OpenAI Realtime Audio POC
Proof-of-concept for OpenAI's WebRTC/Realtime stack that captures microphone input, streams it to the Realtime API, and plays the AI-generated audio reply live in the browser.
- OpenAI
- WebRTC
- Audio Streaming
labs.infocentral.net
Experimental work
AI, WebGPU, browser, and prototype work that is useful to track separately from polished tools.
Proof-of-concept for OpenAI's WebRTC/Realtime stack that captures microphone input, streams it to the Realtime API, and plays the AI-generated audio reply live in the browser.
Repository for a client-side playground for running ONNX transformer models entirely in the browser, featuring text generation, summarization, and token visualization powered by WebGPU execution.
will request permission to download ~200MB model files and transformer library
A minimal WebGPU port of the YOLOv8 tinygrad model that loads <code>net.safetensors</code>, streams the weights into GPU buffers, and runs real-time object detection on live camera input with overlay controls directly in the browser.
will request permission to access camera and run WebGPU inference with Safetensors
Repository for a browser-based Mandelbrot renderer that combines WebGL and WebGPU shaders to deliver smooth zooming, orbit traps, and color palette experimentation directly on the page.
Repository for a single-page assistant that ingests an uploaded PDF, extracts a chosen page or the entire document, and sends your question plus the extracted content to an AI backend using a user-provided API key.
will ask permission to upload PDFs, download LLMs, or request an OpenAI API key