CONCEPT DOC
Synopsis
How do you solve a murder mystery if you can’t trust your own, or anyone’s, memories? The year is 3026, and memory is no longer a given right. Access to each person’s “Memory Core,” implanted at birth, remains to the few and privileged. You are a detective in such trying times, which may not have been the right profession to pick looking back now… Here you are, though, stuck in the middle of an ocean with a bunch of murder suspects and a dead body. You seemingly have all the time in the world to uncover secrets through accessing memory cores. Surely you can solve this before the storm lets up and you head back to the headquarters with your suspect? That’s what you thought at least, until people started dropping like flies.
The game will be centered around talking to other characters, viewing their memory cores, and finding clues around the mansion you happened to be trapped at. We use music and visuals to create an eerie atmosphere for environmental puzzle-solving. Players can explore physical rooms and interact with objects to trigger related memories. Different items unlock separate memory fragments to encourage environmental exploration, and shared objects allow you to jump perspectives. Certain items must be examined in order or combined to reveal full clues. The narrative will slowly reveal that everyone reshapes their memories for personal motives; as the detective, you must uncover the repressed, blurry memories and expose the underlying conspiracy.
Immerse yourself in a game of deception, fading memories, and emotional distortion. How did these people get here? Why are they here? Why are you here? When did it all happen? Can you keep track of your own memory well enough before the memory cores start degrading? Who will you believe, your own intuition or the memory cores? And more importantly, can you trust your memory enough to solve this mystery before you get sucked in too?
Tones
Distrust:
The subjectivity of memory cores and how they degrade with time inherently create a very difficult time for the player detective. We want to emphasize this sense of distrust, making them doubt everything from clues to characters. As the player watches more and more contrasting viewpoints of the same scenario, they’ll naturally start doubting more and more.
Paranoid/Foreboding:
As more people start getting killed, the player will feel more scared that they’re next. The other characters will express a similar form of paranoia as they increasingly refuse to speak to anyone (including the detective player). Getting access to their memory cores will get more difficult; the player will also start to be wary of being in the room with certain characters alone. We will strengthen the music and atmospheric elements here to further creep out the player.
Mystery Solved?:
Above all, the players are solving puzzles and piecing together an answer to the mystery. How do we define truth when memories are so shaky? Who and what do we trust? Physical evidence? People? A possible stretch goal would be to make this game have multiple possible “truths” as endings, and any finding of them would achieve success. But the point is that maybe there is no definitive “truth”?
Setting
The game unfolds in the year 3026. Every human receives a Memory Core implanted at birth, a biotech device that archives lifelong recollections. Developed initially to treat cognitive decline and aid psychotherapy, this technology reshaped law, privacy and society. Memories are no longer private inner experience — they can be accessed, examined and even altered.
A baffling murder shakes the city. Stymied by traditional interrogation, a detective is assigned an unconventional task: rather than questioning witnesses face-to-face, they synchronise with Memory Cores to dive directly into others’ memories.
Crucially, stored memories are not objective footage. Upon entering a memory, the detective experiences events strictly through the host’s first-person eyes. Every recollection carries personal bias, repressed trauma, omissions and emotional distortion. Two core rules govern memory immersion:
First, the longer the detective lingers within a memory sequence, the more the scene degrades. Vision blurs, details fade, and the line between fact and subjective imagination softens, eroding perceptual reliability.
Second, memories can be deliberately forged. People may rewrite their recollections out of fear, guilt, grief, or a desire to protect others. Fabricated fragments mix with naturally warped memories, with no immediate visual cue to separate falsehood from genuine experience.
As the detective uncovers memory fragments from victims, suspects and bystanders, conflicting accounts of the same incident surface. Spotting contradictions between separate perspectives becomes the investigation’s key breakthrough. Players cross-reference disjointed viewpoints, filter distortions, distinguish natural bias from intentional forgery, and piece together a coherent version of events.
The game’s core theme: there is no single, absolute objective truth. All human recall is filtered through subjectivity. Multiple logically consistent interpretations of the murder can be assembled from available clues, supported by branching multiple endings. Players may identify a suspect who meets legal criteria for conviction, yet they can never verify whether their reconstruction matches the unknowable “true” reality. Their conclusion remains merely one viable interpretation among many.
