Samsung One UI 7.1 → 9.0 · AI Imaging · Lead Designer

Portrait
Studio

Designing AI identity at scale — where cultural empathy, prompt engineering, and user research converge into a feature used by millions.

AI / GenAI Prompt Engineering UX Research Employee of the Year 2024 One UI 7–9

Portrait Studio is Samsung's AI portrait & style engine — shipped across Galaxy Flip, Fold, and S-series. I led it from zero: the research, the prompt architecture, the style strategy, the cultural frameworks, and the team that scaled it globally.

Role

Lead Product Designer
Multimedia Team

Timeline

One UI 7.1 → 9.0
2024 – Ongoing

Scope

Research · Strategy
Prompt Eng. · UX

Style Generation Engine — Real-time portrait transformation · One UI 9.0

The AI was
erasing who you are.

In 2024, prompt engineering for large vision models was uncharted territory. The first thing we discovered wasn't a UX problem — it was an identity problem.

Indian subjects rendered with wrong hairstyles — the 1–3B hair spectrum completely ignored
Turbans vanished entirely for Sikh men
Bindis & sindoor omitted from portraits
Nose rings placed on the wrong side of the face
Lips thickened, facial structures altered — every face mapped to a Western default

A designer's feedback shaped the AI model itself — not just the interface.

I raised these concerns and directly influenced Google's team in Mountain View to retrain Imagen 1/2. What started as a UX observation became an input to the model's training pipeline.

15+

Forensic papers
studied

Facial parameters
mapped

Key facial
saliency areas

Cultural markers — good to have features

Cultural markers — facial saliency mapping

Cultural markers — must have constraints

Phase 01 — Foundation

Building the
practice from zero.

Imagen 1/2 · Cultural markers · 60/40 Framework · Prompt Architecture · Seoul secondment · Influencing Google Mountain View

The 60/40 Framework

Every prompt I engineered is governed by this principle — 60% accuracy, 40% creativity.

Accuracy

Cultural markers preservation
Facial structure integrity
Hair type accuracy (1–3B spectrum)
Beard shape & ornament placement
Eye direction & expression fidelity
Nose ring & bindi positioning

Creativity

Material & medium experimentation
Background variety & depth
Lighting conditions & mood shift
Texture, grain & finish quality
Color palette transformation
Contextual emotional enhancement

Accuracy preserves what makes you you. Creativity adds what makes you wow.

✦ Positive Prompt — 10 Slots

01Style Description

02Orientation of Subject

03Expression of Subject

04Hairstyle

05Outfit

06Environment

07Background

08Lighting Conditions

09Medium & Textures

10Colour & Mood

✕ Negative Prompt — 10 Guardrails

01No harsh/flat lighting

02No wrong medium bleed

03No clashing tones

04No anatomy distortions / extra limbs

05No background clutter

06No expression exaggeration

07No style bleed from adjacent styles

08No watermarks / logos

09No low-res / blur

10NSFW · distorted facial ID guardrails

Every word is a design decision. At 480 tokens, there's no room for ambiguity.

Prompt Anatomy in Practice — Clay Style Example Click any section to explore

01 Style Description

3D realistic image in a clay material style, with smooth rounded forms, visible fingerprint-like textures on surfaces, miniature diorama proportions

02 Orientation

Looking away from the camera at a ¾ angle, slight head tilt conveying natural warmth

03 Expression

Minimalistic facial features with subtle expression, gentle closed-lip smile

04 Hairstyle

Styled in a simple bob or tied back with gentle flow lines, clay-sculpted texture

05 Outfit

Plain outfit like a clay-detailed dress or top, soft fabric folds sculpted into material

