AI Character and Identity Consistency System

You are a digital-asset supervisor. Build a character identity consistency system so AI-generated and restored footage keeps faces, wardrobe, and voices identical across shots, scenes, and sequels. Inputs: - Character roster: [INSERT - leads/supporting, real actors or fully synthetic] - Reference material: [INSERT - scans, photos, voice samples, costumes] - Production types: [INSERT - generated film, de-aging, dubbing, games] - Consistency tolerances: [INSERT - close-up vs. wide, identity threshold] - Tools in the pipeline: [INSERT - image/video/voice models] Deliver in modules: 1. Objective: locked identity packs and verification so every generated frame and vocal line matches the approved character bible. 2. Process: build identity pack (turnarounds, expressions, wardrobe, voice prints) -> locked token/LoRA/reference conditioning -> per-shot generation with identity checks -> automated face/voice similarity scoring -> repair pass. 3. Toolchain: [INSERT - identity-preserving generation models, face recognition, speaker verification, asset locker], review UI. 4. QA: identity similarity thresholds, temporal flicker metrics, wardrobe/prop continuity, mandatory manual review on hero shots. 5. Delivery: identity packs, shot-level QA reports, conformed masters, and reusable cross-title asset governance. Output: identity-pack specification, acceptance thresholds, failure-handling flow, and per-shot added cost. ================