Stray Memories: What the Cat Adventure Taught Designers

Stray memories and the design lessons hiding in a cat adventure

A two-button character who cannot speak, a walled city that treats the player as a tourist, and a robot cast that carries the entire emotional load. Several years after its release, Stray still returns to design conversations because its choices are unusually legible. The phrase “stray memories” tends to surface in two different ways: as a vague nostalgic tag for a player’s own experience inside the game, and as shorthand inside studios for the design ideas the title left behind. Both readings point at the same thing. The game made a small set of mechanical and narrative commitments and then committed to them hard enough that other teams can study them without guessing at intent.

This article is written for game developers, designers, and technical leads who want to do more than recall a favorite moment. The goal is to extract the design patterns the title depends on, the production constraints that made them possible, and the limits of those patterns when a team tries to translate them into a new project. A few core questions frame the rest of the page: how did a four-legged protagonist become readable, how did the world signal its own rules, where did the narrative weight actually come from, and which of those choices survive contact with a different scope or budget?

Because “stray memories” can also name a design retrospective, the analysis below treats the title as a case study rather than a review. The Wikipedia entry on the Stray video game is useful for a baseline on development history and release context, and a long-form Stray review from The Verge provides an external critical perspective on the design choices discussed here.

Why a cat protagonist changed the design conversation

Designing for a non-human body forces every system to earn its place. Jump arcs, collision, camera placement, interaction radius, animation priority, and even menu logic all change when the player character is a four-legged animal. The team behind Stray leaned into that constraint instead of fighting it, and the result is one of the clearer examples in recent years of how a strong character silhouette can carry gameplay communication.

Several consequences matter for other projects:

The cat’s movement vocabulary is small but readable. Walk, run, jump, climb, and a single contextual interaction let players predict what is possible at any ledge or prop.
The camera orbits at a height that keeps the cat’s body silhouette visible against the environment, so the player always understands the spatial relationship between avatar and geometry.
Animation blending has to be more forgiving than usual, because felines move through postures humans do not. A team that has not budgeted for additional blend states will feel the cost quickly, usually during the first vertical traversal in front of a playtest group.
Sound design for footsteps, landings, and idle breathing has to be re-recorded against quadruped contact points. Reusing a human footstep library reads as wrong even when the volume mix is fine.
Save and interact prompts have to be repositioned to the cat’s eye line. A team that ships prompts anchored to a human height will break the avatar contract on every menu screen.

For developers, the takeaway is that a non-human avatar pays for itself in clarity only when the rest of the design surrenders to that avatar. A cat in a control scheme built for a human will feel like a costume swap. A cat in a scheme rebuilt around quadruped motion, vertical surface interest, and small-object interaction will feel like a different genre.

Environmental storytelling without quest markers

The neon alleys and residential pockets of the game’s walled city are dense with readable props. Mural fragments, cables strung between balconies, stacked crates, and lamp posts tell the player where to go without a quest arrow. That design choice is harder than it looks, because it pushes the team to invest in art direction, lighting, and prop placement instead of guidance UI.

A useful mental model is to think of each environment as three layers that designers tune together:

Anchor props are large, contrasting elements that establish the player’s next broad direction. A glowing sign, a shaft of light, or a colored doorframe can carry this role.
Path props are mid-scale items that confirm the route between two anchors. Cables, awnings, and ledges belong here.
Texture props are the small details that make a path feel inhabited rather than staged. A dented can, a torn poster, or a sleeping robot fill this slot.

When all three layers are present and aligned, designers can ship fewer explicit waypoints. When the texture layer is missing, the path starts to feel like a level rather than a place, and players begin to ask for the compass icon. Stray is a useful case study for teams who want to retire that compass, because the title shows what it costs to do so in pure production time, and where the discipline has to be written into the brief rather than left to environment artists to figure out on the fly.

Carrying narrative weight through companions, not monologue

The robots in the city do most of the emotional work. The cat is expressive through movement, but does not speak. That division of labor is one of the strongest design decisions in the project, and it is also the decision that is hardest to copy without planning. A team that puts a silent protagonist next to talkative companions has to accept that all the script weight, all the cinematics budget, and most of the localization cost now lives on the supporting cast.

