3D asset presentation · Human-guided AI
From Verified 3D Asset to Interactive Digital Exhibit
A human-guided AI workflow that transforms a validated 3D reconstruction into a six-environment, multi-camera, music-aware presentation experience.
The approved master GLB, geometry, UVs, atlas, materials, signage, and freeze records remained unchanged.
01 · Presentation architecture
Two clearly separated modes turn one asset into six coherent viewing experiences.
Day Cycle is an atmospheric timeline. Studio is a set of curated display stages. Their controls stay grouped so the viewer always understands whether they are changing time or changing the exhibition context.
DAY CYCLE
Day · Dusk · Night
Smooth background, sun/moon direction, ambient, key, fill, shadow, exposure, and music transitions form one continuous timeline.
STUDIO
Review · Japanese · Showcase
A neutral review stage, a Japanese night setting, and a restrained dark gallery each receive their own lighting and soundtrack without changing model scale.
02 · Non-destructive runtime layer
The exhibit changes the presentation—not the approved asset.
Lighting, tone mapping, background blending, window glow, and signage illumination are applied at runtime through cloned materials and non-exported presentation state. The frozen GLB remains the immutable source of truth.
MASTER GLBRead-only · unchanged
GEOMETRY / UV216 meshes · existing atlas preserved
MATERIAL EFFECTSRuntime clones · not exported
PUBLIC BOUNDARYPrivate evidence and workstation metadata excluded
03 · Camera and sound direction
Framing and music became part of the product language.
Front, driveway, opposite, back, and top views share fixed orthographic profiles across all six environments. Day Cycle and Studio keep separate 3/4 compositions, while all three Studio presets share one consistent perspective. Each scene selects an original track and crossfades with the environment.
Consistent inspection
Five orthographic views preserve distance and visible scale across every environment.
Context-aware 3/4
Day Cycle and Studio use different approved perspective profiles to match their backgrounds.
Optional cinematic motion
Slow rotation remains user-controlled and works from every camera family.
Music-aware scenes
Six Gemini-generated original tracks are mapped one-to-one to the six environments.
04 · Human-guided generative production
Generative tools supplied options; human review created the final art direction.
The project owner evaluated scene fit, rejected studio directions that felt too decorative or too gold, and chose a restrained dark Showcase. ChatGPT and Codex supported the final background generation and iteration. Gemini generated the six original music tracks. Human camera tuning then matched the model to every scene.
Human direction
Defined scene purpose, rejected mismatched visual styles, supplied final camera profiles, and approved presentation behavior.
ChatGPT + Codex
Supported background exploration, final image generation, implementation, interaction repair, and visual QA.
Gemini
Generated six original tracks that give each time or studio preset a distinct sonic identity.
05 · Interactive digital exhibit
Move through all six environments in the browser.
Switch between Day Cycle and Studio, blend time smoothly, select fixed cameras, enable cinematic rotation, and hear the scene-specific music. Public presentation controls intentionally omit development telemetry and scale-reference tools.
Browser autoplay policy may wait for your first click before music begins.
06 · Presentation QA
Thirty fixed-camera combinations and every transition were checked without touching the master.
The pass verified six backgrounds and six audio assets by hash, exercised all five orthographic views across six environments, confirmed one shared Studio perspective profile, tested smooth day-night blending and scene audio, round-tripped the public GLB, and recorded no console or page errors.
Presentation QA: passed · Frozen master modified: no · Public scale and telemetry tools: excluded
07 · Scope and rights
A presentation layer can enrich the asset without rewriting its history.
Approved asset boundary
No building geometry, prop geometry, UV, atlas, material architecture, texture, signage copy, or frozen record was modified.
Runtime-only glow
Night window and signage effects exist only in presentation state and are not written back to the downloadable GLB.
Generated media disclosure
Final backgrounds were developed with ChatGPT and Codex; the six original tracks were generated with Gemini under project-owner direction.
Independent presentation
This remains an independently produced visual reconstruction and digital exhibit, not an official manufacturer model or survey-grade record.
Sixth FlowForAI case
The verified asset is now a complete exhibit.
Together, the two Fujiwara cases show both halves of the product journey: evidence-to-asset reconstruction, then asset-to-experience presentation.