This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine
What I Built
I built Radio Ghost, an interactive, vintage-inspired AI radio experience that broadcasts mysterious stories from an imagined past. Users can tune into different radio stations, listen to eerie broadcasts, explore live transcripts, and make choices that influence how the story unfolds.
I interpreted the theme by imagining an old radio that receives strange transmissions from another time. The project combines a retro radio interface, AI-generated storytelling, and an atmospheric listening experience. I was inspired by vintage radio dramas and the unsettling idea of hearing a voice from somewhere in the past.
Radio Ghost is designed to make storytelling feel less like reading a story and more like tuning into a mysterious transmission.
Demo
Code
https://github.com/ZikraRahman/ghost-radio
How I Built It
Radio Ghost is built around local AI inference using Ollama, with Gemma 3 (4B) as the open-weight language model. The model generates supernatural stories and mysterious radio transmissions based on the selected station and the listener’s choices.
The application uses a React, TypeScript, and Vite frontend with a FastAPI backend. The backend connects to the locally running Ollama model, allowing AI-generated content to be produced on the user’s own machine rather than relying on a hosted AI API.
This local-first approach makes the experience more private and accessible while demonstrating how open-weight AI can power interactive storytelling. Browser-based speech synthesis provides narration, allowing users to experience the broadcasts as an eerie radio drama.
Why Does Open Innovation Matter?
Open innovation made Radio Ghost possible as a local-first AI experience. Using open-weight models through Ollama gave me the freedom to experiment with AI-generated storytelling without depending on a paid, closed API.
A closed API could generate stories, but it would limit my control over model choice, deployment, and access to the underlying system. With open-weight AI, I can run the model locally, customize the experience, reduce API dependency, and keep story generation on the user's machine.
It also makes the project easier to extend in the future—whether by trying different models, improving the storytelling, or making the experience more accessible. Open innovation turns Radio Ghost from a simple AI-powered interface into an experiment I can genuinely build on.
The Spirit of It
The idea behind Radio Ghost is simple: what if an old radio suddenly picked up a voice that should not exist?
I wanted to create a moment where the listener pauses and wonders, “Wait… did that transmission just answer me?” The combination of a mysterious broadcast, an old-fashioned radio interface, and choices that shape the story turns a simple listening experience into something personal.
Radio Ghost is my attempt to bring back the feeling of vintage radio mysteries while using AI to make every transmission feel like it could be happening right now.
Prize Categories
Category: Best Use of Gemma
Model: Gemma 3 (4B)
Inference: Local, through Ollama
Top comments (0)