Product design challenge focused on optimizing Amazon Music for Ai-driven discovery & engagement.
Everything has a sound. What does your world sound like?
Amazon Music already has access to massive distribution. It still needed a stronger reason to be opened, tried, and returned to.
What
A concept feature for Amazon Music that uses a phone camera to turn real-world scenes into personalized playlists.
Role
Product Designer
Scope
Research, concept development, interaction design, ecosystem strategy
Timeline
1 month
Outcome
An end-to-end product concept for contextual music discovery using AI-powered visual interpretation.
Amazon Music did not need more catalog. It needed a discovery experience that better translated real moments into listening decisions.
Problem statement
The friction isn't taste. It's translation.
01
Music discovery today is built around search bars, genre categories, and algorithmic recommendations that recycle the same patterns over time.
02
Most listeners fall into passive habits, replaying familiar playlists rather than exploring something new, because the effort to find fresh music is too high.
03
No major streaming platform takes real-world context into account. Where you are, what you see, and what your moment feels like are completely invisible to the product.
Why this matters
People do not just listen to music. They use it to process feelings, set the tone for experiences, and create memories. A product that understands emotional context can form a much stronger bond with its users.
Users increasingly expect technology to adapt to their environment. From smart home routines to location-based recommendations, passive intelligence is becoming baseline, not a luxury.
When users find music that genuinely matches their moment, they listen longer, return more often, and associate the platform with a feeling of being understood. Discovery quality directly impacts engagement and stickiness.
Point the camera at any moment and let Amazon Music turn what you see into what you hear.
Capture your current scene and generate a playlist that matches the exact visual mood, lighting, and energy around you.
Analyzing scene for Soundtrack moment
Use food visuals to translate flavor cues into music tone, then generate a playlist that feels spicy, sweet, rich, or mellow.
Analyzing scene for Taste to music
Analyze interior atmosphere and ambient cues to generate a playlist that fits your room's current aura.
Analyzing scene for Room vibe
Turn an ordinary daily snapshot into contextual discovery and generate a playful playlist from routine life moments.
Analyzing scene for Everyday life
Scan your pet's scene and generate a mood-matched playlist that feels curious, light, and emotionally sticky.
Analyzing scene for Pet POV
The framework helped structure the work from open exploration through focused delivery, keeping the concept grounded in user needs at every stage.
Phase 01
Explored how people currently find music, where friction exists, and how real-world moments influence listening choices across different platforms.
Phase 02
Synthesized research into key insights and framed the core opportunity: bridging the gap between physical context and digital music discovery.
Phase 03
Explored multiple concept directions including voice-first prompts, mood-selection interfaces, and AI-powered camera input before converging on the strongest path.
Phase 04
Designed the full interaction flow, ecosystem integration strategy, sensory modes, and edge-case handling for the AI Camera concept.
Market gap
01
No streaming platform currently uses visual or environmental input as a discovery mechanism. The camera as an input for music is entirely unclaimed territory.
02
Existing discovery still relies on what users explicitly search, browse, or have already listened to. The model is reactive and backward-looking, not contextual or forward-looking.
03
There is a clear opportunity for a product to own the space between what a user sees and what they hear, creating a new category of real-world-driven music discovery.
Core opportunity
No streaming platform uses real-world visual context as a discovery input. That space is wide open.
The research was designed to answer one question quickly: where could Amazon Music create a discovery behavior competitors were not already owning?
Method 01
Conducted conversations with 12 regular music listeners across ages 20 to 35 to understand how they discover music, where effort shows up, and what breaks the feeling of a moment.
Method 02
Distributed a survey to 85 respondents measuring satisfaction with current discovery methods, frequency of passive listening, and interest in context-aware music features.
Method 03
Analyzed Spotify, Apple Music, Amazon Music, and Tidal across discovery mechanics, personalization depth, and contextual awareness to identify where the market leaves an opening.
Target audiences
Primary
Secondary
User stories
Age 26 · Commuter and passive listener
Scenario
Commuting home after a long day, standing on a crowded train with headphones on. Wants to unwind but does not want to scroll through playlists to find the right vibe.
Need
A playlist that matches the evening wind-down mood instantly without any search or input effort.
Age 22 · Student in a focus flow
Scenario
Studying late in her apartment with warm lamp lighting and a cup of tea. Needs background music that matches the calm atmosphere without breaking concentration.
Need
Music that reads the room and adapts to a low-energy focus state without manual selection.
Spotify and Apple Music already perform well when the user knows they want music. The open territory is helping users discover music from the world around them.
Platform
Discovery
Conventional rows, carousels, and search-based browsing within the app.
Personalization
Leverages listening history and Alexa voice data but lacks a signature discovery feature.
Context-awareness
Has ecosystem signals from Prime, Alexa, and devices but does not use them for contextual music input.
Platform
Discovery
Strong algorithmic playlists like Discover Weekly, Release Radar, and Daily Mixes.
Personalization
Industry-leading taste profiling based on listening history, skips, and saves.
Context-awareness
Light mood language in playlists, but no real-world environmental input layer.
Platform
Discovery
Editorial curation, radio stations, and human-curated playlists.
Personalization
Growing algorithmic personalization but still leans heavily on editorial taste.
Context-awareness
Polished and premium but entirely screen-first and manual in discovery flow.
Platform
Discovery
Artist-driven editorial and high-fidelity audio positioning.
Personalization
Moderate personalization, more focused on audio quality than discovery depth.
