Technical Game Design, 2025
UNREAL ENGINE
·
BLENDER
·
FIGMA
·
NEW YORK
·
SUBSTANCE PAINTER
·
CONSOLE & PC
·
INDEPENDENT GAME DEVELOPER & DESIGNER
·
WET WORK
·
UNREAL ENGINE · BLENDER · FIGMA · NEW YORK · SUBSTANCE PAINTER · CONSOLE & PC · INDEPENDENT GAME DEVELOPER & DESIGNER · WET WORK ·
Current Project
Tomorrow: Vector Six
A Mossad agent races to find those responsible for a shocking and suspicious terrorist attack before he is forced to take the fall.
-
The systems and visuals shown here; including player locomotion, animation, and character rigging, represent work which as been completed since development began in August 2025. While a lot of progress has been made, these systems are still under active development and subject to iteration as Tomorrow continues to evolve.
This page isn’t meant to reveal every technical or artistic detail, but rather to demonstrate my process as a Technical Game Designer — how I approached the problems presented by the project, integrated design intent with engineering, and translated ideas into an interactive experience.
Most features, systems, and narrative elements of Tomorrow are intentionally being kept under wraps. This is a developing commercial product, and this presentation focuses on what can be shared publicly while preserving its broader creative identity.
-
Tomorrow’s locomotion system has been developed from scratch but leverages the Unreal Engine 4 skeleton, which has been removed from many of the engine’s starter packages and templates, but once imported into Unreal Engine 5, can act as a more performant and well supported foundation for a game.
The UE5 Skeleton does make some improvements that do not require any of its new bones and additional complexity however, so Tomorrow is implementing a hybrid skeleton of sorts. Character Models have been created from the ground up for the UE4 skeleton, but many adjustments have been made. For example, to reduce the scale of the hands and arms.
-
Although the base mesh of both the physique and face are hand crafted, Metahuman technologies are being leveraged to elevate them. The implementation of those systems would have been much more seamless using what is now the recommended skeleton, but the result of this unorthodox approach (including the exclusion of motion matching and/or root motion within core movement, is a character that looks like something from the present, while feeling like something from the past.
This is how the rest of Tomorrow has been designed mechanically and narratively, too.
Masks On
From this point on, the most recent* versions of the character models will be used to showcase the technical design of Tomorrow. Many of the systems and mechanics being showcased here were designed and developed before the character models shown later were implemented, but you can learn more about that process here, where the process of creating the character models from block-out through implementation is outlined.
However, the decision to determine the right foundational skeleton early on was crucial, as it meant less rework would be required outside of the design department as the project progressed.
It helps that Tomorrow is a very singular and focused idea, rather than an attempt to be an infinitely scalable everything game. But even within those limitations it is possible to build superfluous technologies and systems, and better to build ones that scale into the software’s niche instead.
Tomorrow Vector is a startup disguised as a passion project. An inch wide, and a mile deep.
* Most recent assets as of 20/10/2025.
-
Tomorrow is intentionally avoiding the use of motion matching or root motion - at least for core locomotion. It doesn’t see anything wrong with those techniques but they clash with the creative intent of the project. As a result, it isn’t leveraging many of the animation templates provided by Unreal, and we have some basics to cover instead:
Character and Animation Blueprints handle basic movement, with left stick input (via the InputAxis mappings for MoveForward and MoveRight) driving both character translation and animation parameters through the Animation Blueprint.
The character movement component handles velocity and acceleration, while the Animation Blueprint interprets that data to drive blendspaces and transitions within the locomotion state machine.
-
The project leverages Unreal Engine 5’s default Foot IK control rig; however, it has been repurposed to work with the modified skeletal hierarchy. Luckily, this only required some bones to be renamed (and turning some things off) to align with the UE4 rig used in Tomorrow.
IK nodes (via the Two-Bone IK node and Control Rig adjustments) ensure accurate foot placement relative to the terrain underfoot. The default system uses per-foot line traces to adjust effector targets in real-time, maintaining somewhat believable grounding even across uneven surfaces, but there will be a more sophisticated solution put in place later.
