I’m building a fully automated, always-on “virtual host” for TikTok Live whose sole purpose is to entertain. The character must feel friendly and cheerful on camera, react to chats in real time, and keep viewers engaged with upbeat banter, jokes, and quick shout-outs. Here’s what I need from you: • A complete AI streaming pipeline that plugs into TikTok Live (RTMP or TikTok Live Studio/API). • Real-time comment ingestion and intent detection so the avatar replies within a couple of seconds, never missing a viewer question. • Language generation powered by GPT-style models (open-source or OpenAI) fine-tuned for a playful, positive tone. • Natural-sounding TTS that syncs to the avatar’s mouth; think lip-sync via tools like OBS, FaceRig, Animaze, Live2D or similar—whatever you prefer, as long as it feels alive. • A simple control panel where I can adjust personality sliders, banned words, and fallback responses before each show. • Clear hand-off: setup scripts, model prompts, avatar files, and a step-by-step guide so I can launch future streams without engineering help. Acceptance criteria 1. During a 30-minute test broadcast the AI answers at least 80 % of unique comments, stays on-brand (friendly, cheerful), and never repeats the same line twice in a row. 2. Average response latency under 2 seconds. 3. All components run on my local PC or a single cloud instance I can replicate. If you have previous examples—especially anything involving OBS, Python, Node, or the TikTok Live SDK—please mention them so I can see your approach. Looking forward to turning this digital host into the next must-watch live stream.