[+_+] FOUNDATION // BOOT 000
loading generated stills
AI ASSET▸▸STATE▸PLAY
PLAY ↘
R/ 3Q · STILL 04
01

AREA // 01-A
RESEARCH

USE WITH CURIOSITY AND AT YOUR OWN PACE.

Foundation studies how AI-generated video and image assets become interactive, high-quality experiences: micro-games, simulations and learning & development scenarios.

// generated video,
made playable.

FOUNDATION®BY K28 LAB

X/ EXPLODED · STILL 05
ENTER THE LAB
loading research…
◂

AI VIDEO AND IMAGE ASSETS,
REBUILT AS INTERACTIVE WORLDS.

⚡01‹
002 / Thesis

An asset is
a recording.

Here it is
a state.

Scroll or drag. What you are turning is eight seconds of generated video, scrubbed frame by frame by your input. No 3D model exists.

Interaction

⟵ drag ⟶ she turns with you. On the cover, she follows your cursor.

000°
003 / Method

Break the
asset apart.

01Assetconsistent cast sheets & one keyframe per moment+
02Stateeach clip becomes a game state, loops close on the same frame+
03Engineinputs, timing windows, rules and endings, written as data+
04Experienceplayable, measurable, re-renderable in 4K+
BREAKDOWN 000%

Rebuild it
as a system.

Same pixels. New behaviour. The interface is drawn live on top of the generated image.

// ENGINE & STATES

clip · inputSTATEloop  |  cut
0playable games
0generated clips
0shared engine
03D models

Every clip is a state and every input is a cut. Timing windows inside loops, reaction rings, hotspots, choices, targets and meters are written as data.

One file per experience. Rules and endings decide what plays next, so the same engine runs a boss fight, a heist or a training simulation.

005 / Demos

Playable demos

Built end-to-end from AI-generated assets. Same engine, different data. Desktop: keyboard + mouse. Touch supported.

006 / Applications

Where it goes next

:A1MICRO-GAMES
FEEL · LOOPS · SCORE

Micro-games

Prototype game feel in days: generate the world, wire the states, play it before any art pipeline exists.

[+_+]
:B2SIMULATIONS
BRANCH · TIME · CONSEQUENCE

Simulations

Scenario footage becomes branching, timed decisions with consequences rendered as video, not diagrams.

[+_+]
:C3LEARNING & DEV
DRILL · SCORE · DEBRIEF

Learning & development

Decision drills for teams: realistic situations, scored reactions, a debrief screen and repeatable runs.

[+_+]
:D4PRE-VIS
PITCH · TRAILER · TEST

Pre-visualisation

Test interactive concepts, trailers you can play and pitch demos with production-grade visuals.

[+_+]
007 / Method

Open tooling

Video Game Forge

The provider-agnostic skill and engine behind every demo. Bring any image or video model (MCP or API): it designs the state graph, keeps the cast consistent, turns clips into states, tunes timing from pixels and renders 4K gameplay frame by frame.

github.com/Kayariyan28/video-game-forge ↗
ZSH // ~/.claude/skills
git clone https://github.com/Kayariyan28/video-game-forge \
  ~/.claude/skills/video-game-forge

# then ask Claude:
"Make a playable stealth game from AI video."
  1. 01Pick providers
  2. 02Design states
  3. 03Sheets → keyframes
  4. 04Image-to-video
  5. 05game.json
  6. 06Playtest
  7. 07Render 4K
008 / Research log

Research log

R-01Boss-fight grammar: reactions, abilities, phasesshipped · DUSTFALL
R-02Stealth: timing windows inside generated loopsshipped · NIGHT PAWS
R-03Reflex arcade: read-the-motion outcomesin progress
R-04Point-and-click investigation with clue hotspotsplanned
R-05L&D decision simulation with scoring & debriefplanned
R-06Generated voice dialogue treesexploring