The model-facing generation stack - route each deliverable to the right model, speak camera instead of adjectives, lock the campaign to references, and select like an art director.
Seven skills that operate an AI image and video pipeline rather than explain one. Route each deliverable to the model whose failure mode it survives, replace adjective-soup prompts with camera vocabulary at an 80-percent-checkable-token floor, freeze a campaign to reference blocks so assets cohere, direct one intention per video clip, keep legible product UI out of generated footage, and select from candidate sets with a forensic tell-checklist. Model names live in refreshable platform-facts blocks so the craft survives monthly model churn. Consumes the shot specs from product-photography-direction.
Click to play with sound.
Arranged in the author's recommended order. Walk through them in sequence, or open any one on its own.
Routes each creative deliverable to the right AI image or video model by its failure mode – typography, vector, photoreal with brand hex, campaign consistency, dialogue, human motion – and compiles one creative spec into each model family's prompt dialect.
Writes photorealistic product and editorial image prompts using camera vocabulary as the control surface – mandatory focal length, aperture, shot type, one named directional light, material optics, and an imperfection budget.
Locks a multi-asset AI campaign to one visual identity using reference mechanisms instead of re-described prose – a frozen content-hashed set block, a curated style-reference code library, brand-style locks, and multi-image references.
Decides where words in AI-generated imagery get rendered – in-image with a typography-capable model only when the text is a short quoted string within the ceiling this skill's platform-facts block states, composited in the design layer otherwise.
Directs AI-generated video by decomposing a brief into single-intention shots – one camera move plus one subject action per 5 to 8 second clip – then serializing the shot list to the target model's multi-shot mechanism with timestamp-exact audio cues and…
Runs the selection and revision discipline that separates shipped AI assets from lucky ones – a minimum four-candidate set per final asset, a tell checklist covering rendered-text read-back, hands, logo geometry, shadow, depth-of-field.
Escalates AI image and video work from prompt-only to pipeline mode when fidelity is contractual – structure conditioning for layout-exact generation, hex color conditioning for brand palettes, pinned seeds and saved workflows for reproducibility.