Context-awareness
No contextual awareness layer. Discovery is catalog-driven rather than moment-driven.
Key takeaway
No one owns real-world contextual discovery.
Key insights
People often want the result of good music selection without doing the work of describing what they want. The less effort required, the more magical the experience feels.
Design implication
Discovery should begin from low-effort signals instead of requiring explicit search intent.
Most music apps know what users liked before, but not what is happening around them right now. There is a disconnect between past preference data and present-moment relevance.
Design implication
A product that reads moment, environment, and energy can open a new discovery layer no competitor owns.
When users are moving through real life, even small discovery friction feels disproportionately annoying. The fastest path to the right music wins.
Design implication
A single-action concept can feel more premium than a denser, more configurable interface.
Knowing someone likes indie pop is useful. Knowing they are looking at a rainy city street at dusk is more immediate and emotionally resonant.
Design implication
Context should complement taste history, not replace it. The strongest results come from layering both.
Strategic opportunity
Users should not have to translate their mood into search terms. By starting from a visual capture, the product removes the hardest step in music discovery and makes the experience feel effortless.
Amazon already has the ecosystem reach through Prime, Alexa, and Fire devices. AI Camera gives the music product a distinct, memorable reason to be opened, tried, and returned to, driving both acquisition and retention.
The strongest direction wasn't the one that added the most features — it was the one that let users capture a moment without having to describe it.
Explored, then rejected
Voice fit the Amazon ecosystem through Alexa, but it still forced users to describe a feeling in words. That added cognitive effort instead of removing it.
Explored, then rejected
Mood chips and sliders were easier than search, but still made people pause and translate emotion into interface terms. The interaction felt functional, not magical.
Chosen direction
A camera-based input lets users capture a real moment instantly. It felt more novel, more Amazon-native through AWS infrastructure, and more aligned with zero-effort discovery than any alternative.
Instead of asking users to search for the right playlist, the feature starts from a scene they are already in and translates that moment into music.
AI Camera Generated Playlists
The user captures a moment with their phone camera. The system interprets visual context and generates a playlist that matches the energy, mood, and setting of that scene.
AI Analysis
The system reads lighting conditions like warm golden hour, dim ambient spaces, or bright daylight to gauge mood temperature and energy level.
Environmental cues like crowded spaces, quiet corners, outdoor scenery, or cozy interiors signal the kind of energy the music should match.
Objects and settings like food on a table, a desk setup, a park, or a cityscape tell the system what kind of moment the user is in.
5 Sensory Modes
01
Capture your current scene and generate a playlist that matches the exact visual mood, lighting, and energy around you.
02
Use food visuals to translate flavor cues into music tone, then generate a playlist that feels spicy, sweet, rich, or mellow.
03
Analyze interior atmosphere and ambient cues to generate a playlist that fits your room's current aura.
04
Turn an ordinary daily snapshot into contextual discovery and generate a playful playlist from routine life moments.
05
Scan your pet's scene and generate a mood-matched playlist that feels curious, light, and emotionally sticky.
Preset Modes
For moments when users want faster results without pointing the camera, preset modes offer one-tap shortcuts for common contexts like workout, cooking, study, commute, and wind-down. Each preset combines default environmental assumptions with the user's listening history.
End-to-end flow
Step 01
The user points the camera at a moment, whether it is a room, skyline, desk, pet, meal, or commute. The interaction is immediate and requires no typing or browsing.
Step 02
The system reads visual cues like lighting, setting, energy, objects, and atmosphere to infer the contextual signals behind the scene using AWS-powered interpretation.
Step 03
Amazon Music returns a playlist tailored to that specific moment instead of a generic search result. Discovery feels responsive, contextual, and surprising.
Step 04
The generated mix is refined using the listener’s history, saved artists, and skip behavior so the result feels personal rather than random.
AI scan
Generated playlist
AI Camera is the product hook. The larger strategy is acquisition plus retention: bring users in through the ecosystem, then give them a reason to come back inside the music product.
200M+
Prime members globally
55M+
Amazon Music subscribers
<5%
Prime-to-Music conversion rate
Integration points
The gap
Amazon has the reach. It needs the product moment.
Edge cases
When the AI reads a scene incorrectly, the system should offer a quick correction flow with alternative mood or energy options rather than forcing the user to re-capture.
In dark or unclear environments, the system should detect low confidence and offer a fallback path such as a simple mood prompt or recent context suggestion.
The product must make it transparent what is analyzed, what is stored, and how quickly a photo is discarded. Users should have full control over their visual data.
Users need lightweight controls to steer the result after generation. Thumbs up or down, energy sliders, or a regenerate option let them refine without starting over.
Impact
AI Camera gives Amazon Music a stronger first-use moment than another recommendation rail or onboarding survey. The novelty and low friction create a memorable activation experience.
By starting from real-world context instead of search history, the feature surfaces music that feels genuinely fresh and relevant rather than algorithmically familiar.
No competitor owns real-world contextual discovery. This concept gives Amazon Music a clear, defensible story in a crowded streaming market.
The project was most useful when it stopped being about a feature in isolation and started becoming a system-level argument for how Amazon Music could grow.
Music discovery is not just a functional task. It is an emotional handoff between life and interface. The strongest concept was the one that honored that relationship.
The concept worked best when AI handled the heavy lifting but still gave users light ways to correct or refine the result. Full automation without control felt unreliable.
The most valuable part of the idea was not just the feature itself. It was how AI Camera could connect Prime, Alexa, AWS, and Amazon Music into one acquisition-to-retention story.