For narrative designers, several points tend to come up in postmortems on this kind of structure:

Companion dialogue must move even when the player is exploring. Lines that only trigger during cutscenes feel like padding once the player has spent twenty minutes with the world.
Voice direction becomes a high-leverage skill. A small cast of well-acted robots can feel richer than a large cast of competent but unremarkable humans.
Quest pacing has to thread the needle between solitude and company. The cat is the lens, the robots are the meaning, and a designer who forgets either side of that relationship loses the player’s investment quickly.
Repetition is a script risk, not a writing risk. Companions greet the player when the player returns to a hub, and a script that has not planned for return visits will feel like a bug rather than a feature.
Localizing a companion-heavy project takes longer than a silent project, because every line of every companion has to be re-recorded, not just subtitled. Teams that plan only for subtitle cost usually overrun during voice production.

Studios thinking about a similar companion-driven structure should plan the script budget before they plan the level count. A common failure mode is to write a small script and then stretch it across a large world. The result feels thin even when the environments are dense, and playtesters will name the thin script as the reason the world feels empty.

Verticality as a genre signal

One of the more underappreciated choices in the title is how verticality is treated. Rooftops, beams, pipes, and interior balconies all stay in the player’s mental map. The cat adventure sits at one end of a spectrum of readability choices, with rich geometry and dense prop placement used to keep the player’s mental map intact across chapters.

Vertical design depends on three things staying in sync: camera framing, jump arc, and recovery affordance. If any one of those drifts, the player loses confidence in their own position. The table below summarizes the most common failure modes and the diagnostic signal that usually appears first.

Failure mode	What the player reports	Diagnostic signal	Cheapest fix
Camera loses the avatar against the wall	“I can’t tell where my cat is”	Frame-by-frame silhouette is below 8 percent of screen area for more than two seconds during a climb	Add a slight orbit offset when a vertical surface is detected
Jump arc is too flat for the spacing	“I keep missing that beam”	More than one failed attempt per ten vertical traversals on a representative playtest build	Extend arc apex and add a small magnet assist on the next ledge
Recovery from a fall is punishing	“I died because of one missed jump”	Death clusters around specific vertical sections in telemetry	Add a forgiving respawn near the last successful interaction, not at the section start
Companion pathfinding breaks on ledges	“The robot just stood there”	Companion idle timers above a few seconds on routes the player has already cleared	Snap-to-path logic on marked ledges rather than free navigation
Climb prompt competes with context prompt	“I pressed the button and nothing happened”	Input log shows the same button triggering different actions within one second	Prioritize the prompt closest to the avatar’s facing direction

The most expensive mistake is usually the camera. It looks like an art problem, but it ships as a control problem because players blame themselves for missed jumps that the framing actually caused. Teams that catch this in playtest will save a lot of post-launch support work.

Audio cues as the player’s second minimap

Because the cat cannot speak, audio carries more weight than it would in a voiced-led adventure. Distant music swells, the clatter of robots in the next room, the hiss of steam, and small ambient loops tell the player what kind of space they are about to enter before they see it. Designers working on similar projects often underestimate how much of that work happens during sound design rather than composition.

A useful production approach is to budget audio the way a team would budget a level. Each environment gets a short brief that names the one or two signals the player must hear. If the brief is missing, the soundscape becomes wallpaper, and players stop using it as a map.

Define a primary audio anchor for each new region. A single identifiable loop or motif that the player can recall later, even if the music has shifted into a more passive state.
Treat companion voices as a spatial signal as well as a story beat. A voice that comes from a believable direction doubles as navigation, and a flat mix will lose that benefit.
Keep a low-frequency bed constant across most environments so that new high-frequency events still register as events rather than as part of the texture.
Re-record or re-source any sound that was originally built for a human avatar. The cat’s contact with the world is different, and a footstep library that worked for a boots-led project will not work for paws on concrete.
Plan for accessibility. A project that uses audio as a map needs a parallel visual cue, because players with reduced hearing still need a way to follow the same signal.

How stray memories show up in a postmortem document

When a studio writes a postmortem after a release, the entries under “things we would keep” and “things we would change” are usually where the phrase stray memories starts to make sense as a category. These are the design memories that outlive the project: choices the team would defend again, even knowing the cost. For a cat adventure, the list is unusually short and unusually specific.

