@jairodri/huckleberry-cli 1.0.2 → 1.1.0

package/skills/creative/references/VISUAL_INTELLIGENCE.md

@@ -0,0 +1,1233 @@
+# Visual Intelligence for AI Creative Production
+
+**The missing layer between "generate an image" and "create something that actually works."**
+
+This reference document contains the visual psychology, anti-generic techniques, style vocabulary, platform-specific strategies, prompt construction guidance, and brand consistency techniques that separate forgettable AI output from images that stop scrolls, drive clicks, and convert.
+
+---
+
+## Table of Contents
+
+1. [Visual Psychology — What Makes People Stop](#part-1-visual-psychology--what-makes-people-stop)
+2. [Direct Response Visual Principles](#part-2-direct-response-visual-principles)
+3. [Anti-Generic AI Aesthetic Principles](#part-3-anti-generic-ai-aesthetic-principles)
+4. [Anti-AI-Slop Playbook](#part-4-anti-ai-slop-playbook)
+5. [Style Vocabulary Taxonomy](#part-5-style-vocabulary-taxonomy)
+6. [Leading Brand Visual Identities](#part-6-leading-brand-visual-identities)
+7. [Platform-Specific Visual Strategy](#part-7-platform-specific-visual-strategy)
+8. [Prompt Construction for the configured image model](#part-8-prompt-construction-for-image-generation)
+9. [Text Rendering & Typography in AI Images](#part-9-text-rendering--typography-in-ai-images)
+10. [Brand Consistency Across Multiple Generations](#part-10-brand-consistency-across-multiple-generations)
+11. [Prompt Templates](#part-11-prompt-templates)
+12. [Quality Checklist](#part-12-quality-checklist)
+
+---
+
+## Part 1: Visual Psychology — What Makes People Stop
+
+### The Hierarchy of Visual Attention
+
+The human brain processes visual information in predictable patterns. Understanding these patterns is the difference between images that command attention and those that dissolve into the feed.
+
+**Processing Order:**
+1. **Faces** (highest priority — neurologically wired)
+2. **Movement / implied motion**
+3. **High contrast elements**
+4. **Unexpected / novel elements** (pattern interrupts)
+5. **Text** (last — requires conscious processing)
+
+### Why Faces Work
+
+Humans are neurologically wired to notice faces before anything else. Mirror neurons activate when we see facial expressions, making us *feel* what we see.
+
+**Effective Face Usage:**
+- **Exaggerated expressions** outperform subtle (shock, excitement, curiosity)
+- **Gaze direction matters**: Eyes looking at your CTA guide viewer attention there
+- **Close-ups** beat full-body shots for emotional impact
+- **Asymmetrical angles** create more interest than straight-on
+
+**Anti-Pattern:**
+Generic stock smile → use specific emotions that match the content's emotional payload
+
+### The 3-Second Threshold
+
+Research shows:
+- **65%** of viewers who watch the first 3 seconds continue to 10 seconds
+- **45%** will reach 30 seconds
+- First impression forms in **50 milliseconds** (before conscious thought)
+
+**Implication:** Your visual must communicate value/intrigue/emotion before conscious processing begins.
+
+### Pattern Interrupts That Stop Scrolling
+
+The brain filters out expected patterns. To stop the scroll, you need *novelty without confusion*.
