Three-step method + 4 design tips
1 image model + 2 prompts, three steps turn text into knowledge cards; plus 4 tips so AI stops guessing style.
The soul of the course: 1 image model (Qwen Image 3.0) + 2 prompts, three steps turn text into knowledge cards. The 4 design tips are the "fuse" of this method — so AI stops gacha-ing every time and you can build a coherent set.
Three-Step Method Overview
Deconstruct
Long text → image-ready copy (understand · filter · reorganize)
The universal deconstructor turns any text into "image-friendly" copy.
Generate
Structured copy → knowledge card image
Generate the first card by "card type + locked style", then only swap topics — never the style.
Polish
Manually audit each card using the checklist
Find issues → go back to the model to fix text / layout / style → re-check the table → ship only if passing.
Step 1: Deconstruct
Turn long text into "image-friendly" copy — not a rephrasing. The 8-step deconstruction chain:
Summarize topic → map info layers → extract core points → break key terms → mark contrasts/causal → list process steps → organize reusable sentences → suggest structure
Key decision:
- Single topic, moderate info density → one card copy
- Multiple topics, steps, or parallel concepts → 3-8 card copy, each focused on one topic
Step 2: Generate
Output by "card type + locked style". Pick the card type from the 6 categories in Lesson 1, pick the locked style from tonight's main "warm notebook" or 4 alternatives.
Warm notebook style (tonight's main): light cream paper background + warm brown ink strokes + hand-drawn line border + retro bookmark + ballpoint pen illustration in the corner; top has title + one-line overview; left side has core concepts, middle has method steps, right side has a quote; overall flat, no 3D shadow, no decoration blocking text.
Cream paper + warm brown ink + hand-drawn border + bookmark illustration, flat with no shadow.
Light blue rice paper texture + light ink lines + vermilion accents + ink brush strokes, 16:9 landscape.
White notebook paper + black handwriting + fluorescent highlights + margin annotations.
Cream background + large color blocks + rounded cards + linear icons + thin arrows, 3:4 portrait.
Aged parchment texture + double-line border + serif font + cinnabar seal, black · deep blue · cinnabar tri-color.
Step 3: Polish
A must-do after every card — no prompt needed, just use your eyes. The 6-step QC loop:
Zoom in on text
Especially small text, formulas, super/subscripts, LaTeX — AI's most failure-prone areas.
Verify numbers & formulas
Numbers, dates, percentages, version numbers — verify them yourself.
Check element occlusion
Are bookmarks, seals, illustrations covering text? Are edges clipped?
Check positions
What goes in left / center / right / top / bottom? If positions drift, add position constraints back to the model.
Check colors
Hard-code color values + "no cool blue" type hard constraints to avoid drift.
Check corners
Any watermarks, copyrights, author signatures you don't want? Decoration elements stealing the show?
Core principle: find issues → go back to the model to fix text / layout / style → re-check the table → ship only if passing.
4 Design Tips
Core Principle
AI handles 70%; the remaining 30% depends on how clearly you write your requirements. The three steps are the skeleton; the 4 tips are the muscles — missing any piece and you can't build a coherent set.
Next lesson we run 3 hands-on demos using The 7 Habits of Highly Effective People — taking the three-step method end-to-end.
Knowledge cards: fundamentals & concepts
5 core concepts + 6 categories + 4 use cases + 8 high-frequency scenarios + tool selection — see the lingo, boundaries, and the right tool before making cards.
3 hands-on demos: The 7 Habits
Demo 1 deconstructs the book → Demo 2 generates cards → Demo 3 audits + bakes the style into a Skill. End-to-end three-step method in one lesson.
Tutorials