Amplifier Showcase

Amplifier
UX Analyzer

Computer vision for UI understanding. Extracts colors, layout, elements, and text from screenshots so AI agents can see interfaces.

✓ Active

Built by ramparte (Sam Schillace) · February 2, 2026

The Problem

AI agents can't see
user interfaces.

🖥️

Screenshots are opaque

A screenshot is just a grid of pixels to an AI. No structure, no semantics, no understanding of what's a button versus a toolbar.

🎨

Design intent is lost

Color palettes, spacing hierarchies, and layout relationships are invisible without structured extraction from the raw image.

🔄

Recreation requires understanding

Building pixel-perfect UI recreations demands structured knowledge — regions, elements, text positions, and color values — not just a flat image.

The Solution

A four-stage vision pipeline

Screenshot in, structured JSON out.

📸

Screenshot

PNG / JPEG input

→

🎨

Color Extract

k-means clustering

→

📐

Layout Detect

Contour analysis

→

🔤

Text + OCR

EasyOCR extraction

OpenCV NumPy scikit-learn EasyOCR Pillow

Capabilities

What it extracts

🎨

Color Palette

k-means clustering identifies dominant colors from the image. Returns RGB values, hex codes, and relative frequency for each color.

📐

Layout Regions

Contour analysis detects structural areas — toolbars, content zones, sidebars, status bars — with bounding boxes and hierarchy.

🔘

UI Elements

Fine-grained contour detection identifies individual buttons, controls, and interactive elements within detected regions.

🔤

Text Extraction

EasyOCR reads all visible text with confidence scores and bounding box positions, enabling precise text placement in recreations.

🖼️

Visual Output

Generates annotated screenshots showing detected regions overlaid on the original image, for visual verification of extraction quality.

📊

JSON Output

All extracted data is returned as structured JSON — ready for programmatic consumption by AI agents or downstream tools.

Impact

Giving AI agents
spatial understanding

Instead of describing a UI in words and hoping an agent gets it right, UX Analyzer provides the structured data an agent needs: exact colors, precise regions, identified elements, and extracted text — all from a single screenshot.

Before

"Make it look like this screenshot" — agents guess at colors, approximate layout, and miss text content entirely.

After

Agents receive exact hex values, bounding coordinates, element counts, and OCR text with confidence scores.

Result

A foundation for pixel-perfect UI recreation — structured data replaces guesswork.

Development

Built in a single day

415

Lines in main script

2

Commits total

1

Day of development

6

Test files included

Sole contributor: ramparte (Sam Schillace) · Both commits on February 2, 2026
First commit 12:07 PST → second commit 16:29 PST · Complete tool with tests in under 5 hours

Transparency

Sources & Methodology

Data as of: February 20, 2026

Feature status: Active — repo exists with code present, no updates since February 3, 2026

Primary contributor: ramparte (Sam Schillace) — 100% of commits

Research performed:

GitHub API query for repo metadata (created, updated, description)
Local filesystem inspection of ~/dev/ANext/amplifier-ux-analyzer
git log --oneline → 2 commits confirmed
git log --format → commit timestamps: Feb 2 12:07 PST, Feb 2 16:29 PST
wc -l ux-analyzer.py → 415 lines
README.md reviewed for feature descriptions and dependency list
Test file count verified via filesystem: 6 test files

Known data gaps:

No performance benchmarks found in repo — no timing claims are made in this deck
No browser bundle PR was found (a prior version of this deck mentioned one — unverified, removed)
No GPU acceleration benchmarks — prior "10-50x faster" claim removed as unsourced
No GitHub description was set on the repository

Numbers in this deck: All figures (415 lines, 2 commits, 1 day, 6 test files) come directly from git log and filesystem inspection. No estimates or projections are presented as facts.

Get Started

See your UI
through new eyes.

Screenshot in, structured understanding out. Give your AI agents the vision they need to work with user interfaces.

ramparte/amplifier-ux-analyzer

Built with Amplifier · February 2026

AmplifierUX Analyzer

AI agents can't seeuser interfaces.