What is codefall-face?

Codefall Face is a JavaScript project that renders a talking face for an AI agent using nothing but falling green and white glyphs on a black background, similar to the visual style of digital rain. Instead of drawing a 3D head or using an avatar image, the code builds the face out of moving text characters that shift and reform in real time. The face speaks, changes expression across nine emotions, glitches, and appears to dissolve back into the stream of characters when interrupted. You can try it without installing anything beyond Python or Node: run a simple local web server, open the page in a browser, and press the demo button. This mode uses your browser's built in speech features, so no accounts or API keys are needed. There are two visual themes you can switch between, plus two face shapes, and the project respects settings for reduced motion. For real time voice conversations, the project can connect to Azure Voice Live, a service for streaming audio back and forth over a WebSocket connection. Because browsers cannot safely include a secret key in that kind of connection, a small backend server is included whose only job is to hold that key and pass audio through. Without an Azure key configured, the app automatically falls back to the browser's own speech tools. Two other options exist too: Piper, which runs a voice model entirely on your own computer with no cloud service involved, and Lacy.ai, described as a fallback that currently only supports text based replies over its API rather than streaming audio. Under the hood, each frame of the face is calculated mathematically rather than played back from images, using shapes for the jaw, eyes, and mouth that are then converted into characters and rendered efficiently for smooth animation. The project has no listed license information in what was reviewed, and has not yet attracted stars from other users.