“ Yes, both of them “
The game employs both directional and strafing movement depending on context. Directional input vectors are normalized and converted into animation parameters (Direction and Speed), which drive a 2D Blendspace in an Animation Blueprint.
The Blendspace enables smooth interpolation across eight directional locomotion animations (forward, backward, left, right, and four diagonals), producing responsive and continuous movement transitions.
This same system is extended to handle gait variations — walk, jog, and run. By blending between multiple blendspaces or leveraging Blendspace by Speed parameters, the character transitions naturally between walking and running without abrupt animation pops.
An Attack on the state
Tomorrow includes 3 stances; Standing, Crouching, and Prone, each with multiple sub-1.0 speeds and unique animation sets.
Standing and Crouching are implemented within a shared state in the Locomotion State Machine. Partly because there are plans for additional speed-specific animation sets for both of them and much of the functionality is shared. But primarily because prone required a much more complex and bespoke solution. There is some overlap however, since many transitions in both are handled through Animation Layer Interfaces and conditional blendspaces based on a Stance enumerator and speed values.
Capsule half-height and camera offsets are also modified dynamically to match stance transitions in both cases too, with a timeline used to soften the changes to relevant variables. This same timeline will be used to soften the movement of the custom camera setup too, but the more traditional camera setup shown here is used for demonstration purposes only.
As mentioned, prone is handled as a separate locomotion state due to its unique IK, collision, and movement logic. This will be covered in more detail later.
Hyperrealism supersedes reality
by being more coherent,
more emotionally legible,
or “truE” than reality itself.
-
Transitional animations like the stopping animations shown here have been added very sparingly.
This one was implemented by duplicating the blendspaces used for walk, crouch and prone movement, and replacing the movement animations with the respective stop animations, therefore syncing the directional blending logic across both. It was then placed within its own state in the locomotion state machine, and triggered whenever the ShouldMove boolean becomes false during movement.
Most of the complexity can be found in the timing of the triggers that mask the transition back to the idle pose. Tomorrow checks how much time is remaining in any stop animation currently playing (TimeRemaining (Ratio)) to perfectly synchronise it with the idle pose - or get close enough that the difference is imperceptible.
In other words, over-engineered. I believe this timing could/should be implemented within the animation assets themselves, but for now animations are still in development and subject to change.
-
This is the ONLY logic driven by animation currently.
Tomorrow has been designed to perfectly mirror play input and character output. These transitions are considered a safe zone to implement animation driven logic because player input in these scenarios is, by definition, zero. This logic can also be overridden by the player, since there are NO start animations, and the stop animations showcased here are cancelled as soon as the character starts moving again.
Tomorrow's theory is that immersion doesn't come from weight or graphical fidelity, but from a world the player really is altering using the buttons on a plastic controller.
-
I’m not an animator by trade — but Tomorrow had some specific requirements for its crouch animations, and I was responsible for creating them.
A crouch animation asset was exported from Epic’s Lyra template, retargeted to the UE4 skeleton, and then modified directly (and significantly) in Unreal Engine’s Sequencer. The movement style and rhythm provided a great base reference to work from, but the overall height of the character while crouching had to be much closer to half of its default standing height (or crouching would not offer unique gameplay benefits over walking), it had to adapt nicely to sub-1.0 playback rates (explained why later), and it had to have a particular vibe (see II and III).
-
Unlike many other games, Tomorrow does not have a Stealth Mode.
Crouching lowers the character’s overall height, which can sometimes allow you to hide behind small objects, or enter areas you otherwise couldn’t, but it doesn’t magically reduce your sound profile (more on that later) or make you less suspicious.
This is partly because crouching doesn’t reduce movement speed - or change it all. Tomorrow Vector doesn’t disconnect stance from stealth, but it stops assuming lower = better. Some stances (such as crouching or prone) can instead have a negative impact on your stealth rating depending on the context. For this reason (among some others) Tomorrow is careful not to call itself a stealth game.
