StrideCast

StrideCast is a mobile-first Progressive Web App (PWA) designed to create a personalized podcast for your run, regardless of cellular data availability. The app generates an audio narration based on the runner's chosen duration, pace, mood, and topic, all while tracking the route in airplane mode. StrideCast was created as part of the DEV Hacktoberfest Open-Source AI Challenge Week 1, with the theme "Touch Grass."

The setup process takes only about 30 seconds, allowing users to configure their preferences for duration, pace, mood, and the podcast topic. Once set up, users simply start the run by pressing a single button. The app then handles the rest, offering an offline-ready experience with a user-friendly interface.

StrideCast utilizes open-source AI, specifically the Gemma2:2b model served by Ollama, to generate the podcast script. The backend calculates the number of words needed for the run, which is about 150 per minute, and then requests structured, segment-based scripts of 2-3 minutes each from the model. Voice synthesis is handled by Kokoro, which converts each segment into audio.

The frontend of StrideCast is built using React 19, TypeScript, and Tailwind v4, while the entire application is packaged as a Progressive Web App using vite-plugin-pwa (Workbox). Finished audio is saved into the browser's Cache API, enabling offline playback. User settings and run history are stored using Zustand and IndexedDB, ensuring all data remains on the device and never leaves it.

The app's backend is built with FastAPI and Pydantic v2, and it communicates with the frontend using Server-Sent Events to provide real-time updates on the script writing and voice recording progress. StrideCast works seamlessly without any map tiles or external map data, making it an ideal companion for runners and trail joggers who want company on their runs without relying on cellular data or an account.