+
+**Effective Pattern Interrupts:**
+- Unexpected color combinations (not random — intentional clash)
+- Scale violations (tiny person, giant object)
+- Temporal mismatch (retro aesthetic + futuristic elements)
+- Context displacement (product in impossible environment)
+- Visual juxtaposition (opposites in same frame)
+
+**Example:** Perplexity's metallic asterisk on mossy green earth texture — digital + organic = attention
+
+### High-Arousal Emotions Drive Sharing
+
+Content that triggers **high-arousal emotions** (positive or negative) gets shared more than neutral content:
+
+**High-Arousal (Shareable):**
+- Awe, wonder, amazement
+- Excitement, anticipation
+- Outrage, anger (use carefully)
+- Surprise, shock
+- Joy, delight
+
+**Low-Arousal (Less Shareable):**
+- Contentment
+- Sadness
+- Calm
+
+**Visual Translation:**
+- Awe → vast scale, impossible beauty, dramatic lighting
+- Excitement → movement, vibrant colors, dynamic composition
+- Surprise → unexpected juxtaposition, impossible scenarios
+
+---
+
+## Part 2: Direct Response Visual Principles
+
+### From Ogilvy, Schwartz, and Modern DTC
+
+The advertising legends tested what works. Here is what still applies:
+
+### Images vs. Copy: Complementary Strengths
+
+**Images communicate with FORCE:**
+- Immediate emotional impact
+- Works before conscious thought
+- Creates desire, not justification
+
+**Copy communicates with PRECISION:**
+- Specific claims and details
+- Overcomes objections
+- Provides rational scaffolding for emotional decision
+
+**Implication:** Use images to create emotional pull, copy to close the gap.
+
+### The Ogilvy Rules That Still Work
+
+1. **Finished outcomes beat processes**
+   - Show the result, not the work
+   - Image of delicious meal > image of raw ingredients
+   - After transformation > before transformation
+
+2. **Specificity beats generality**
+   - "A 42-year-old founder in Austin" > "an entrepreneur"
+   - Precise details trigger identification
+
+3. **Reader-first composition**
+   - People look at illustrations first, then headlines, then body
+   - Visual hierarchy must match this sequence
+
+4. **Authenticity over polish**
+   - Real contexts beat sterile studios
+   - Lifestyle imagery reduces psychological distance
+   - User-generated aesthetic often outperforms professional
+
+### Visual Hierarchy for Conversion
+
+```
+ATTENTION SEQUENCE:
+1. PRIMARY FOCAL POINT (the hook)
+   ↓
+2. SUPPORTING CONTEXT (why care)
+   ↓
+3. CALL TO ACTION (what to do)
+```
+
+**Hick's Law:** More choices = slower decisions = lower conversion
+- One dominant visual element
+- One clear message
+- One obvious next step
+
+### The Rule of Thirds (Actually Useful)
+
+Divide frame into 9 segments (3x3 grid). Place key elements on intersecting lines, not dead center.
+
+**Why it works:**
+- Creates visual tension and interest
+- Matches natural eye scanning patterns
+- Leaves room for composition to "breathe"
+
+**Anti-Pattern:** Centered subjects feel static and amateur
+
+---
+
+## Part 3: Anti-Generic AI Aesthetic Principles
+
+### Understanding "AI Slop"
+
+AI slop = high volume + low effort + statistical averages = forgettable
+
+**Why It Happens:**
+- Models trained to predict "most likely" output
+- Without specific direction, AI defaults to training data averages
+- Generic prompts → generic results
+- "Beautiful woman in fashionable clothing" → Eurocentric defaults, conventional poses, sterile compositions
+
+### The Generic AI Markers (Avoid These)
+
+**Visual Tells:**
+- Perfectly centered composition
+- Uniform, directionless lighting
+- Overly smooth/plastic skin texture
+- Symmetrical everything
+- "Too perfect" Disney aesthetic
+- Hands that do not look right
+- Gibberish text
+- Repetitive background elements
+- Unnatural reflections
+
+**Aesthetic Tells:**
+- High saturation + no color strategy
+- Gradient backgrounds to nowhere
+- Generic "tech blue" palette
+- Stock photo poses
+- Inspirational quote + sunset vibes
+
+### The Anti-Generic Framework
+
+**1. Strategic Specificity**
+Instead of describing what you want generically, specify:
