I wanted to explore whether AI could help a writer perceive a scene differently, not just generate more text. Writers often know what a scene should feel like but may not immediately notice the relationships between background sounds, distance, attention, and atmosphere.
Sound Map lets a writer describe an imagined scene. AI proposes an initial arrangement of available sounds, and the writer can listen, move sounds closer or farther, remove what does not belong, and save the result as a sonic blueprint to return to while writing.
The current café scene is a proof of concept using three recordings: rain, room ambience, and sweeping. Unsupported sounds remain visible as notes instead of being silently replaced.
How we built it. The prototype uses React, Vite, Node.js, Web Audio, and OpenAI models. AI is used for the initial interpretation and optional reflection, while sound movement, playback, saving, and reopening are handled locally in the browser.
Challenges we ran into. The hardest part was making the sound relationships feel meaningful without pretending the prototype was a full room simulation. I also had to balance AI suggestions with writer control, handle unsupported sounds honestly, and test whether distance and prominence were actually perceptible.
Accomplishments. I’m proud that Sound Map grew from a simple idea into a working creative prototype with a clear purpose. I learned how to shape an AI-assisted product without letting AI take over the writer’s decisions, tested multiple model choices, built a usable sound-mapping interaction, and worked through real issues around perception, unsupported sounds, saving, privacy, and reflection ownership. I’m also proud that the final prototype stays honest about its limits while still showing the broader potential of the idea.
What we learned. I learned that AI can be more useful as a creative perception partner than as a content generator. The strongest interaction was not the AI producing a final answer, but the writer hearing a proposed arrangement, changing it, and deciding what the scene meant.
What's next for Sound Map. This prototype is intentionally narrow. A future version could support richer sound libraries and many kinds of environments, but this submission only demonstrates the café proof of concept.
Log in or sign up for Devpost to join the conversation.