Dopamine Level AI Analyzer

Inspiration

It started with a question every parent in the digital age should ask: What is my child feeling while watching?

As an animation creator and a dad, I realized that calm, poetic stories give children a completely different experience than fast, overstimulating nursery videos. Yet, we had no shared language to describe that difference. We have nutrition labels for food – but nothing for the content that feeds our children's minds for hours each day.

When I discovered Gemini’s multimodal capabilities, I saw an opportunity to change that. If an AI can perceive rhythm, emotion, and sensory load, maybe it could help us measure the impact of what we watch. That’s how Dopamine Level AI Analyzer. (D-Level AI Analyzer) was born – a bridge between human empathy and machine perception.

What it does

In one line: D-Level AI Analyzer is a Chrome extension that rates YouTube videos by their neural stimulation profile – like a digital “nutrition label” for attention.

It provides a simple, understandable D-Level score (0–100) that reflects how intensely a video stimulates the viewer’s nervous system. The tool operates in a hybrid AI mode:

On-device Promise Analysis (Gemini Nano): Privately evaluates a video’s title, thumbnail, and metadata to estimate its emotional and sensory tone.
Cloud Full Video Analysis (Gemini 2.5 Pro): Performs multimodal analysis where the AI “watches” and “listens” to the video, generating a detailed perceptual report.

The MVP focuses on children’s content, helping parents understand stimulus levels and make more conscious media choices.

How I built it

Tech stack

Frontend: React, TypeScript, d3.js
Chrome APIs: Manifest V3, content scripts, background service workers
AI: Gemini Nano (on-device), Gemini 2.5 Pro (cloud)
Backend: Firebase AI Logic, REST API integration for hybrid orchestration

Hybrid AI architecture

The system routes lightweight metadata analysis to Gemini Nano and offloads full multimodal reasoning to Gemini Pro. The connection layer manages privacy boundaries while maintaining consistent output formatting.

Prompt design

At the heart of the system is the D-Level Checker – a structured system prompt built on E-Prime language (avoiding absolute terms to maintain objectivity). Gemini evaluates each video through five perceptual layers:

Sensory: visual rhythm, color, sound
Affective: emotional trajectory, coherence
Cognitive: clarity, message depth
Rhythmic: pacing and neuro-balance
Functional: whether stimuli serve narrative or mere attention retention

Over 20 variables (ASL, BPM, saturation, etc.) are aggregated into a 0–100 D-Level score and five dopamine types: Reflective, Creative, Affective, Sensory, and Chaotic.

Challenges I ran into

Building my first Chrome extension required rethinking architecture – dealing with background messaging, state management, and isolated DOM contexts.
The bigger challenge was engineering a stable AI prompt that produces consistent multimodal analysis within a 32k-token reasoning window.

Accomplishments that I'm proud of

Created a working hybrid AI system balancing on-device privacy with cloud-level insight.
Built an original “digital nutrition label” concept that translates complex AI analysis into a form parents can easily understand.
Designed a systemic prompt that aligns large-scale AI reasoning with human well-being and child safety.

What I learned

Technically, I learned how to architect Chrome extensions and how to orchestrate client-side and server-side AI communication.
Conceptually, I learned that the art of prompting lies in role design – guiding the model as a specialized expert, not a question-answer engine.
Most importantly, I learned that transparency in AI helps people see how technology affects them, instead of leaving them in the dark.

What's next for D-Level AI Analyzer

Public Database & Crowdsourced Accuracy

Build an external API with a searchable database of analyzed videos.
Use multi-analysis averaging to generate more reliable D-Level scores.

YouTube Filtering & Parental Tools

Add a YouTube filter mode that highlights or hides videos by D-Level range.
Enable batch analysis of entire playlists or channels.

Digital Wellness Coach

Develop a “digital diet” dashboard that helps parents monitor children’s exposure to high-intensity content and suggests balanced alternatives.

Goal: not to eliminate stimulation – but to make it conscious, balanced, and meaningful. With Chrome’s built-in AI for local perception and Gemini’s deep reasoning in the cloud, we can finally build technology that truly understands what we watch.