06 Environment

Handmade, warm, approachable atmosphere — like a clay workshop studio

07 Background

Solid pastel colour (cream or soft mint), clean and uncluttered

08 Lighting

Soft ambient lighting highlighting sculpted curves, diffused shadows

09 Medium & Textures

Clean and crafted feel, soft matte textures, visible hand-sculpted imperfections

10 Colour & Mood

Natural clay tones — warm beige, terracotta, blush pink, ochre accents

Full Assembled Prompt ~480 tokens

3D realistic image in a clay material style, with smooth rounded forms, visible fingerprint-like textures on surfaces, miniature diorama proportions. Looking away from the camera at a ¾ angle, slight head tilt conveying natural warmth. Minimalistic facial features with subtle expression, gentle closed-lip smile. Styled in a simple bob or tied back with gentle flow lines, clay-sculpted texture. Plain outfit like a clay-detailed dress or top, soft fabric folds sculpted into material. Handmade, warm, approachable atmosphere. Solid pastel colour background (cream or soft mint), clean and uncluttered. Soft ambient lighting highlighting sculpted curves, diffused shadows. Clean and crafted feel, soft matte textures, visible hand-sculpted imperfections. Natural clay tones — warm beige, terracotta, blush pink, ochre accents.

Click sections on the left to highlight Clear highlight

✕ Negative Guardrails 10 constraints

No harsh shadows No metallic/plastic No synthetic feel No neon colours No uncanny valley No anime/2D bleed No watermarks No blur/low-res No NSFW No distorted ID

Seoul Secondment

The framework that crossed borders.

The 60/40 framework and prompt architecture were validated during a secondment at Samsung HQ in Seoul. This foundation became the system used to expand into more styles across subsequent releases.

Testing was a critical blocker — security restrictions at HQ and ORCs meant server-side validation could only happen in Korea. The framework had to be robust enough to guide remote iteration without direct access.

Seoul HQ Security Constraints Remote Iteration Framework Validated

What We Established

60/40 accuracy-creativity ratio as the north star for every prompt

480-token modular prompt anatomy — scalable to any new style

Cultural markers as non-negotiable accuracy constraints

→ Returned to India with a validated system. The Seoul team trusted our research-backed approach.

Phase 02 — Expansion

From pets to trends
to 10 shipped styles.

Pet portrait research · Ghibli moment · 78-user study · 40 → 10 style selection · Cross-border leadership

Pet Portraits

Understanding how people see their pets.

When we came back from Seoul with the framework validated, the team was given an exercise: study which styles could be introduced for pet portraits. I had recently become a pet parent — 3 cats — and that personal connection shaped the research lens.

We pulled extensive research on how people associate with their pets and how it fundamentally differs from people portraits. Pets are family, not subjects — and the emotional language users used confirmed this.

Emotional value of pet's name rated 5/5 55.6%

Users rank 'Expression' most important 57.1%

Users rank 'Pose' as most important 85.7%

User Relationship Language

"biological daughter" "family member" "my sister" "my whole world"

Research Context

How people photograph and relate to their pets

Four Macro-Trends

During pet portrait research, the Ghibli trend exploded globally. Models were getting stronger, and the project expanded beyond pet portraits to entire photo generation — we needed to suggest more styles with careful consideration as our parameters were evolving and our USP was in question. Cross-referenced Adobe Creative Trends 2024, social AI reports, and 5 niche apps.

Retro Nostalgia

Y2K revival, high contrast, warm grain. Vintage, Funky, Retro-Fusion substyles. ↗ Adobe Creative Trends

2.1B

TikTok views

Wonder & Joy

Cutesy, Ghibli-adjacent, whimsical. Expressive, Pastel substyles. The Ghibli moment.

150M+

ChatGPT Ghibli users

Future Punk

Neon glow, cyberpunk, volumetric lighting. Tech Graphics, Glow substyles.

$307B

Gaming market backing

Material Play

Felt, clay, handcrafted, plush textures. Soothing, Tactile substyles.

$19.8B

Plush toy market

Cross-referenced → Adobe Creative Trends 2024 Social AI Reports 5 Niche Competitor Apps TikTok/Instagram Trend Data Regional Popularity Mapping

The UX Glance Study

As models grew stronger, the core question became user intent and use cases. We mapped them through a structured user study. 78 users · 4-step methodology · 65/35 M/F · 80% millennials.

Users tested across people & pet portrait styles

12+

Use-case surfaces mapped inside & outside ecosystem

User intents identified driving style & surface decisions

The Discovery: People ≠ Pets

Two completely different behavioural profiles emerged from the same product.

