Descriptions:
“Gap Focus” is a cinematic composition technique centered on strategic obstruction — placing foreground elements that block 40-60% of the frame, forcing the viewer’s eye through a narrow visual channel toward the subject. The effect creates a voyeuristic, narrative-charged quality particularly effective for Film Noir, Ancient Fantasy, and Cyberpunk aesthetics. This Veteran AI tutorial explains the underlying principles (credited to a Twitter post by @voxcat) and then encodes them as a reusable AI agent Skill for use with Google’s Antigravity development tool.
The Gap Focus Storyteller Skill is a structured Markdown (skill.md) file that guides an AI agent through four steps: defining the foreground obstruction type and coverage percentage, specifying the gap shape (window panes, cracked doors, gaps in foliage or buildings), constructing a subject with an explicit narrative action, and adding atmospheric lighting cues like the Tyndall effect. A built-in review checklist validates tunnel vision, center placement, foreground blur contrast, and overall narrative quality before the final prompt is output.
The live demonstration runs in Antigravity’s Planning mode using a Gemini model, generating a detailed image prompt for a ladybug climbing between grass blades — showing how the agent interprets the Skill file, generates a step-by-step execution plan for user review, and then produces a prompt ready for text-to-image generators. The resulting images are designed to feed directly into video generation pipelines like Wan 2.2. This is the second Skill tutorial in the Veteran AI series, following the earlier dynamic video prompt generator.
📺 Source: Veteran AI · Published January 27, 2026
🏷️ Format: Tutorial Demo
