AudioWorld

Inspiration

On Tuesday, I ended up in the ER with a severe migraine that caused temporary deafness. During my visit, I had to go through a very rudimentary hearing test, a process that felt outdated and disconnected from how people actually experience sound in the real world. When I dug deeper, I discovered that most audiology tests are still performed in sterile, quiet rooms, even though patients struggle most in noisy, everyday environments. That gap inspired me to build AudioWorld: a system that finally brings real-world scenarios into hearing assessments, ensuring patients are tested and treated for the situations that matter most.

What it does

AudioWorld Test simulates real-world listening environments (like cafés, subways, or offices) so audiologists can evaluate how patients hear and understand speech in everyday noise. Audiologists purchase credits through Flowglad, assign tests to patients, and results are automatically scored and exported as a PDF for easy EMR integration. A Honcho AI agent monitors missed follow-ups or failed tests, ensuring no patient slips through the cracks

How we built it

I rapidly prototyped AudioWorld using Lovable to scaffold a React web app with a Node backend and Supabase for data connextor. Flowglad powers the payments system for audiologists to purchase test credits, while Honcho provides an AI agent that tracks follow-ups and escalates missed tests. For the patient test experience, I combined speech synthesis with layered background noise to create realistic audio environments, then built automated scoring and PDF export to simulate EMR integration.

Challenges we ran into

Simulating lifelike audio environments under hackathon time constraints was tough — balancing speech synthesis with background noise required careful tuning. Integrating Flowglad payments and Honcho agents into a single, seamless workflow was another hurdle, since we had to wire together two systems that don’t normally talk to each other. Finally, making the demo both clinically meaningful and simple enough to run end-to-end in just a few hours pushed us to focus on one test flow that proved the concept.

Accomplishments that we're proud of

I built one of the first generative world models for audio environments, and I believe the first for an auditory training and analysis application. Utilizing synthetic audio data, integrating AI agents for follow-up, and developing rudimentary auditory tests + automatic scoring was a great way to develop a functional prototype to meaningfully improve hearing care.

What we learned

Learned how to integrate various technologies in payments, agents, orchestration, and synthetic audio to build a productized world model for audiologists. It was also great to work under time pressure.

What's next for AudioWorld

I plan to expand beyond the initial “busy café” test by adding more real-world environments (subway, classroom, office meeting) and multi-language support. On the clinical side, I want to integrate directly with EMRs for seamless record-keeping and validation with audiologists to refine scoring. Longer term, I see opportunities to partner with hearing aid manufacturers, turning AudioWorld into a platform to build better devices.

Built With

• asr
• audio
• flowglad
• generative
• honcho
• lovable
• n8n
• supabase
• synthetic