With all of that that in mind, this particular crouch animation had to make movement that is very difficult to quickly (or slowly) in reality look as natural as possible.
-
Crouching carries symbolic weight. In other games and media, it often communicates predatory confidence — power, control, precision. In others, it evokes fear or powerlessness — a defensive, retreating posture.
But when children play hide and seek, both of those sensations exist simultaneously. They’re not really afraid, and they’re obviously not predators either. But they are being mischievous.
That sensation — the excitement of bending the rules, of playful subversion — is what defines this aspect of Tomorrow ’s gameplay. The protagonist is a professional and capable of great violence, but his defining trait is mischief. The crouch animation attempts to reflect this: a physical language that suggests he’s enjoying himself, at least a little, and an invitation to the player to play with the game’s mechanics and systems.
Curated Realities
Transition animations are only added where necessary to maintain Tomorrow’s hyperreal internal logic. There is almost zero transition between walking and crouching for example, but a transition was necessary between both of the above and prone.
By leaning into its identity as a video game, rather than fighting it, Tomorrow is able to give additional care and attention to the parts of reality it adopts. Not only is there specific transitions for each of the 8 possible directions of movement, the system even adapts to changes in input mid-transition.
For example, if the player stops moving while transitioning into, or out of prone, the animation will morph into an in-place animation. This is done using carefully timed (and tested) triggers that can occur while moving in or out of prone, reducing the likelihood of ‘sliding’ animations ocurring.
-
Not unlike how the crouch animation was developed, a prone animation asset was imported as a start point. In this case it was much harder to find a high quality base reference, likely because prone movement is quite tricky to implement and most avoid it unless it is 150% necessary.
Regardless, eventually an asset was found within a project created by Jakub - which is a treasure trove. Although Tomorrow is not using any of the mechanics or systems included in his project, many of its movement animations have been ported across. These have already been modified to various degrees (see II) and as the project progresses they will eventually become unrecognisable, but they were invaluable during more experimental phases.
-
The base animation reference wasn’t a million miles away from the prone movement style required for Tomorrow, but it was slightly too ‘slidey’.
Two possible causes were identified:
1. The hands never clenched, which detached them from the apparent physicality of the action.
2. The feet didn’t appear to be contributing as much to the momentum being generated as the upper legs.In response, the balls of feet now bent on impact and stretched on release, digging into the ground and pushing away more convincingly. While the hands being clenched brought in some much needed determination, which is required for something as unpleasant as crawling. The closed fists also appeared to keep the form clear, making the animation feel more cohesive overall.
-
Hundreds of other tweaks have been made too, such as adjustments to hand/arm positioning to reduce appearance of clipping, and the legs both reach a little further forward than they did originally. This fits with the game’s hyper-real aesthetic, and makes it possible to convincingly bump up the maximum speed of prone movement.
As a result, prone paradoxically felt quicker, but also more weighty and punchy.
on american soil
-
At the top it was outlined that Tomorrow had 2 movement modes, but prone actually required a third movement mode. Turning while prone introduces about 1,451 interconnected and unexpected problems, which means this project naively stumbled into all the same design flaws that the designers of Metal Gear Solid 2 solved for 20+ years ago.
The problem that still exists today is that the player collision capsules of typical Unreal Engine player characters (and most traditional video game characters) are designed to scale vertically. Changing between standing and crouching only adjusts collision along the Y axis, and this isn’t a challenge to account for - Unreal even has crouch functions built into the engine itself.
-
Prone is an entirely different beast. Literally, in a sense. Because it’s more akin to creating a new controllable character than adding a new ability.
It shrinks the Y axis and extends the X axis. This means that most of the existing collision or ledge detection wouldn’t work anymore, without extensive changes and additions.
