I want to spice up my TikTok Lives with a playful automation: each time a viewer sends a gift, a 3-D NPC, study the sender’s profile picture with AI image recognition, then blurt out a short, cartoon-style voice remark. Everything needs to happen in real time inside a single Python-based workflow. What has to happen • Listen to live gifting events through the TikTok Live WebSocket (or another reliable Python library). • Pull the gifter’s avatar URL and feed it into a lightweight vision pipeline—open-source CNN, cloud Vision API, or something equally fast—to extract visual cues such as dominant colors, detected objects, face attributes, or overall “vibe.” • Turn those cues into a fun sentence, keeping it friendly and stream-safe. • Pass the sentence to a TTS engine that can deliver an unmistakably cartoonish voice (e.g., ElevenLabs, Uberduck, Coqui-TTS). • Lip-sync that audio to a rigged 3-D character and render it as an overlay for OBS/RTMP so the audience sees and hears the comment with no more than a ~2-second delay. Tech comfort zone Python is the backbone, so feel free to lean on PyTorch/TensorFlow for vision, PyOpenGL or a quick Unity/Unreal sidecar for rendering, and any proven TTS toolkit. Just keep licensing clean. I’m on Windows and stream through OBS. What I need in hand • Complete Python source with requirements.txt • The NPC model, rigged for real-time lip sync (or clear instructions for plugging in my own) • A small config file where I can edit catch-phrases, filter words, and tweak TTS pitch/speed • Step-by-step setup notes so I can reproduce the environment and go live without surprises Acceptance check I’ll run a private test stream, send gifts from another account, and expect to hear the NPC’s cartoon voice commenting on my avatar before the next gift arrives. If this sounds straight-forward to you and you’ve worked with live streaming, image recognition, and TTS before, I’m ready to get started right away.