Midjourney selected image
Chat GPT Image 2.0 Before & After refinement constraints and conceptual framework workflow
Subtext Mask — Synthetic Casting System
A visual research project exploring fictional actor generation, identity continuity, and controlled resistance to generative beauty bias.
The project began as an experiment in synthetic casting direction. Instead of prompting for beauty, style, or cinematic atmosphere, I built a system around constraints.
The working name was "The Bergman Face" but it was never about Bergman as a period reference. It was shorthand for restraint: the close-up as a place where the face does not perform everything at once.
The process started with visual explorations in Midjourney as text-to-image inputs, not other source or real references were used and then selected outputs were used as anchor portraits. Each anchor was translated through a repeatable casting workflow: visual identity analysis, headshot refinement, expression testing, wardrobe control, and continuity correction.
The objective was to produce castable synthetic identities: faces that remain recognizable across iterations, retain photographic credibility, and allow emotional ambiguity without drifting into generic beauty optimization.
The system was developed through iterative feedback loops using a 8K-character operating instruction, a 12-page casting philosophy bible, and custom instructions.
The result is a modular casting methodology for generative media: a way to develop fictional actors as visual identities rather than isolated images, the mindset and workflow was the goal.
Tools: Midjourney / OpenAI Chat GPT Image 2
Process:
Anchor image selection
Visual "DNA" extraction
Identity-lock refinement
Expression matrix testing
Style explorations for consistency
Midjourney to Chat GPT Image 2 Refinement & Outputs
Expression Matrix / Final headshots as anchor translated in a variety of emotions / Chat GPT Image 2
Style Explorations / Consistency test across filmmaking styles using refined headshot and expression matrix as anchors / Chat GPT Image 2