Utility-Inclined

People Portraits

45 of 78 users voted Cinematic as their top style

Studio Glow Light Cinematic

Primary surfaces

WhatsApp DP LinkedIn Contact Photos Professional Branding

"What if I can make this directly as my WhatsApp DP?"

Mapping 8 Intents × 12 Surfaces

Many got commercialised into devices — cross-team collaborations across mobile divisions.

User Intents

Self Expression

Community Sharing

Partial Anonymization

Adding Uniqueness

Personal Branding

Re-Use / Re-New

Curation

Improve Quality

Use-Case Surfaces — Inside & Outside Ecosystem

Photo Widgets

AR Filters

Playlist Covers

Merch / Gifts

Sticker Sets

Gallery Albums

Chat Backgrounds

Contact Profiles

Flip Cover Screens

Watch Faces

Game Thumbnails

Social Posts

Cross-App Entry Points & Deep Links

Portrait Studio as a capability surfaced across the Samsung ecosystem — every entry point designed for zero-friction access.

Ecosystem Reach

Gallery · Contacts · Messages · Camera
DeX · Tablet · Watch · Z Flip Cover

📱

Gallery Deep-link

Direct entry from Gallery long-press — zero friction for power users

🔗

Contacts & Messages

Apply styles as contact photos or share via Messages directly

🖥️

DeX & Tablet

Sidebar panel layout — same mental model, sized for large displays

⌚

Watch & Flip Cover

Generated portraits as watch face art and Z Flip cover backgrounds

40 Candidates → 10 Shipped

20 portrait + 20 pet candidates filtered through 5 criteria.

Proposed

→

Finalists

→

Shipped

Global Appeal Works equally in Korea, India, US, Brazil

Visual Complexity AI renders it consistently at scale

BG Agnostic Works regardless of photo setting

Draw-assist Free No reliance on manual user masking

Visually Distinct Clearly different from every existing style

Cross-Border Leadership

Seoul ↔ Bangalore.
Led with data, not opinion.

Seoul proposed Studio, K-pop Star, Webtoon. I advocated for Fisheye, Clay Figurine, Sneaky Peek, Bubble Time, Felt Toy, Peekaboo. The result: a hybrid set that's stronger for a global audience.

Seoul Secondment built the cross-cultural trust that made this collaboration possible. Server testing was outsourced to Korea due to India security restrictions — led entirely with research-backed advocacy.

Seoul Proposed

StudioK-pop StarWebtoon

I Advocated

FisheyeClay FigurineSneaky PeekBubble TimeFelt ToyPeekaboo

→ Hybrid set shipped. Stronger for global audiences.

Product Experience

The interface
that shipped.

Real screens. Real UX decisions. From style tray to final output.

Portrait Restyle — End-to-End Flow

Select photo → Choose style → Generate → Result

Select
Photo

Style
Tray

Generate
Portrait

Final
Result

Create Tab — 2-Depth AI feature discoverability

Generative Edit — Erase Object-aware AI erasing

Result Card System Before/after comparison

UX Architecture

How the screen flow maps to the generation pipeline.

The style tray layout directly influences which prompts get triggered. Scroll depth in the tray = user exploration = prompt diversity. UI architecture and AI behaviour are inseparable.

01 Horizontal scroll depth maps to prompt diversity — deeper tray exploration = more creative output variety

02 Card result system encodes before/after comparison — reinforces AI value perception with zero cognitive overhead

03 480-token constraint forces prompt economy — every parameter slot must justify its inclusion via user impact data

Style Output — People & Pets

Final shipped styles across portrait and pet portrait modes.

Portrait Styles

Pet Styles

Desktop & Tablet UX

Designing for Large Screens — Not Just Responsive, Rethought

Scaling to tablet and desktop wasn't about making the mobile layout wider — it was about rethinking the interaction model for how people use large screens. Different posture, different gaze, different expectations.

Gaze & Layout Strategy

F-pattern scanning on desktop vs. thumb-zone on mobile — control panel repositioned to left sidebar
Preview canvas expanded to 65% of viewport — image-first interaction
Style selection → persistent sidebar instead of bottom sheet
Eliminates cover/reveal pattern that wastes desktop real estate

Interaction Model Shifts

Hover previews on style thumbnails — replaces tap-wait-generate cycle
Keyboard shortcuts & drag-drop — Ctrl+Z, drag to compare
Multi-panel layout for simultaneous comparison
Side-by-side original + styled output — no modal transitions

Galaxy Tab S9 Samsung DeX Sidebar Panel UX F-pattern Layout Hover Preview Multi-panel Compare

Features That Evolved With the Project

Not everything ships at v1. Some features emerge from usage patterns.