+- Exact time of day / lighting condition
+- Specific artistic movement or reference
+- Precise color palette with hex codes or names
+- One or two distinctive anchor elements
+- Mood/emotion in specific terms
+
+**Generic:** "A professional looking office"
+**Specific:** "A warm afternoon-lit workspace with golden hour streaming through venetian blinds, casting striped shadows across a worn mahogany desk, muted earth tones, shot on medium format film"
+
+**2. Constraint as Creative Tool**
+
+Limits force originality. Add constraints:
+- "No typical poses"
+- "Avoid centered composition"
+- "Non-uniform lighting only"
+- "Imperfect textures"
+
+**3. Multi-Dimensional Style Architecture**
+
+Do not rely on single style keywords. Specify across dimensions:
+
+```
+STYLE ARCHITECTURE:
+├── Lighting (direction, quality, temperature)
+├── Color Palette (specific palette, not just "colorful")
+├── Composition (rule, framing, depth)
+├── Artistic Reference (movement, artist, era)
+├── Technical Treatment (camera, film stock, processing)
+├── Texture/Surface (material quality, imperfections)
+└── Emotional Register (specific mood, energy)
+```
+
+**Example Build:**
+- Lighting: "Backlit with rim lighting, deep shadows"
+- Color: "Desaturated rust orange and slate blue"
+- Composition: "Dutch angle, tight crop"
+- Reference: "Annie Leibovitz environmental portrait"
+- Technical: "Shot on Hasselblad, shallow depth of field"
+- Texture: "Film grain, natural skin texture"
+- Mood: "Contemplative tension, quiet intensity"
+
+**4. The Iteration Loop**
+
+Professional workflow: Generate → Critique → Refine (repeat)
+
+**Critique Questions:**
+- Did this fall into any generic patterns?
+- What is the most distinctive element?
+- What would make this impossible to ignore?
+- Does this feel intentional or accidental?
+
+---
+
+## Part 4: Anti-AI-Slop Playbook
+
+This section is specifically about avoiding the "obviously AI-generated" look that has become its own form of visual noise. As AI-generated content floods every platform, audiences have developed an intuitive ability to spot and skip AI slop. Avoiding these patterns is now a competitive requirement.
+
+### The 7 Dead Giveaways of AI Slop
+
+**1. Over-Saturated Colors**
+
+AI models tend to push colors 20-40% more saturated than reality. This creates a "candy-coated" look that screams artificial.
+
+**Fix:** Add desaturation language to prompts. Use specific, muted color names instead of generic ones.
+- Instead of: "vibrant red" → "muted terracotta" or "burgundy"
+- Instead of: "bright blue" → "slate blue" or "dusty navy"
+- Add: "desaturated color palette", "natural color grading", "muted tones"
+
+**2. Plastic Skin / Porcelain People**
+
+AI smooths skin to an uncanny perfection. Pores, texture, and natural skin variation disappear.
+
+**Fix:** Explicitly request natural skin in prompts.
+- Add: "natural skin texture", "visible pores", "realistic skin"
+- Add: "shot on film" or "film grain" (introduces natural texture)
+- Specify imperfections: "freckles", "laugh lines", "sun-weathered"
+
+**3. Too-Perfect Symmetry**
+
+Real-world photography is never perfectly symmetrical. AI defaults to bilateral symmetry because it is statistically "most likely."
+
+**Fix:** Break symmetry intentionally.
+- Add: "asymmetrical composition", "off-center subject"
+- Specify: "slightly turned head", "casual posture", "weight on one foot"
+- Use: "candid moment" rather than "portrait"
+
+**4. The AI Glow / HDR Effect**
+
+An unnatural, even illumination where everything is lit from everywhere. No real light source.
+
+**Fix:** Specify directional, motivated lighting.
+- Add: "single light source from upper left"
+- Add: "deep shadows", "chiaroscuro", "motivated lighting"
+- Specify: "overcast daylight" or "window light from the right" — real-world conditions
+- Avoid: "well-lit", "bright lighting" (these encourage flat AI lighting)
+
+**5. Generic Stock Photo Compositions**
+
+The "person smiling at camera with thumbs up" or "diverse group of professionals in meeting room" compositions that feel pre-canned.