No supernatural forces exist in this world. All tension stems from memory’s inherent unreliability and the ethical burden of invading another’s consciousness. Fragmented, disorienting memory visions are rooted in a lived-in near-future city. We reject generic flashy sci-fi aesthetics: even with widespread memory technology, ordinary people retain mundane daily lives. This contrast lends human weight to abstract, fractured memory spaces.
Memory Core technology fuels deep societal friction. Debates rage over the legal validity of memory evidence; citizens fear surveillance and memory theft, while an underground trade for memory manipulation and data trafficking thrives beneath the city. These tensions form the backdrop of the murder investigation and raise the central question: if evidence is inherently flawed and alterable, how can truth ever be defined?
Gameplay
Platform: Desktop-first (Steam-shaped, or simply a GitHub Unity project any teammate’s laptop can build). AR is a possible later stretch (see Stage 3), not a starting requirement.
Players: Single-player only. Suspects are finite-state-machine NPCs, not LLM agents — their lies need to be authored and consistent so players can actually catch contradictions across memories, which a live model improvising fresh dialogue each run would undermine. Multiple player extension using a turn-based mechanic on one device (e.g. not IP address tracking for multiple devices).
***Staged prototyping — build the loop before the look.***
Stage 1 — 2D gridworld, mechanics-first: A top-down tilemap mansion (Unity’s Tilemap + 2D colliders), Game Boy-style presentation. The goal here isn’t atmosphere, it’s proving the actual game: walk up to an object, trigger a memory, log a statement, catch a contradiction. Every core system gets built at its cheapest possible version first.
- Object interaction: trigger colliders, walk-up-and-press-to-examine.
- Memory sequences: a separate tilemap room or simple scene swap, entered on interact, exited back to present-day.
- Dialogue: ScriptableObject-based branching nodes — text, speaker, player-choice array, each choice pointing to the next node (Wayline: dynamic dialogue with ScriptableObjects, branching narratives and player choice).
- Journal/evidence log: a runtime list of flagged statements, tagged by suspect and memory, viewable and comparable side by side.
- If the loop doesn’t work here — if catching a contradiction isn’t satisfying with placeholder art and boxes for rooms — no amount of later polish fixes it. Prove it cheap first.
Stage 2 — richer presentation once the loop holds: With mechanics validated, invest in visuals without necessarily jumping to full 3D or animation:
- Comic-strip memory panels as a serious option, not just a fallback: memories rendered as a sequence of static illustrated panels (with pans/zooms via Timeline, or even just cross-fades) rather than fully animated scenes. This sidesteps animation budget entirely, gives you a strong, deliberate visual break between “present-day gridworld” and “inside someone’s memory” (satisfying the memory-vision UI requirement for free), and fits the theme well — a memory recalled isn’t a live recording, it’s already a flattened, selective reconstruction, which a comic panel looks like in a way a smooth 3D replay doesn’t.
- If budget allows, upgrade the gridworld itself (better tile art, lighting/parallax, sound design) before considering a jump to 3D — same Jenkins framing (enacted story via the mansion layout, evocative space via mood/music) applies whether the world is tiles or full 3D geometry, so this is a visual upgrade, not a mechanical rewrite.
- Full 3D with imported Sketchfab mansion assets remains an option here, not a requirement — only worth it if Stage 1 proves the loop is strong enough to deserve the extra art/engineering cost (model cleanup, camera scheme, navigation).
Stage 3 — AR, experimental: Only attempted once Stages 1–2 are solid, and explicitly understood as narrowing playtesting to the one iPad on hand. If pursued, the natural fit isn’t rebuilding the whole game in AR — it’s a single showcase beat (e.g. one pivotal memory rendered as an AR scene in the player’s real room) layered on top of the already-working desktop game, not a platform switch.
Systems that don’t change across stages:
- Memory sequences — Embedded Narrative (Jenkins): the game space becomes something to decipher backward through what a memory reveals about it. Whether rendered as tilemap, comic panel, or full 3D, the structural move is the same — leave the present, see the past through one person’s eyes, return.