One UI 7.1 — Launch

Core Style Engine

10 portrait + 7 pet styles. 480-token prompt system. Basic single-shot generation with result card.

One UI 8.0 — Evolution

History of Edits

Users wanted to revisit past generations. We added a persistent edit history — every AI generation saved, browsable, reshareable.

One UI 9.0 — Innovation

Recursive Editing

Continue editing from any previously generated image. Users can layer styles, iterate, and diverge from the same source — turning a single photo into infinite creative branches.

Device Reach & Real-World Impact

Portrait Studio ships across Samsung's flagship and foldable ecosystem — reaching hundreds of millions.

37M+

Galaxy S24 series units shipped globally in 2024

5.2M

Galaxy Z Fold + Flip foldables sold in 2024

1B+

Active Galaxy smartphones worldwide

241M

Samsung smartphones shipped in 2025 alone

Galaxy S24/S25/S26 Galaxy Z Fold 6/7/8 Galaxy Z Flip 6/7/8 Galaxy Tab S series Samsung DeX Galaxy Watch Faces

Photo Assist — AI Adoption

30% of Samsung users rank Photo Assist as the most beneficial Galaxy AI feature

Galaxy AI on 200M+ devices — targeting 400M+ by end 2025

84% of Galaxy S24 owners used AI features within the past month

Photo Editor → Photo Assist Evolution

Manual editing → AI-first — evolved from Photo Editor to generative capabilities with Galaxy S24

70% of S25 users actively engage with AI features daily

Fastest adoption ever — Galaxy AI outpaced every previous Samsung service in time-to-scale

Portrait Studio in Action

Live interaction demos showing the AI generation pipeline and UX flow.

Photo Assist — Live Demo Generative AI photo editing flow on Galaxy device

Galaxy AI — Key Visual Samsung's AI imaging ecosystem hero animation

Style Generation Engine Real-time portrait style transformation pipeline

Phase 03 — Scale & Future

An architecture
built to bend.

Open-ended prompts · Seasonal styles · 2026 trends integration · Module redesign for infinite extensibility

The Architecture Shift

Adding styles every year wasn't sustainable. We redesigned the entire module.

Before

Fixed style catalogue
Manual prompt engineering per style
New styles = new dev cycle
Scaling bottleneck
Reactive to trends

→

After

Open-ended prompt section
Users create any style
Modular architecture for scale
New styles = config update
Proactive trend responsiveness

Pet styles in One UI 9

Portrait styles in One UI 9

Seasonal styles launched

The architecture change led to entire module redesign — future-proofing for any trending style, not just the ones we can predict.

Phase 01

Foundation

→ Cultural markers framework
→ Influenced Google's model
→ 60/40 accuracy/creativity
→ 480-token prompt system
→ Seoul framework validation

Phase 02

Expansion

→ Pet portrait research
→ 4 macro trend analysis
→ 78-user Glance study
→ People ≠ Pets discovery
→ 40 → 10 + Cross-border

Phase 03

Scale

→ Open-ended prompt section
→ Architecture redesign
→ Seasonal styles (5)
→ 2026 trends integration
→ Multimedia team vertical

What I Brought to the Project

Five capabilities. One project. Everything I know how to do.

🌍

Ethical AI Infrastructure

Influenced Google's model training. 60/40 framework + cultural audits → responsible AI.

⚙️

Prompt Engineering as Design

Built this practice from zero in 2024. 480-token modular anatomy, open-ended architecture.

🔬

User-Centered Segmentation

People ≠ pets. Enterprise ≠ consumer. 78 users, 12+ surfaces, data-driven decisions.

🚀

0-to-1 Team Building

First person on Multimedia team. Built processes from scratch. Seoul cross-cultural leadership.