There is already a bunch of these solutions implemented within Tomorrow that resolve these issues, such as line traces from head and feet to detect walls, ledges, and slopes for IK (more on that later), but there would still be 100s of flaws if it wasn’t for a single design decision.
To be clear, I don’t think there will ever be a silver bullet to solve for all of the ways prone breaks traditional video game movement. But, if you restrict the number of ways in which players can turn, this movement style can become a lot more manageable.
-
Both of the movement modes implemented to now:
1. Directional. The player character simultaneously follows, and turns towards, the current input direction.
2. Strafing. The player character follows current input direction, but continues to face away from the camera.
The new movement mode:
3. Backtracking. The player character follows current input direction of movement, but does not rotate toward the input direction, or the the orientation of the camera.
-
This new movement mode is only active when prone AND moving backwards, so a new function was created to determine whether the player character was doing so.
If so, orientation towards the camera or input direction is disabled, and a separate blendspace is enabled (Prone_bkwd) with animations that were initially just the same prone animations at -1.0 playback rate, but have since been modified. The player character can therefore move forward and backward while probe without needing to perform a 180* degree turn.
‘Alternate’
-
That might have seemed like a small change, but those 180 degree turns are responsible for the vast majority of problems discussed earlier.
Extending the X axis in the way prone does isn’t much of a problem when moving forwards or backwards, since a single line trace can check if its possible, functionally identical to the rest of the locomotion system, although slightly customised for prone’s unique dimensions.
But turning is different.
What would be a simple turn on the spot while standing turns into an extended pivot across a wide circumference while crawling, which means extensive calculations in multiple directions. And this is just to confirm if a movement is possible without clipping through a wall, it isn’t necessarily providing a means by which to navigate the 3D space.
In other words, even after all of that hypothetical effort, 99% of the time (unless there were bespoke solutions for every possible permutation), the result would be ‘you cannot crawl there’ anyway.
-
Anything below a 120” turn is less complicated, but movement speed and rotation rate are both reduced to varying degrees to prevent the character from spinning in circles or generating movement that doesn’t fit the turning animation that has been created.
Early in development a prone turn animation was implemented as a montage. It was visually impressive, but there were several issues synchronising the animations if the player kept turning, or quickly moving in and out of the turning montage. The solution to those issues was typically to have animation drive movement logic, but this isn’t a good fit for Tomorrow.
The current implementation is surprisingly simple on the Animation Blueprint side, with both turning animations being placed inside the Prone_fwd blendspace.There is a little more complexity on the Character Blueprint side, ensuring the right movement speed and rotation rate is applied to the character in order to make the movement look and feel right, but this is the right fit for Tomorrow’s hyper-reality.
thousands
-
The current implementation of Prone IK is very much a work in progress.
Currently, there is a sphere trace drawn from both the pelvis and the neck, with their respective hit locations (where they are touching the surface below) used to calculate if there is a slope and, if so, its angle. This same calculation is performed for the left and right shoulder bones too.
Then, if there is a large enough slope, the entire player character object is rotated to match the respective angles. This proved the calculations produced the intended results, but not much else.
However, the calculations and logic outlined here will be repurposed soon, using either bespoke animations or systemic adjustments to individual bones to alter the player character position more subtly. This updated approach should also allow the addition of multiple pivot points, so the character can respond realistically to ledges at the top of a slope, or where they meet the ground below.
-
When I experienced VR for the first time I was more blown away by the motion-tracking controllers, which allowed me to pick things up and throw them around like a child using their hands for the first time. It added a tactility that I didn’t even realise was missing.
It’s no longer surprising when a game features photo-realistic environments. But even in 2025 it’s rare for a video game character to respond to the suggested physicality of those environments.
The industry has been adding additional detail at a much faster rate than it can possibly conjure up methods of adding additional interactivity. There are attempts to offset this with modular interactive elements scattered throughout playable (but static) zones, but after more than a decade of use, these elements are so easily recognisable as disconnected from their surroundings (often even more so than the playable character) they might actually be detracting from the experience, if anything.
