AEC hackathon · glazing and life safety

Visual Vibes

Watch how people look at a building.
Specify what works for them.

TeamVision · Voice · Vibe

  • Zack Sickman
  • Mohammad Fasahat
  • Joaquin Martinez
  • Tyler Andersen
  • Darin Holbrook
  • Blake Gibbons

How it works

Recorded demo walk: walking into the apartment, turning to the curtain wall and holding the look until it is captured
From the recorded demo walk · about real time

  1. 01

    Walk

    A first-person walk through a real apartment, in the browser. No install.

  2. 02

    Look

    Only steady looks count: turning under 20°/s, moving under 0.6 m/s.

  3. 03

    Link

    Every pane, mullion and sofa is tagged with its IFC class and specification.

  4. 04

    Specify

    The report names the glass, with its make-up, VLT and g-value.

Visual Vibes · glazing study

The test

Same windows.
Only the glass changes.

Four products swap on both facades at once, every 20 seconds. The walker never touches a control.

Because the geometry never changes, size can’t bias the result.

VLT · clear 76% · neutral 60%
frit 45% · bronze 32%

Observer view: the glass auto-rotates from low-iron clear to neutral solar control

Clear → neutral · observer view

Subject view: the walker drifts along the wall and the glass turns from neutral to bronze with nothing on screen

Neutral → bronze · subject view

The glass turns from bronze to fritted dot, then the facade is seen from the garden

Bronze → frit · then the garden

The fritted dots vanish as the glass returns to low-iron clear mid-look

Frit → clear · mid-look

Visual Vibes · glazing study · four real swaps from one recorded walk, sped up 1 to 3×

The metadata

Every element knows what it is.

13

properties per glass product: make-up, coating, VLT, g-value, U-value, Ra, reflectance, Rw, thickness, weight, cost, carbon, lead time

22

curtain-wall panels tracked one by one, from inside and from the garden: 11 modules × vision and high-light

10

design traits on every element, from curved geometry to privacy, plus IFC class, Uniclass and spec section

One remark, as recorded

“too dark and brown, gloomy”

stood at x 2.6, z 1.2 · heading 167°
eye height 1.6 m
looking at curtain wall, 4.6 m away
glass on show bronze tint
its spec VLT 32% · g 0.28 · Ra 82
tone −2 · counted against bronze

Visual Vibes · glazing study

During the walk

Every look lands on a spec

Observer view from the recorded demo walk: the walker turns to the curtain wall, the look is captured, and the panel shows the element in view with its system, module, transom, frame and glass

Visual Vibes · glazing study · Observer mode, recorded demo walk at about real time

One demo walk · 2 min 10 s · 7 remarks · verdict: leaning

Specify: neutral solar control

6 mm solar-control toughened / 16 mm argon / 6.8 mm laminated · triple-silver coating · 08 81 00

VLT 60%

g 0.33

U 1.0 W/m²K

Ra 94

Rw 36 dB

This walkLow-iron clearNeutralFritted dotBronze tint
Steady looking49.3 s/min40.7 s/min50.2 s/min45.1 s/min
Time on show40 s20 s40 s30 s
Remarks0 (3)+2 (1)−2 (2)−2 (1)
Share of preference31%39%, chosen16%14%

Looked at a little less than clear or fritted glass, but the only glass spoken of warmly. Each remark point moves a glass’s score by 25%.

Visual Vibes · glazing study

Decode walk · spec sheet

The answer arrives with its product data

Glass options side by side: how this person responded, then each product's make-up, coating, VLT, g-value, U-value, Ra, reflectance, Rw, thickness, weight, cost, carbon and lead time

How this person responded

Steady looking, time on show, remarks and share of preference for each glass.

From the product metadata

Make-up, coating, VLT, g-value, U-value, Ra, reflectance, Rw, thickness and weight.

Beside the preference

Cost index, embodied carbon and lead time, so the trade-off sits on one sheet.