🎨

Visual Arts + Systems

BFA deconstructs art movements into prompt parameters. 5+ years builds systems to scale them.

Samsung Unpacked Keynote

Galaxy AI — shown to the world on the biggest stage.

What I Would Do Differently

If I were starting this project over with what I know now.

Prompt Versioning from Day 1

Evolved prompts across 3 model generations (Imagen 1 → 2 → 4) — tracking was manual
Would build a version-controlled prompt repository from the start
Treat prompts as code — diffs, rollbacks, A/B testing baked in

Earlier Cross-Regional Testing

Security constraints meant server testing only happened in Seoul
Would push harder for a sandboxed remote testing pipeline earlier
3-4 week feedback loops slowed iteration during the critical foundation phase

User-Generated Style Input

Open-ended prompt (Phase 3) was the right direction — but came too late
Would introduce mood boards and reference uploads earlier
Research hinted at user intent signals we didn't act on fast enough

Quantitative Cultural Benchmarks

60/40 framework was qualitative by design — adaptable but hard to measure
Would establish quantifiable cultural accuracy metrics (e.g., facial feature preservation score)
Enable automated regression testing as models updated

More from the
Galaxy ecosystem.

5+ years designing AI imaging experiences across 6 form factors, One UI 5.0 → 9.0, and multiple cross-team collaborations.

Photo Assist

Gen AI Photo Editor · Employee of Year

Galaxy AI

Generative editing at scale

Camera & Fun Mode

Samsung × Snapchat · 2B+ activations

Magnifier

Accessibility · One UI · Multi-device

2B+

Samsung × Snapchat Fun Mode activations

Form factors designed across Galaxy ecosystem

Years of AI imaging experience at Samsung Research

02 — Camera & Fun Mode · Samsung Research

Part A — Usability

Camera &
Fun Mode

Improving the core usability of Samsung's AR lens experience — uncovering 20+ pain points through research with Gen Z and Millennial users, and defining actionable design directions for Fun Mode.

AR / Fun Mode Samsung × Snapchat Usability Research Heuristic Evaluation One UI Camera

2B+

Fun Mode activations
worldwide

20+

Pain points uncovered
in research

Research participants
Gen Z + Millennials

100%

Failure rate on key
feature discovery task

What is
Fun Mode?

Fun Mode is the Samsung Camera's AR lens experience — a partnership with Snapchat that brings real-time augmented reality filters directly into the native camera app on Galaxy devices.

The experience features 7 AR lenses always available, 4 Special Wide & Ultra-Wide lenses refreshed quarterly, and 1 daily featured lens updated every day — keeping the experience fresh and engaging across millions of devices.

7 core AR lenses — always available

4 Special Wide/Ultra-Wide lenses — quarterly refresh

1 daily featured lens — updated every 24 hours

Project
Brief

The project spanned two parallel tracks — improving the core usability of Fun Mode, and defining a scalable design framework for sponsored AR lenses entering the Galaxy Camera ecosystem.

Then

Ownership transfer — Fun Mode inherited from a different team with no documented UX rationale

Now

Market response — understanding current device support, user feedback, and feature gaps

Reimagined experience — defining the future direction with mid-term and long-term roadmap

Phase 1 — Usability

Secondary
Research

The research phase began with a structured analysis of existing signals — heuristic evaluation, voice-of-customer data, and competitive benchmarking — before moving into primary fieldwork.

Heuristic Analysis — Top Findings Nielsen's 10 heuristics applied to Fun Mode interface

Systematic heuristic evaluation across Nielsen's 10 usability principles revealed critical violations in visibility of system status, match between system and real world, and user control — validating the need for a ground-up interaction audit.

Voice of Customer (VOC) Analysis Play Store reviews, support tickets, NPS verbatim analysis

VOC data mining across app store reviews, customer support tickets, and NPS survey verbatim surfaced recurring themes: users were confused by the lens discovery model, disappointed by AR element fidelity, and unaware of key features like video capture.

Competitive Benchmark Study Snapchat · Instagram · TikTok · Apple PhotoBooth

Cross-platform competitive analysis across Snapchat, Instagram Reels, TikTok effects, and Apple's PhotoBooth mapped best-in-class patterns for lens preview, content refresh cadence, and sponsored content integration models.

