Direct-on-Reference Frame Directing

von @samifox_ai

Prompt

I'm honestly blown away. GPT IMAGE 2 doesn't "prompt" scenes, it directs them. Mark actions and camera moves right on a reference frame → 10x more context and near-total control in one shot. #gptimage2 #seedance2

Why this prompt works

This community post showcases a workflow rather than a scene: marking actions and camera moves directly on a reference frame to give the model far more context and near-total one-shot control. It frames GPT-Image-2 as 'directing' rather than 'prompting', highlighting annotation-driven control over composition and motion.

Variations to try

·Annotate a reference frame with arrows for a character's walk path
·Mark a camera dolly-in plus a subject gesture on one still
·Add labeled zones on the frame to control background versus foreground action

Style & lighting notes

Demonstrates a direct-on-reference-frame directing technique: action and camera-move annotations drawn onto a still grant high contextual control. Emphasis is on workflow and one-shot precision rather than a fixed visual style.