Developer creates open-source browser rhythm game powered by local neural networks
A developer has built an open-source, browser-based rhythm game called Struna that uses four lightweight neural networks running via WebAssembly to analyze MP3 files and procedurally generate gameplay levels entirely on the user's device. This project demonstrates the growing capability of client-side AI, showing how multiple specialized neural models can run efficiently in a web browser without requiring cloud servers or compromising user privacy. The game uses BeatNet, Spotify's Basic Pitch, a Spleeter vocals network, and SwiftF0, totaling about 22 MB, running on onnxruntime-web with WebAssembly and falling back to digital signal processing if needed.
## BACKGROUND
WebAssembly (WASM) allows high-performance code written in languages like C++ or Rust to run in web browsers at near-native speed. ONNX Runtime Web is a JavaScript library that executes machine learning models directly in the browser utilizing CPU, WebGL, WebGPU, or WebAssembly.