AI Influencer Live AMA Platform

I run a virtual influencer built with generative-AI and I’m ready to take her interaction to the next level: real-time “Ask Me Anything” sessions streamed on Instagram or TikTok Live where viewers speak their questions through their microphones and receive spoken, perfectly lip-synced answers in her own voice. Here’s the flow I need working end-to-end: • Incoming audio from the live stream is captured and converted to accurate text fast enough to feel natural. • That text is fed into the character’s existing personality engine (GPT-style prompt chain already in place) to craft a short, on-brand reply. • The reply is turned back into speech with the exact voice model we use, then lip-synced to her face model and pushed straight back to the ongoing live video. • All of this must run continuously during a live broadcast on either Instagram or TikTok, keeping latency low so the audience feels they’re in a real conversation. Key points - You’ll be working with the current character assets (voice model, 3D face rig, style prompt library) and can choose the underlying stack—Unity, Unreal, OBS, WebRTC, or another pipeline—provided it meets the speed and quality targets. - The solution must survive typical live-stream chaos: overlapping voices, background noise, and large audience spikes. - A brief dashboard or CLI toggle to start/stop the pipeline, monitor latency, and log each question/response will help with show moderation. Deliverables 1. Running prototype demonstrating one full conversation loop on Instagram or TikTok Live. 2. Documented build/install instructions plus environment list. 3. Source code (Python/C#/C++ or mixed) with clear comments. 4. Post-launch support window to patch issues discovered in the first live event. Acceptance criteria • End-to-end round-trip (spoken question to spoken answer) consistently under 5 seconds. • Word error rate ≤ 8 % on noisy user audio. • Lip-sync visibly aligned within one video frame. • No streaming drops over a 30-minute test run. When you reply, focus on relevant experience with real-time GenAI character systems—links or brief case studies are perfect. Add your hourly rate and how long you expect the full build (through prototype) to take.