- Mansion exploration — Enacted Story: the plot is your movement through the space; the layout paces the mystery (a locked wing withheld until you’ve earned a key, a forced detour that puts you alone with a suspect). True at every visual fidelity level.
- Cross-referencing contradictions — Emergent Narrative: the actual puzzle layer. We don’t script the “aha” — the journal holds contradictory flagged statements side by side, and players construct the accusation themselves. This is pure data/UI, entirely stage-independent.
- Evocative Spaces: the isolated-mansion-in-a-storm premise does atmospheric work by loading a pre-existing genre association before a single clue is found — carried by music/sound/lighting, which can be as simple or as produced as each stage’s budget allows.
UI, at a glance (stable across stages):
- Examine prompt — context-sensitive prompt on interact-eligible objects.
- Memory-vision cue — Stage 1: a simple screen-border color shift or filter; Stage 2: the comic-panel transition itself does this job.
- Journal/contradiction log — persistent UI listing flagged statements, comparable side by side; the actual mechanical carrier of “who do you believe.”
Challenges
Key Challenges for Design
Making contradiction-catching actually fun. The whole game lives or dies on the moment a player notices two memories don’t line up — if that moment isn’t satisfying, nothing else saves it. The journal needs to make it easy to compare different statements side by side, but it can’t just point out the answer for you.
Keeping the mystery exciting when there’s no single right answer. Multiple endings sound great, but they risk making players feel like it doesn’t matter what they pick. Every ending needs to demand a solid chain of evidence so players feel like they actually figured it out, not just guessed.
Balancing time between memories and the mansion. If players spend too long in memories, the mansion becomes a loading screen; if they spend too long in the mansion, it feels like busywork. The two sides need to feed each other — finding something in the mansion unlocks a new memory, spotting a detail in a memory tells you where to look back in the mansion.
Key Challenges for Tech
Memory degradation system. The longer a player stays in a memory, the more the visuals and audio need to fall apart — even Stage 1 needs a basic version of this. How fast it degrades, what details disappear first, whether we warn the player — all of that needs to be easy for design to tweak, not hardcoded.
Keep core systems separate from visuals. Dialogue, journal, memory transitions, contradiction tracking — all of this has to work independently from whatever the game looks like, or we’ll be rewriting everything each time we move from 2D to comic panels to 3D.
Dialogue system needs to be flexible enough. ScriptableObject-based dialogue nodes are fine as a foundation, but our needs go beyond basic branching — every line needs to be taggable by suspect, memory, and topic, and we need support for conditions like “this line only shows up after you’ve seen Memory X.” This has to be solid in Stage 1 because everything else builds on top of it.
Key Challenges for Art
Making Stage 1 feel moody on a tight budget. The Game Boy-style top-down tilemap is cheap on purpose, but a murder-on-a-stormy-island setup still needs at minimum some lighting, rain particles, and sound to set the mood. A consistent color palette and a screen shake on thunder go a long way toward keeping the prototype from feeling too flat to properly test whether the tension mechanics work.
Players need to instantly tell reality from memory. You have to know right away whether you’re in the mansion or inside someone’s head, and different people’s memories should look different too. A color shift, a change in framing, or slightly different tilesets per character can get across the difference between one suspect’s nervous recall and another one’s self-serving version of events.
Memory degradation has to look intentional. Degradation is part of the gameplay, not just a visual effect — it needs to clearly tell the player “this memory is getting less reliable” without looking like a glitch. Art and programming need to build this together, because what it looks like matters just as much as how fast it happens.
Intended Audience
The intended audience is people 17+.
Players who enjoy narrative-driven mystery and puzzle games — fans of titles like Return of the Obra Dinn, Disco Elysium, or Her Story that reward close attention and cross-referencing over reflexes — and who are drawn to slow-burn psychological tension over action. Likely college-age-and-up given the murder-mystery premise and themes of memory, guilt, and unreliable truth, but accessible to anyone comfortable reading dialogue and piecing together a narrative rather than needing fast execution or combat skill.
Appendix (Team Tainotherium Blogs)