Method 1

Heuristic Eval

Expert review using Nielsen's 10 usability heuristics across all Fun Mode touchpoints

Method 2

VOC Mining

Thematic analysis of 1000+ user reviews, NPS verbatim, and support ticket clustering

Method 3

Competitive Audit

Feature parity matrix and UX pattern mapping across 4 competing AR platforms

Method 4

Moderated Testing

8 participants, think-aloud protocol, task-completion scoring, affinity mapping

Primary Research

8 Participants.
20+ Pain Points.

Moderated usability sessions were conducted with 8 participants across Gen Z and Millennial demographics — the core Fun Mode user base. Sessions uncovered over 20 distinct pain points spanning discoverability, content quality, and feature awareness.

Primary research summary — 8 participants, Gen Z + Millennials, 20+ pain points

Top 5
Pain Points

Five critical usability failures emerged from primary research — each backed by observed behaviour and quantitative task-completion data.

Discoverability

Ultra-Wide Lens Hidden

8 out of 8 participants failed to discover the ultra-wide AR lens — a 100% failure rate on a core differentiating feature.

Pain point 1 — ultra-wide lens discoverability

Feature Awareness

Video Capture Unknown

Only 5.64% of captured media was video — users were unaware Fun Mode supported video capture at all.

Pain point 2 — video capture unawareness

Preview Quality

Static Thumbnails Mislead

8 out of 8 participants failed to understand a lens from its static thumbnail — expecting animated previews similar to Snapchat and Instagram.

Pain point 3 — static thumbnail previews

Content Quality

Low-Quality AR Elements

AR lens elements perceived as low-fidelity — users compared unfavourably to competitor experiences, reducing motivation to use or share.

Personalisation

No Way to Save Favourites

Users had no mechanism to save or quickly return to favourite lenses — leading to frustration when trying to re-find preferred content.

Ideation &
Concept Directions

Structured ideation translated research findings into actionable design directions through HMW exercises, a cross-functional design critique, and three distinct concept frameworks.

HMW Ideation Exercise

Design Critique Session

UX Concept Directions — 3 Goal Framework

From 20+ usability pain points to three structured concept directions — primary research with 8 participants revealed that Fun Mode's core experience needed a fundamental interaction redesign before scaling to sponsored content.

Measurable Impact

How the UX work moved the needle

2.5B+

Total Fun Mode
activations by Sept 2022

100%

Key discovery issues
addressed in next release

3×

Concept directions
advancing to One UI roadmap

Markets reached from
India to Brazil

Positive VOC shift — Post-release user feedback on the Play Store showed marked improvement in lens discoverability and feature awareness — the two highest-priority pain points from research

Video capture ratio improved — Video share of total media creation rose from 5.64% to 12.3% after shutter redesign addressed the visibility and affordance gaps

Samsung × Snapchat partnership expanded — from initial launch on Galaxy A52/A72 to a global rollout across A, F, and M series devices

Part B — Sponsored Lenses

The Magnifier project was a cross-studio endeavour spanning SRIB UX and Seoul UX, with engineering collaboration from EP Group 1 and SRIN RnD.

UX Studios

SRIB UX · Seoul UX

Engineering

EP Group 1 · SRIN RnD

Evolution
Timeline

Magnifier's journey from inception through to premium overhaul — spanning core enhancements, competitive differentiation, and a drive toward inclusive design at scale.

Timeline: Project Inception to One UI UX Enhancements

Basic Enhancement

Foundational usability improvements across core Magnifier features

Competitive Advantage

Benchmarked against Apple and Google to define a differentiated position

Empower Inclusivity

Designing for low-vision users across every Galaxy device and form factor

Competitive
Landscape

A systematic review of Apple Magnifier and Google Lens uncovered what users praise — and what Samsung could do better. This analysis directly shaped the enhancement roadmap.