Visual Vibes · glazing study · report

Decode walk · glass explorer

Any property against attention

Scatter of embodied carbon against attention per minute on show, one dot per glass on each facade, coloured by remark tone, with a trend line

Pick a field

VLT, g-value, cost, carbon, lead time, or any of the ten design traits.

Like with like

Each dot is one glass on one facade. The geometry never changes, so size can’t win.

A lead, not a verdict

r shows the trend across elements. One walk is a lead to test with more people.

Visual Vibes · glazing study · report

Decode walk · design DNA

What drew the eye, trait by trait

Design DNA: plain-language findings above bars for ten traits, sought out versus in view, with positive and negative remark weights

Ten traits per element

From curved geometry to privacy, scored 0 to 1 in the metadata.

Sought out vs in view

Bars right of centre: the person centred their view on that trait more than it surrounded them.

Remarks by trait

Each remark is weighted by how strongly the element in view carries the trait.

Visual Vibes · glazing study · report

The facade, module by module

Where on the curtain wall people look

The south curtain wall unrolled with attention heat, then bars per module from inside and from outside, strongest on modules 4 to 6

M05 · 20.6 s

most-looked module

2%

of looking went above the 2.7 m transom

62.8 s · 6.9 s

steady looking from inside, and from the garden

Visual Vibes · glazing study · report

The remarks

Every remark is pinned where it was said

A remark is spoken: the words appear in the note bar and become note 1, tied to the curtain wall and the low-iron clear glass
Said while looking: the words become a note tied to what was in view and the glass on show · 1.25×
The 3D review: stepping back from the wall to see five numbered remark pins where they were said, in front of the heat on the curtain wall
The 3D review: each numbered pin stands where its remark was made, in front of the heat · up to 2×

Visual Vibes · glazing study · recorded demo walk

Decode walk · element schedule

A BIM schedule with behaviour in it

Element schedule rows for each curtain-wall glass and the north windows: IFC class, Uniclass and spec section, key properties, time shown, attention, captures, remarks and a preference bar

Classified

IFC class, Uniclass code and spec section for every element as seen.

Specified

Key properties straight from the metadata: make-up, coating, VLT, g, U and Ra.

Measured

Time on show, attention, captures, remarks and a preference index. Exports as CSV.

Visual Vibes · glazing study · report

Any platform · Rhino 9 plug-in

Any platform you need.
We built it for Rhino.

The walk, the tagged metadata, the heat map and voice notes run as a native plug-in on the Rhino model itself. One more step toward a fully digital design process, with the data to make the best call for the client.

Rhino 9 with the test room in the Perspective view, an attention heat map painted on the walls, floor and sofa, beside the Vision, Voice, Vibe window listing time on each tagged object and the event log
Heat map painted on the Rhino model
Rhino 9 looking at the W-02 ribbon window with its tag and heat, beside the Vision, Voice, Vibe window with voice on, the window's glass and frame spec, and a spoken note linked to W-02
Voice note linked to W-02 and its glass spec

Visual Vibes · Rhino 9 for Mac · screenshots from a live session

Rhino to the web app

Send a Rhino model up.
Walk it in the browser.

The plug-in also works the other way: objects from Rhino go up to the web app, where a client can pick them as a layout, walk them, and see exactly which ones held their attention.

A ring-shaped pavilion with four wall openings and a floating roof disc, modelled in Rhino's Perspective view

01

From Rhino

Send the model’s objects up to the web app

The web app's Choose a layout screen: 01 Original study, with the Glazing Study, Garden Apartment and Open Loft, or 02 Rhino geometry, the embedded model with navigation, dwell captures and a surface heat map

02

Pick a layout

The original study or the Rhino geometry

The Rhino ring walked in the web app with heat cells on its surfaces, beside the Rhino model observer: the mesh being looked at, session time, captures, heat cells and an attention ranking of every mesh

03

Deep dive on any object

Surface heat, and every object ranked by look time

Visual Vibes · Rhino geometry in the web app

Stock falling off pallet racking across an aisle, with dust, beside a goods lift

Second study · life safety

Factory
Egress

Same engine, a harder question: can people find their way out?

