I am ready to embed a fully-customized, ChatGPT-powered voice assistant into my website and need a developer who can take the project from concept to live deployment. The assistant’s core role is to listen to a visitor’s spoken question, send it to ChatGPT, speak the answer back, and display the text on-screen. When ChatGPT cannot confidently respond, it should fall back to a friendly, generic clarification message rather than escalating to a human agent. Scope of work • Build or extend a lightweight JavaScript/TypeScript widget that taps the Web Speech API (or similar) for real-time speech-to-text and text-to-speech. • Connect the widget to the OpenAI Chat Completion endpoint, passing conversational context so answers feel natural. • Expose configuration hooks so I can adjust system prompts, voice type, language model parameters, welcome message, and UI colours without touching source code. • Ensure the solution loads quickly, works on all modern browsers and mobile, and plays nicely with my existing front-end (vanilla JS with a small React component set). • Deliver a short README and commented code so I can maintain and extend it later. Acceptance criteria 1. Voice assistant installs via a single script tag or npm package and initialises with one line of code. 2. Spoken queries are transcribed with >95 % accuracy under normal conditions. 3. Responses are read out within three seconds on a standard broadband connection. 4. Unanswerable questions trigger the predefined generic response every time. 5. No API keys or secrets are hard-coded; they must be injected via environment variables. If you have previous work with Web Speech API, OpenAI, or similar voice interfaces, please include a quick demo link or repo in your bid so I can gauge fit.