional penalties, with limited slots you can expand or forget Wikipedia.

How big is it really?
Steam lists 24 skills, 80+ clothes, 14 tools, 60 thoughts, full voice in Final Cut Steam. Minimum 20 GB, 2 GB RAM.

Bottom line

Disco Elysium proves a heretical production rule: the best engine for a masterpiece can be the most boring one.

Unity did not make Disco beautiful. Oil paint and sentences did. Unity made Disco shippable, clickable, portable, and cheap to fail in — and then got out of the novel’s way. If your game’s core verb is “read and decide,” copy that humility: rent the book-press, spend on words, let failure be content.

Top comments (0)