🍎

Apple Magnifier — User Reviews

🟢

Google Lens — User Reviews

Apple Magnifier — Strengths

Activity, Display, Settings — clean 3-tab UI with colour filter presets and reader mode

Netizen praise — "I use Magnifier all the time! It's wonderful!" — users love simplicity and reliability

Detection mode — dedicated text detection with high-contrast readability overlay

Google Lens — Strengths

Real-time translation — overlays translated text on signs and menus with AR positioning

Colour filter customisation — full filter palette and more settings for accessibility users

User love — "A wonderful app for 62-year old eyes! I use it to read warranties on tools" — real utility

Future
Directions

Three strategic directions define the long-term roadmap — from a fresh UX overhaul to AI-powered assistance and deeper system integration across One UI.

Direction 1

New Feature & Fresh UX

Core interaction overhaul and new accessibility-first feature set

Direction 2 · One UI 9.5

AI Assistance

Intelligent scene understanding and real-time guidance powered by on-device AI

Direction 3 · One UI 12.5

System Integration

Deep integration across Galaxy ecosystem — lock screen, quick panel, and Bixby

Large Device
Implementation

Magnifier was designed to scale beyond the phone — with tailored experiences for the Galaxy Z Fold (cover and unfolded states) and Tab S11 (capture and post-capture landscape modes).

Next Proposals

Voice Assist
& Discoverability

Two forward-looking proposals define the next generation of Magnifier: a Voice Guided Assist feature using LVM/VQM proof-of-concept to provide real-time verbal descriptions, and a targeted effort to dramatically improve feature discoverability across the One UI surface.

Voice Guided Assist

LVM/VQM PoC — AI-powered real-time scene narration for low-vision users

Improving Discoverability

Surfacing Magnifier capabilities where users need them most across the OS

From foundational zoom controls to AI-powered voice guidance — Magnifier evolved from a utility into a best-in-class accessibility platform built for every Galaxy device.

EAA

European Accessibility Act

Designing for
EAA Compliance

Magnifier required deep collaboration with Samsung's Accessibility team to ensure full compliance with the European Accessibility Act (EAA) — going beyond minimum standards to create a truly inclusive experience.

Visual Accessibility

Contrast ratios — all UI elements meet WCAG AAA (7:1) for critical controls, AA (4.5:1) minimum across all surfaces

Reduced transparency mode — background blur and overlay opacity adjustable for users with low contrast sensitivity

Iconography audit — every icon tested for recognisability at 200% zoom with alternative text labels

Interaction Accessibility

Touch target areas — minimum 48×48dp for all interactive elements, 56dp for primary actions

TalkBack integration — full screen reader support with semantic content descriptions for every magnified view

Switch Access & Voice Control — full operability without touch for users with motor impairments

7:1

WCAG AAA contrast
for critical controls

48dp

Minimum touch
target size

100%

TalkBack coverage
across all screens

EAA

European Accessibility
Act compliant

Beyond Mobile

Hardware Vision:
Smart Glasses

The most compelling next frontier for Magnifier is wearable AR — specifically Samsung's upcoming smart glasses ecosystem. Magnifier's camera-based zoom and enhancement pipeline translates directly to heads-up display (HUD) rendering for low-vision users, enabling real-time environmental augmentation without requiring users to hold a device.

Glasses × Magnifier — Use Cases

👓

Real-Time Environmental Magnification

On-device neural processing magnifies text on menus, signs, and labels — projected directly onto the AR display at user-preferred zoom levels

🎙️

Voice-Guided Navigation

LVM-powered scene understanding paired with spatial audio cues — guiding low-vision users through complex environments hands-free

🔍

Adaptive Contrast Enhancement

Dynamic colour filter overlays that adjust based on ambient light conditions — improving readability for users with colour vision deficiency

Technical Architecture — XR Pipeline

On-device NPU inference — Magnifier's image processing pipeline runs entirely on Samsung's Exynos NPU, enabling sub-20ms latency for real-time HUD rendering without cloud dependency

Spatial UI paradigm shift — Translating Magnifier's 2D touch-based controls (pinch-to-zoom, pan, filter toggle) into gaze-tracking, head-tilt, and voice command interaction models for XR

Multi-modal accessibility fusion — Combining visual enhancement (magnification), auditory feedback (TalkBack integration), and haptic cues (wrist-based alerts) for comprehensive accessible XR

Device Reach
& Accessibility at Scale

Magnifier ships across every Galaxy device — from the entry-level Galaxy A14 (34M+ units) to flagship foldables — with model-specific UX adaptations for each form factor. Designing for 1 in 4 adults globally who live with some form of disability.

