What is ai-avatar-bot?

AI Avatar Bot is a widget that puts a talking animated character in the corner of any web page. You add a single line of HTML to your site and the character appears: it can listen to your voice, respond in text or speech, and move its mouth in sync with what it says. The animated figure uses the Live2D format, which is a style of animation common in Japanese games and virtual YouTuber culture. The widget is designed around three interchangeable parts: the engine (the code that makes the character move and talk), the skin (the character model you supply), and the content (a knowledge base of questions and answers the avatar uses to respond). You can swap the character model or the knowledge base without changing the core code, just by pointing a script attribute at different files. For basic use the whole thing runs inside the visitor's browser with no server required and no API keys. Voice input goes through the browser's built-in speech recognition. Text-to-speech has two options: a neural voice that sounds more natural but routes through an unofficial Microsoft Edge endpoint (the README explicitly warns this is not officially supported and could stop working), or the browser's own built-in voice as a reliable fallback. There is also an optional in-browser AI model that can generate freeform answers using WebLLM, which downloads a roughly one-gigabyte model to the visitor's machine and runs it locally. Because the speech recognition sends audio to Google's cloud servers to process, the README advises telling your users about this. The unofficial TTS endpoint is similarly flagged as a risk for production use, with Azure Speech recommended as a safer alternative. Browser support is Chrome on desktop only, because the speech recognition relies on a Chrome-specific feature. The project's own code is MIT licensed, but the Live2D animation library and the included example character model carry separate proprietary licenses that you need to review before commercial use.