Design choice	Why a team would defend it	Cost a team has to accept
Four-legged avatar with a small move set	Clearer player intent, stronger identity, less tutorial UI	More animation states, more camera tuning, more collision work
Companion-driven narrative	Emotionally readable without voice for the avatar	Heavy script budget, large voice direction cost, tight pacing constraints
Quest-marker-light guidance	Stronger world reading, fewer UI overlays	Heavier art direction, more prop passes, longer environment sign-off
Vertical-first level design	Mental map stays coherent, traversal feels distinctive	Companion pathfinding, camera framing, and recovery logic all need extra time
Audio as a second minimap	Players can navigate with sound off the screen	Larger sound design budget, more briefs per environment, stricter mixing pass
Hub-and-chapter pacing	Companions earn screen time without crowding the avatar	More save-state logic, more companion pathing in confined hubs

Studios borrowing any of these choices should think of them as a package. Adopting the avatar without the camera work, or the vertical design without the audio work, usually produces a product that feels like the original in stills but loses the player’s trust within twenty minutes of play. The package framing is also the most useful answer to a producer who asks why the team needs more time on a reference title that “looked simple”.

Lessons that do not transfer cleanly

Not everything from a cat adventure survives a different brief. A few patterns look portable on paper and break in practice, and they are worth naming explicitly so other teams can avoid the most common misreads. For additional context, The Verge’s long-form review of the title is one useful reference for how those choices read in front of a general audience.

Silent protagonist does not equal weak protagonist. A silent human reads as a marketing placeholder. A silent animal reads as a deliberate design choice. Teams that want a silent human will need stronger camera acting, more readable body language, or a different framing device.
Companion-heavy scripting is not a shortcut for a small world. A short, dense script can outstay its welcome in a small world because the player meets the same companions too often. The inverse is also true: a long, thin script will not carry a long world. Length and density have to be matched deliberately.
Quest-marker-light guidance only works when the art direction is consistent. A team that has not aligned palette, lighting, and prop scale across the project will find that visual cues compete rather than guide. The result is the worst of both worlds: no markers and no readable path.
Verticality is not a difficulty toggle. A vertical world is a structural commitment, not a feature flag. Teams that want to ship a flat version later usually find the level kit was designed for a vertical read, and rebuilding it costs more than the original.
The cat does not scale to every genre. A quadruped avatar in a tactics game, a management sim, or a first-person shooter will read as a gimmick, because the silhouette fights the camera and the controls. The avatar works when the genre is built around traversal and observation, not when the genre is built around shooting or stacking menus.

A practical reading order for designers studying the title

Studios that want to use the game as a reference tend to get more out of it when they approach it with a checklist rather than a free play session. A focused reading order helps a team pull specific decisions out of the experience without losing them to the emotional memory of the moment.

Play through the first chapter once without commentary, paying attention to where the player already knows where to go without any UI prompt.
Play through the same chapter a second time, this time noting the moment a prop, a sound, or a camera choice was the cue.
Watch a recorded playthrough of a player who is not the team, and count how many times they looked for a quest marker that was not there.
Read the public development commentary, the Stray Wikipedia entry, and at least one long-form critical review, such as the Stray review on The Verge, to separate what the team intended from what the audience noticed.
Translate the cues from step two into a one-page brief that names the anchor, the path, and the texture layer for each environment the new project is about to enter production on.
Run the brief past a designer who has not played the reference. If they can predict the cue from the brief alone, the brief is doing its job.

The last step is the one most often skipped. A reference study that ends at “this was great” stays as stray memories in a folder. A reference study that ends at a brief becomes input for the next milestone, and the brief is also the artifact that survives staff turnover on a multi-year project.

How the lessons interact with different project sizes

Design lessons do not move between scopes for free. A pattern that fits a small indie project can become a tax on a mid-size project, and a pattern that fits a large production can feel cheap on a small one. The table below maps the most common lessons from the cat adventure to the kind of team that is likely to benefit from each one, and to the kind of team that should treat the lesson as a warning instead.