1B+

Active Samsung Galaxy
devices with Magnifier

34M+

Galaxy A14 units sold
— most accessible smartphone

300M

People with colour vision
deficiency worldwide

Form factors with
model-specific UX

Galaxy A14 / A16 / A36 Galaxy S24 / S25 / S26 Galaxy Z Fold 6/7/8 Galaxy Z Flip 6/7/8 Galaxy Tab S10/S11 Samsung Smart Glasses (Vision)

Model-Specific Design Adaptations

Phone (Bar)

Single-hand zoom gesture with haptic feedback anchoring · Bottom-sheet control panel optimised for reachability · Portrait-locked magnification with auto-rotation for document scanning

Foldable (Fold/Flip)

Cover screen quick-launch for on-the-go magnification · Unfolded mode: split-screen with live view + captured frame side-by-side · Flex mode: tabletop stabilisation for extended reading sessions

Tablet (Tab S)

Landscape-first capture mode with S Pen precision controls · Post-capture annotation layer for marking key text · Multi-window compatibility for side-by-side magnification + notes

Final
Designs

Shipped Magnifier screens across the Galaxy ecosystem — a consistent accessibility experience that adapts to every device and context.

Main View

Camera-based magnification with flashlight and filter controls

Live Capture

Real-time magnification with zoom controls

Multi Capture

Batch capture and review for detailed inspection

Post-Capture

Frozen frame for detailed inspection

DesigningAI at Scale.

Portrait Studio

The AI waserasing who you are.

Building thepractice from zero.

The 60/40 Framework

The framework that crossed borders.

From pets to trendsto 10 shipped styles.

Understanding how people see their pets.

Four Macro-Trends

The UX Glance Study

The Discovery: People ≠ Pets

Mapping 8 Intents × 12 Surfaces

Cross-App Entry Points & Deep Links

40 Candidates → 10 Shipped

Seoul ↔ Bangalore.Led with data, not opinion.

The interfacethat shipped.

Portrait Restyle — End-to-End Flow

Style Output — People & Pets

Designing for Large Screens — Not Just Responsive, Rethought

Features That Evolved With the Project

Device Reach & Real-World Impact

Portrait Studio in Action

An architecturebuilt to bend.

The Architecture Shift

What I Brought to the Project

What I Would Do Differently

More from theGalaxy ecosystem.

Camera &Fun Mode

What isFun Mode?

ProjectBrief

SecondaryResearch

8 Participants.20+ Pain Points.

Top 5Pain Points

Ideation &Concept Directions

How the UX work moved the needle

SponsoredLens Framework

Research &Lens Audit

UX Principles in Detail

Content Guidelines& Monetisation

UX × PM × Business — Driving QBR Outcomes

Device Reach& Scale

Technical Architecture & Design Systems

Magnifier

Collaboration& Scope

EvolutionTimeline

CompetitiveLandscape

FutureDirections

Large DeviceImplementation

Voice Assist& Discoverability

Designing forEAA Compliance

Hardware Vision:Smart Glasses

Device Reach& Accessibility at Scale

Model-Specific Design Adaptations

FinalDesigns

Designing
AI at Scale.

Portrait
Studio

The AI was
erasing who you are.

Building the
practice from zero.

From pets to trends
to 10 shipped styles.

Seoul ↔ Bangalore.
Led with data, not opinion.

The interface
that shipped.

An architecture
built to bend.

More from the
Galaxy ecosystem.

Camera &
Fun Mode

What is
Fun Mode?

Project
Brief

Secondary
Research

8 Participants.
20+ Pain Points.

Top 5
Pain Points

Ideation &
Concept Directions

Sponsored
Lens Framework

Research &
Lens Audit

Content Guidelines
& Monetisation

Device Reach
& Scale

Collaboration
& Scope

Evolution
Timeline

Competitive
Landscape

Future
Directions

Large Device
Implementation

Voice Assist
& Discoverability

Designing for
EAA Compliance

Hardware Vision:
Smart Glasses

Device Reach
& Accessibility at Scale

Final
Designs