Proactive AI Companion App Build -- 2
Publicada el 2026-07-23
Descripción de la oferta
I want to take the idea of an AI friend far beyond the usual “ask and it answers” model. The goal is a fully proactive companion that runs on both iOS and Android, notices when I’m around through the camera, starts talking on its own, keeps the conversation flowing, and handles live video and audio in real time. Here is what the finished product must feel like in practice: I open the app, set the phone down, and walk away. When the camera recognises I’m back in the frame, the AI greets me, offers a thought, or simply comments on what it last “heard” a moment ago. If I go quiet, it doesn’t sit idle; it fills the silence naturally—exactly the way a person in the room would. All of this needs to happen with sub-second latency so the exchange feels alive rather than scripted. To achieve that, I’m expecting: • Native or high-performance cross-platform code that runs smoothly on recent iOS and Android devices. • Real-time video pipeline for presence detection (camera-based) and a lightweight on-device inference loop where possible. • A speech stack that pairs streaming recognition with expressive TTS so responses begin while the user is still finishing a sentence. • An LLM-powered dialogue engine tuned for proactive initiation instead of purely reactive replies. • Backend services (if required) that scale, but a clear plan for caching, rate limiting, and privacy of audio/video data. Acceptance criteria: 1. The app starts a contextually appropriate utterance within one second of detecting the user. 2. Conversation never stalls for more than three seconds of silence unless the user clearly ends the session. 3. Video, audio, and LLM calls run continuously for 30 minutes on a modern handset without overheating or noticeable lag. If you have shipped anything similar—especially involving proactive dialogue, live video overlays, or camera-triggered interactions—please point me to it. A short technical outline of your proposed architecture plus a demo or repo link will help me choose quickly so we can move straight into milestones and make this companion come alive. use high quality pre built character models with motion tracking For visual realism: use high-quality pre-built character models (Live3D, Ready Player Me, or similar) with realistic motion tracking and natural facial animations. Integrate these with a real-time video pipeline using Unity or Unreal Engine. Use emotion detection through models like DeepFaceLab or MediaPipe to match expressions to what the AI is feeling in the moment. Experience with AI/ML, real-time video/audio processing, chatbot systems, and character animation is required. If you've built similar interactive AI projects before, please share examples.
Skills
Fuente original: freelancer