I think crawling is featured in so few games because; paradoxically, interactivity has not been the focus for a long time.
-
Tomorrow feels like something from the past because it leans into the idea that video game innovation has been frustratingly focused on graphical fidelity, and that going back 2 decades (almost exactly) has very little impact on game design, besides gutting always-online requirements, roadmaps and an overabundance of UX refinements that only served to shave off any edge games used to have.
Tomorrow goes back 2 decades and wonders what would happen if games learned from movies instead of trying to become them - or if games could became an interactive exploration of truth - or if they became an interactive exploration of truth.
But more than that, it wonders what happens when a person is willing to commit everything to the study of a single craft.
Relax ,its a game
not real
-
There are 3 distinct locomotion state machines implemented in Tomorrow; Base, Pistol and Rifle. An Items enumerator is then used to determine which of the inputs to use, all of which are wired into a Blend Pose node. with a 0.1 transition time - roughly the time it takes for a button to reset after being pressed.
There are no plans for equip or unequip animations, because items are equipped and unequipped with the single press of a button - or a menu that exists outside of the play space (and time).
Only if the mechanics of item selection were to change in the future would the inclusion of equip animations be reconsidered.
-
[REDACTED]
-
Because Tomorrow goes back a little over 2 decades and then reimagines a new Tomorrow for games (and the world) there isn’t (what would now be considered) a traditional over-the-shoulder camera setup.
However, turn-in-place animations were still required, so these animations were converted into montages and called in the character blueprint whenever the player aims a weapon, including 90 and 180 degree turns for both standing and crouching - a very different system will be implemented for prone later.
Note: A traditional over-the shoulder setup has been implemented here for demonstration purposes, but the [RECACTED] will be showcased at a later time.
-
These montages effect the entire body, but they’re called before the base, pistol and rifle locomotion state machines within the Animation Blueprint, so the upper body animations are safely overwritten by the relevant state machines whenever the player character is holding or aiming a weapon.
just a GAME
-
Tomorrow doesn’t feature a stealth mode, or anything else that would undermine mechanical depth in service of a streamlined user experience, because it does not view simplification as a solution to interpretability.
The slower you move, the quieter you are. This means pushing the analogue just enough is part of the challenge, as well as learning how much sound is generated at a given speed, but note that this isn’t a game where quieter and slower are always the best options (more on that later).
To implement this, Tomorrow sets an isSneaking variable based on whetherplayer speed is more or less than 0.5, then that variable is plugged into a blend by bool node in the Animation Blueprint that blends between the slow and regular version of the walk, crouch or prone animations.
In addition to this, both the slow and regular animation playback rates (how fast the animation plays) are set by the amount of player input. For example, if player is pushing the analog a quarter of the way forward, the slow animation will be selected, but it will also play at a very specific 0.25 speed.
This is also occurring for the regular movement too, which makes all of it feel unmistakably representative of player input.
-
The UX department has requested a toggle for movement speed.
-
The UX department has requested a toggle for movement speed, again. This time they’ve included player data.
-
The UX department considers the accessibility option proposed sufficient.
-
The base aim rotation of the Character is called from inside the Animation Blueprint, then this value is used to systemically modify the transform of the player character’s spine.
-
Tomorrow will not be cancelled due to the political climate in the United States of America, or anywhere else in the world.
Tomorrow will maintain its creative identity across all regions, and will choose not be sold in regions where this creative identity is not respected - rather than undermine what it claims to stand for.
Tomorrow will be playable without an internet connection and it will likely be possible to obtain a copy illegally. Please do not do this - unless you are living in a region in which Tomorrow is not being sold.
Tomorrow may implement online features, but these will be optional, and it will stand as a creative work independent of those features.
Tomorrow will receive no post-launch updates or patches. There will be no roadmap. It will not get better, or worse. It will be.
Tomorrow will not be perfect.
-
Tomorrow is coming soon.
A deeper look at Tomorrow’s modelling, rigging, and animation pipelines.
