SafeSpace

Inspiration

The physical world wasn’t built for everyone. 15% to 20% of the global population is neurodivergent (including those with ADHD and Autism). In chaotic public spaces, sensory overload drains cognitive performance, and it takes an average of 23 minutes to recover deep focus after a single environmental distraction. Current assistive tech alters the screen, but ignores the physical room. We built SafeSpace to use software to actively filter and calm the user's physical environment.

What it does

SafeSpace is a full-stack web application that treats the physical world as a live input stream and dynamically dampens its chaos. By capturing the user's live webcam and microphone feeds, it applies real-time visual and acoustic filters to create a personalized sensory shield.

Our core implementation is The Deep Sea Space—an immersive, instant decompression room. It captures the user's video feed, blurs out distracting background room artifacts, and applies a deep underwater chroma-tint. Users can interact with autonomous, gesture-responsive AI fish (which swim into an open palm or flee when hand velocity spikes) and pop floating, canvas-rendered ASMR bubbles. Concurrently, it routes live microphone data through a dynamic low-pass filter to muffle harsh room noises into a dull, comforting underwater rumble—increasing cognitive focus by up to 30%.

How we built it

We engineered SafeSpace as a full-stack JavaScript monorepo using Vite, React, Tailwind CSS, and Express/Node.js.

Video/Audio Pipeline: Video tracks are processed via real-time HTML5 Canvas loops. To achieve zero-latency audio buffering without heavy assets, we used the native browser Web Audio API, routing live microphone data through a dynamic BiquadFilterNode low-pass filter.

Architecture & Workflow: The Express backend leverages an asynchronous local JSON database to save and load user presets (GET/POST /api/presets). We built this concurrently using Google Antigravity 2.0, assigning parallel subagents to compile separate frontend modules and API endpoints.

Challenges we ran into

Browser Security Restraints: Modern browsers strictly block camera and microphone access until explicit user interaction. We resolved this by engineering a custom gating modal that securely instantiates the AudioContext and media tracks on a single click.

Merge Collision Risks: With four developers coding simultaneously, we strictly partitioned the codebase into decoupled architecture files (UI, Canvas, Web Audio, and Express) before writing code, entirely eliminating version conflicts.

Accomplishments that we're proud of

Zero-Latency Client Edge Processing: Accomplished high-definition video frame manipulation and audio filtering entirely on the client edge with zero lag, ensuring a smooth experience free of sensory-inducing latency.

Full-Stack Persistence: Successfully linked algorithmic canvas states to a backend database for seamless user profile loading.

Data-Driven Empathy: Designed a functional assistive tool grounded directly in cognitive and neurological science.

What we learned

Web Audio Utility: Mastered low-level browser nodes for real-time acoustic manipulation without external latency.

Agentic Orchestration: Learned to scope and constrain AI subagents inside Antigravity 2.0 to hyper-accelerate our software output during a crunch.

What's next for SafeSpace

The Customizable & Fidget Spaces: Our immediate next steps are deploying our next two planned rooms—a Custom Space with step-less slider controls for step-less isolation tuning, and an ASMR Fidget Space for micro-gesture tactile grounding (wax melting, keyboard tapping).

Native Mobile Port: Porting the core engine to React Native so users can activate their sensory shields on the move using wireless earbuds and phone cameras in public transit or busy streets.

Biometric Automation: Integrating smartwatch APIs via WebBluetooth to automatically activate SafeSpace's dampening filters the moment a spike in heart rate or anxiety stress markers is detected.

Biometric Automation: Integrating smartwatch APIs via WebBluetooth to automatically activate SafeSpace's dampening filters the moment a spike in heart rate or stress markers is detected.