#root{display:none}.faq__panel{grid-template-rows:1fr!important;visibility:visible!important}[data-reveal]{opacity:1!important;transform:none!important}Skip to content

Daylight baseDusk result

1440 × 960 pxDrag to pan. Scroll or pinch to zoom.

Back to home

Pavilion evening study

Ready for review

Included exampleDusk

Procedural pavilion, not a client project. No AI edit.

This is the included example, running in your browser. Compare the two images, then check the three groups and sign. Nothing you do here is saved or sent.

Dark areas need your review

New dark areas appeared at dusk. Check that they hide no building detail; nothing here is approved automatically.

How this is decided

Newly dark pixels: 31,487
Largest dark area: 10,814 px
Newly bright pixels: 0
Largest bright area: 0 / 900 px

Policy dusk-review/1. Raw detector: fail. At dusk, new dark areas are expected, so they come to you instead of failing the render. New blown-out light still blocks export. The detector’s own result stays in the record as measured.

An explicit review policy, not a calibrated universal image-quality metric. Darkness is never automatically approved.

Your review

Check each group in the viewer, tick it, and write what you saw (8 to 2000 characters).

1. Design and details

Compare the camera, building outline, windows, text, objects and small details.

Pending item: Design: not reviewed
Openings
Thin features

I inspected this group

Observation

2. Lighting and shadows

Confirm that bright areas and dark shadows retain the detail you need.

Pending item:

I inspected this group

Observation

3. Overall finish

Check for unwanted objects, smears, repeated details or other visual defects.

Visual artefacts

I inspected this group

Observation

What was measured

2 for your review
2 not measurable
2 passed
1 informational
4 not applicable

Show the 11 checks

A check passes only when it was measured and is within its limit.

A threshold marked “hypothesis” is an engineering hypothesis, not a validated limit.

ClippingYour review
Newly bright pixels
0
Largest bright area
0
Newly dark pixels
31,487
Largest dark area
10,814
Threshold
900 px bright hypothesis
Newly clipped bright or dark areas compared with the original.
Instrument
Dark areas need review for lost building detail. Intentional dusk shadows may be acceptable.
Visual artefactsYour review
No measurement
Smears, repeated details and other visual defects. Not measured in this release.
Instrument
artifact detection is not implemented in M1; a human reviews

2 not measurable

Could not be measured does not mean unchanged. Look at these yourself before you sign.

OpeningsNot measurable
No measurement
Whether windows and doors kept their count and position. Not measured in this release: check them by eye.
Instrument
No annotated features supplied; requires human review
Thin featuresNot measurable
No measurement
Whether mullions, railings and other thin elements survived. Not measured in this release: check them by eye.
Instrument
No annotated features supplied; requires human review

Supply outline masks with the next project to measure the building outline.

Feature annotations are not part of this release; these details stay with the reviewer.

2 passed

Camera framingPassed
Median move
0.42 px
95th percentile
2.12 px
Matched points
51
Threshold
2.6 px
Whether the camera moved between the original and the result, from matched keypoints.
Instrument
frozen instrument checks.run_all, config_hash ed14867ffa97f341
Edge driftPassed
Edge drift
0.0243 %
Threshold
0.5 % hypothesis
How far matched edges moved, as a share of the image diagonal.
Instrument
framing median inlier displacement 0.42 px over the canvas diagonal 1730.7 px, in percent; pass if at most 0.5

1 informational

Protected pixelsInformation
Not used: this project protects no area.
How many protected pixels are byte-identical. Informational only.
Instrument
per preserve region: fraction of byte-identical pixels and max channel delta; informational, never pass or fail on its own; the plan has no preserve region

4 not applicable

Building outlineNot applicable
No measurement
Whether the building outline is the same, from the supplied masks.
Instrument
plan has no silhouette_mask
Edit spillNot applicable
No measurement
Whether the edit changed pixels outside the permitted region.
Instrument
plan has no edit_mask; there is no region outside which a change counts as spill
Locked coloursNot applicable
No measurement
Whether colour-locked regions kept their colour.
Instrument
plan has no color_locked_regions
Signage textNot applicable
No measurement
Whether protected lettering reads the same. Not measured in this release.
Instrument
no OCR in M1; signage is reviewed by a human

Details

Size: 1440 × 960 px
Last change: Sep 9, 2026, 9:50 PM

Evidence

On your computer this is where the full audit and the unreviewed draft open. For the example, read the sample receipt.

See a sample receipt How the workspace works

Disclosure: AI-assisted or physically rendered image of an architect’s design; not a photograph; not a certification of the built work.

Review: 0 of 3 groups