Lesson	Fits small team well	Fits mid-size team well	Fits large team well	Risk to watch for
Non-human avatar as identity	Yes, when the move set is small	Yes, with a dedicated camera owner	Yes, with a clear animation budget line	Underestimating blend state and collision work
Companion-driven narrative	Yes, if script length is bounded	Yes, with a dedicated writer and voice director	Yes, with a script supervisor across chapters	Companions repeating lines in a small world
Quest-marker-light guidance	Yes, with a strong art director	Yes, with prop and lighting leads aligned	Rare, because UI overlays are usually cheaper to maintain	Direction drift between art and level design leads
Vertical-first level design	Yes, when pathfinding is simple	Yes, with a dedicated traversal designer	Yes, but only if the kit supports it from day one	Companion pathfinding breaking on ledges
Audio as second minimap	Yes, with one good sound designer	Yes, with briefs per environment	Yes, with a dedicated audio director	Soundscape becoming wallpaper without briefs
Hub-and-chapter pacing	Yes, with a small hub area per chapter	Yes, with a hub designer role	Yes, with a content director watching the pacing curve	Hubs feeling like a tax on traversal time

The interesting cells are the ones where the lesson does not fit the team size. A small studio can sometimes ship a quest-marker-light title with a single art director. A large studio will struggle to keep that consistency across multiple art leads without a clear brief, and the cost of the discipline often outweighs the savings on UI. Teams that spot the mismatch early can either downscope the lesson or scale the team up to match it. Teams that ignore the mismatch usually find the cost in late production, when the change is most expensive.

Where stray memories sit in a studio’s design culture

Most studios have a small set of references that everyone on the team has either played or heard about. Those references are useful because they let a producer say “we want this to feel like the rooftop chapter” instead of describing a target from scratch. A cat adventure is one of those references for many teams now, in part because its design choices are easy to point at. The next step for a studio is to decide which of those choices are real design goals and which are just shared memories of a personal experience.

Several questions help separate the two:

Is the choice we want to copy a mechanic, a moment, or a feeling? Mechanics are portable. Moments depend on art and pacing. Feelings depend on the whole project and rarely transfer without a lot of supporting work.
Does our team have the production time to make the choice work, or are we picking it up because it sounds good in a slide deck? The cheapest version of any of these lessons is still a real cost.
Would our target player recognize the choice as a strength, or would they only notice it if it failed? Strong references stay invisible when they work. If the audience would notice the choice as a feature, it is a marketing decision, not a design decision.
Has anyone on the team shipped something similar before, or is everyone reasoning from memory? Memory of a reference is a starting point, but a shipped experience with that reference is a much better predictor of cost.
Is the new project’s audience similar enough to the reference’s audience to value the same cues? A cat-shaped readability signal works for the audience the original targeted, and may read as noise to a different audience.

The honest answer to all of these for a cat adventure is “it depends on the team”. That is the version of stray memories that survives a project’s planning phase: not a feeling, but a set of constraints the team is willing to defend. It is also the version that helps a producer push back on a designer who wants to add a feature without naming the cost.

A short checklist for teams borrowing from the title

Before locking a reference like this into a design document, a producer or lead designer can run a short checklist against the team’s current scope. The list below is a starting point, not a rule, and each item should be answered in one or two sentences in the actual document.

Do we have an animation owner who has shipped a non-human avatar before, or do we need to budget extra time for blend states and collision?
Have we written the script length per companion, or are we hoping the writers will “find the right size” in production?
Is the art direction brief specific enough to support a quest-marker-light world, or will the level designers have to compensate with overlay UI?
Is the level kit built for vertical play, or will we need a separate traversal designer to make the world feel consistent from the avatar’s eye line?
Does the audio brief name the one or two signals per environment that the player must hear, or are we shipping a soundscape and hoping the mix holds?
Has a voice director signed off on the companion cast size, or are we planning to record more than we can direct?
Is the companion pathfinding tested on the same ledges the player can climb, or will the companions look broken in every vertical section?
Is the save and interaction UI anchored to the avatar’s eye line, or are we shipping a human-height UI in a non-human project?

If the answer to more than two of these is “we will figure it out”, the reference is probably a memory rather than a design input. Either the team is comfortable paying the production cost up front, or the team is collecting stray memories that will not survive contact with the milestone plan. Either answer is fine, as long as the team is honest about which one it is.

What the lessons look like in a one-page design brief

Most of the value in a reference study comes from the act of writing the brief, not from playing the reference a second time. A one-page brief that names the cue, the system that delivers it, and the cost of delivering it is usually enough to take a designer from a memory to a plan. A brief that stops at “we want it to feel like the rooftop chapter” is closer to a mood board than a design input.

One structure that has worked for small teams is a four-row table per environment. Row one names the anchor prop and its color or light signature. Row two names the path props and their spacing. Row three names the audio anchor and its expected mixing level. Row four names the camera framing and the recovery affordance for any vertical surface in the area. Each row gets one or two sentences, and the whole table fits on a single sheet.

