What is lineridervisualizer?

{"i":0,"repo":"BaselAshraf81/lineridervisualizer","stars":1,"lang":"TypeScript","ok":true,"char_count":1903,"text":"LineVisualizer is a web app that turns a song into a live, animated scene in your browser. You upload an audio file, and the app uses a neural model to separate the song into individual instruments like drums, bass, vocals, guitar, and piano. It then draws small balls that fall under gravity and hit physical lines at the exact moment each instrument plays. The result is a music-driven animation in the style of DoodleChaos, where the track is built to match the music, rendered live on a web canvas as the song plays.\n\nTo get started, you install Node.js dependencies and set up a Python environment. A setup script handles the Python side, which includes the neural separation model and audio processing libraries. An NVIDIA graphics card is strongly recommended for speed. The app falls back to CPU processing without one, but that is much slower. Once running, you open the app in your browser, drop in an audio file, and press Generate. The first run downloads a roughly 170 megabyte model file automatically.\n\nWhen you generate a scene, the app sends the audio to a Python tool that splits it into six instrument tracks. Drums are further divided into kick, snare, and percussion. Each instrument becomes its own colored ball. The app plans a path so that every ball lands on its line precisely on the beat, using gravity-based arcs to connect the notes. The animation is deterministic, meaning every visual frame is tied directly to the audio playback time, so pausing or seeking stays perfectly in sync.\n\nYou can customize the scene in the browser. By default there is one ball per instrument, but you can drag instrument labels between balls to group them or split them apart. You can rename balls, change their colors, hide them, and use sliders to adjust line height and angle, giving you control over how the balls weave across the screen.\n\nThe project is a TypeScript monorepo with three main parts: shared data types, a neural audio analysis package that wraps the Python tools, and a web package containing the server and React client where the live scene is rendered. The codebase includes a detailed map file for new contributors. The project is private and not published as an installable package.","error":null}