Every lesson, exercise, and deliverable you'll complete.

Choose only visible identity anchors and check whether the references support product-only, detail or human-use proof.

Use ZH or EN, append one matching category preset and generate a single 4:5 draft without new advertising copy.

Review product fidelity before beauty, freeze passed areas, repair one defect and choose Pass, Revise, Hold or Reject.

The 8 visual tells that make AI content feel fake: wax skin, dead eyes, floating objects, physics breaks, smooth-everything, generic scenes, impossible reflections, and scale drift.

Product truth, scene logic, physical behavior, camera logic, light + materials, natural imperfection, and continuity + correction. How they stack and why fixing one layer without the others still looks fake.

Side-by-side analysis of weak vs. directed AI outputs across food, beauty, home goods, and services. Practice naming which layer breaks in each case and what prompt change fixed it.

How to document geometry, packaging, label hierarchy, color, finish, and approved facts so your AI never invents product details. Includes templates for food, beauty, electronics, and home goods.

Constructing believable person, place, moment, and action context. Why "luxury lifestyle" is not a scene and how to replace it with specificity that AI can actually render.

The anatomy of a realism prompt: product lock, scene direction, camera spec, material behavior, exclusions, and self-check. How to structure prompts that produce reviewable results, not random outputs.

How to direct viewpoint, framing, focus plane, motivated light, and category-correct material behavior. Making glass look like glass, food look edible, fabric look textured, skin look real.

Adding selective texture, asymmetry, wear, and environmental detail. Why "too perfect" reads as fake and how controlled imperfection can reduce an overly synthetic visual feel.

Freeze what passed, identify one defect, correct it, re-check. The iterative loop that turns a "close enough" generation into a reviewable image draft. Includes copy-paste correction prompts for common failures and a mandatory human QA gate.

The gap between "looks good" and "answers a buyer question." Understanding buyer context, visual hierarchy for commerce, and how to direct images around a measurable hypothesis — not just look pretty.

A practical AI Designer workflow for Shopee listing sets and promotional posters: buyer job, base artwork, exact copy, visual hierarchy, safe areas, export checks, and versioned revision briefs.

Complete guided walkthrough of the 5-stage Prompt Builder: Diagnose → Message → Angle → Creative → Test. Use it to systematically build conversion-focused creative direction.

A structured scorecard for evaluating AI-generated images across all 7 realism layers. How to objectively compare variants and pick the strongest direction — not just the "prettiest" one.

Designing tests that isolate one variable at a time. How to build a test card, set observation windows, define keep/revise/retire rules, and avoid false conclusions from noisy data.

Documentation that compounds: how to maintain a correction library, variant history, and quality baseline that makes every future project faster and more consistent.