When the brief is written that way, the milestone plan can point at it directly. A level designer can ask whether a new blockout matches the anchor and path rows. A sound designer can ask whether the new area has its audio anchor in the brief. A producer can ask whether the camera framing row has been tested with a playtester who has not seen the area before. Each of those questions is a small, specific check, and that is the kind of check that survives a project.

Closing practical note

The phrase stray memories works both as a tag for a player’s experience and as a quiet warning to a design team. The title is a useful reference because its design decisions are visible, but the visibility is the result of a production discipline that did not advertise itself. A team that wants to learn from it should treat the lessons as a package, plan the production cost up front, and write the answers to the checklist above into the design document before the first level goes into blockout. The design choices that survive the planning phase are the ones that will still be worth defending at the postmortem, and the ones that will not need a marketing pitch to explain to a new hire.

Frequently asked questions

What does “stray memories” usually mean in a game design context?

In a game design context, the phrase usually means either a player’s emotional memory of the cat adventure Stray or a studio’s internal label for the design choices the title left behind. Both readings point at the same thing: a small set of legible decisions that other teams can study, including a four-legged avatar, companion-driven narrative, and quest-marker-light guidance.

Why do designers keep returning to the cat adventure as a reference?

Designers return to it because the title’s choices are unusually easy to point at. The move set is small, the camera framing is distinctive, the supporting cast carries the script, and the environment does most of the guidance work. That combination makes the project useful as a reference for both small indies and larger studios trying to retire some UI overlays.

What is the hardest design choice in a cat adventure to copy?

The hardest choice to copy is usually the companion-driven narrative. The cat does not speak, so every line of script, every cinematic beat, and most of the localization cost sits on the supporting cast. Teams that underestimate this cost tend to ship a small script stretched across a long world, and the result feels thin even when the environments are dense.

Can a small studio realistically ship a quest-marker-light world?

Yes, but only if a single art director owns the palette, lighting, and prop scale across the entire project. Quest-marker-light guidance depends on environmental readability, and readability drifts quickly when more than one person sets the tone. A small studio with a strong art direction brief can do it. A larger studio without a brief will usually fall back to UI overlays because they are cheaper to maintain.

How does verticality change camera and pathfinding work?

Verticality changes both systems. The camera has to keep the avatar silhouette readable against walls and beams, the pathfinding has to handle ledges and climbable surfaces for companions, and the recovery logic after a missed jump has to be forgiving without erasing the player’s sense of risk. None of those changes are free, and they all need to be designed together rather than in isolation.

Why is audio so important in a silent-protagonist project?

Audio carries spatial information that would normally be carried by voice or UI. Distant music swells, the clatter of a companion in the next room, and small ambient loops all act as a second minimap. A soundscape that does this work is the result of briefs per environment, and a soundscape without briefs usually becomes wallpaper that the player stops using for navigation.

What production role is most likely to be underestimated on a project like this?

The voice director is often the most underestimated role. A small cast of well-acted companions can feel richer than a large cast of competent but unremarkable humans, and the difference is almost entirely in direction rather than script. Teams that budget voice recording without budgeting direction usually end up with a longer recording schedule and a flatter result.

How can a team turn a reference study into actual design input?

A team can turn a reference study into design input by ending the study with a brief. The brief should name the anchor, the path, and the texture layer for each environment the team is about to enter production on, and it should connect each cue to a specific system in the project. A reference study that ends at “this was great” stays as stray memories in a folder. A reference study that ends at a brief becomes input for the next milestone.

Are the design lessons from the title still relevant for non-animal avatars?

Some lessons transfer and some do not. The companion-driven narrative and the audio-as-second-minimap patterns can work with a human protagonist, although the silent-protagonist framing reads differently. The non-human avatar and the vertical-first level design are more specific to a quadruped silhouette. Teams borrowing lessons for a human-led project should pick the ones that survive the change of avatar, and treat the others as context for the original.

What is the cheapest way to start a reference study without overcommitting the team?

The cheapest start is a one-day pass with a single designer and a single reference checklist. The designer plays the first chapter twice, fills in the anchor, path, and texture rows for one environment, and presents a one-page brief at the end of the day. If the brief is useful, the team can decide whether to expand the study. If the brief is empty, the team has lost a day rather than a sprint.