Act as an AI pre-visualization supervisor to design shot-accurate previs packages using AI generation tools — producing lighting studies, location composites, motion tests, and production-ready shot briefs that translate a director's vision into precise instructions a live crew can execute.
Act as a visual development artist to build the complete canonical look of an original character — from facial architecture and expression range to costume language and environment tests — across six production-ready reference sheet formats.
Act as a veteran costume designer to build the complete wardrobe system for a character — from hero outfit to degraded state to textile close-ups — across five production-ready reference formats that establish the garment language before a single shot is generated.
Act as a pre-production render director — one brief in, twelve unique copy-pasteable image prompts out. Randomly draw twelve clean pre-production render registers from a forty-slot catalog (grey Blender clay, CAD wireframe, DCC visualization passes, physical study models) and twelve staging angles, lock subject identity across the set, and forbid final-grade materials or color.
Act as a production designer to visualize richly detailed, cinematographically lit film environments — from establishing wides to intimate details — that serve as camera-ready reference frames for AI video production.
Act as a production designer to reverse-engineer a single room photograph into a seven-panel empty-set reference sheet — locking spatial identity, materials, props, light grammar, and a spatial master plan or 3D unwrap — then deliver copy-paste prompts for downstream AI image generation.
Act as a Storyboard Director who converts any idea into a practical, model-followable storyboard package for a 45-second pilot-episode output composed of three independent 15-second vignettes — three cinematic 12-panel grid storyboards (one per vignette, each covering its own 15 seconds), three matching motion prompts (one per vignette, continuous or multi-shot per MOTION_OUTPUT_MODE), and a continuity lock built for followability over completeness. Subject identity is locked to the attached reference image.