Night shift · alarm · falling stock · power cut · every exit sign carries its own metadata

Factory egress · the spawn

The farthest point from every exit

Walking-distance field over the factory plan: light near the three exits, dark far away, with 5 m contours and the four farthest spawn points circled

29.8 m

on foot to the nearest exit from this spawn

3 exits · 5 decoys

two stairs and an external door; a lift, two shutters, an office and a sealed exit

14 signs

illumination, size, mounting and rated viewing distance on each

Visual Vibes · factory egress · walking distance on a 0.25 m grid, lighter is closer to an exit

Factory egress · the scenario

Pressure, one switch at a time

An alarm and a shake start the run. Stock then falls a few metres ahead on the walker’s own shortest route, always leaving another way out.

Camera shake has its own switch for anyone prone to motion sickness.

Stock falling across the aisle

Stock falls onto your route

Racking aisle in a power cut with a lit exit sign

Power cut: emergency light only

Smoke in a dark aisle with a hanging exit sign

Smoke builds up

A former exit door barred shut with a NO EXIT sign

Decoys: sealed exit, lift, shutters

Visual Vibes · factory egress

Factory egress · the signs

Every sign is judged on what people did

In view

On screen, facing you, nothing in between, and enough light and clear air for its type.

Readable

Within its rated distance: Z factor × sign height, 200 for lit signs and 100 otherwise.

Looked at

Held within 3° of the centre of view for a quarter of a second, view steady.

Followed

Within 4 s, moved the way it points, or towards its door.

Observer panel during a run: nearest exit 24.4 m on foot, hesitating, aisle blocked and power cut, looking at sign SG-05 with its pictogram, standard, colour, illumination, battery duration and size

Visual Vibes · factory egress · Observer mode, live

Factory egress · one run, decoded

The route against the shortest one

Route map: walked path in blue against the shortest route in green, a hatched blockage, hesitation circles, wrong-turn crosses, numbered remark pins and sign triangles

Out in 1:02

by Stair B, after the aisle was blocked

56% efficient

82 m walked against a best possible 46 m

3 wrong turns

37 m of detours, 11 s inside the dead end

1 wrong door

tried the goods lift during the power cut

Visual Vibes · factory egress · one scripted demo run

Factory egress · sign audit

Every sign gets a verdict

Sign table: each of the 14 signs with type, mounting, rated distance, time in view, readable and looked at, first-look distance, followed, an arrow check and a verdict

Worked

SG-14, a lit hanging sign: looked at from 4.7 m, then followed east.

Missed

SG-08, low-level photoluminescent, and SG-13: readable, never looked at.

Wrong

SG-09, a superseded text sign, points west. The shortest route heads north-east.

Visual Vibes · factory egress · report

Factory egress · findings

The report says what to fix

Findings list: exit and route, time at the spawn, signs that changed the route, signs readable but never looked at, the lift tried, the blockage and re-route, the power cut, the longest blind stop and the dead end
A superseded EXIT sign above a pallet block at the end of the dead-end aisle

SG-09 points straight into a pallet block. Behind it is the sealed former exit.

Visual Vibes · factory egress · report

Where it goes next

From one room to every project

Load real projects

Bring in a Revit or Rhino model; its IFC properties become the metadata.

Life-safety reviews

Test signs, decoy doors and routes with people before construction.

Real eye tracking

Run the walk on a VR headset with eye tracking instead of the screen centre.

Group studies

8 to 12 people per study, with decoded vs stated picks as the accuracy check.

Product libraries

Use manufacturers’ glass, finish and sign data as the options people test.

Visual Vibes · github.com/MF-Little/AEC-Hack-Project