+
+**Fix:** Describe specific, narrative moments.
+- Instead of: "team in meeting" → "founder sketching on whiteboard at 2am, coffee cups scattered, laptop glow on face"
+- Instead of: "happy customer" → "woman mid-laugh at a cafe, looking at her phone, afternoon light"
+- The more specific the narrative moment, the less generic the output
+
+**6. Impossible Backgrounds / Object Soup**
+
+Background elements that dissolve into vague shapes, or objects that have no real-world logic.
+
+**Fix:** Describe the background as specifically as the foreground.
+- Add: "on weathered concrete surface" not just "on surface"
+- Add: "brick wall with peeling paint in background" not just "urban background"
+- Every object in the scene should have a reason to be there
+
+**7. The "AI Art Station" Aesthetic**
+
+An overly polished, hyper-rendered look that blends concept art, CGI, and illustration into a generic "AI art" house style. Too many highlights, too much bloom, too many particle effects.
+
+**Fix:** Ground the image in a real-world photographic reference.
+- Add camera references: "shot on Fujifilm X-T5", "35mm film photography"
+- Add processing references: "minimal post-processing", "straight out of camera"
+- Avoid: "ultra-detailed", "hyper-realistic", "8K" — these push toward the overprocessed AI look
+- Instead use: "professional photography", "editorial quality", "commercial photography"
+
+### The Authenticity Spectrum
+
+```
+MOST AUTHENTIC ◄──────────────────────────────► MOST ARTIFICIAL
+
+Film grain        Professional    HDR/saturated    Hyper-rendered
+Natural texture   photography     Everything lit    AI Art Station
+Visible wear      Clean but real  Plastic smooth    Fantasy bloom
+Candid moment     Posed natural   Stock pose        Perfect symmetry
+```
+
+**Target the left side.** As AI content floods the right side, authenticity becomes the differentiator.
+
+### Quick Anti-Slop Prompt Additions
+
+Add these to any prompt to reduce AI-obvious markers:
+
+```
+Anti-slop base layer (pick 2-3 relevant ones):
+- "natural color grading, not over-saturated"
+- "realistic lighting with visible shadows"
+- "shot on [specific camera], [specific lens]"
+- "natural skin texture" (for people)
+- "slight film grain"
+- "asymmetrical composition"
+- "candid, not posed"
+- "environmental context, not studio"
+- "minimal post-processing"
+```
+
+---
+
+## Part 5: Style Vocabulary Taxonomy
+
+### 2025-2026 Commercial AI Aesthetics
+
+These are the dominant styles being used commercially right now:
+
+### Tier 1: Chrome & Metallic
+
+**What It Is:** Reflective metallic surfaces, liquid metal, chrome textures
+**Communicates:** Premium, technological, futuristic
+**Best For:** Luxury brands, tech, products
+
+**Keywords:**
+```
+chrome, liquid metal, metallic, mercury-like, reflective surface,
+polished chrome, brushed aluminum, iridescent metal, holographic chrome,
+high-gloss metallic, molten silver, mercury droplet, chrome sculpture
+```
+
+**Technical Notes:**
+- Requires precise lighting specification
+- Include background that will reflect convincingly
+- Works best against dark or minimal backgrounds
+
+**Example Prompt Fragment:**
+"...rendered as sculptural chrome object with liquid mercury reflections, dramatic studio lighting creating sharp highlights and deep shadows, set against matte black surface..."
+
+---
+
+### Tier 2: Organic & Biomorphic
+
+**What It Is:** Nature-derived forms, flowing organic shapes, biomimetic textures
+**Communicates:** Authenticity, sustainability, warmth, growth
+**Best For:** Wellness, sustainability brands, humanizing tech
+
+**Keywords:**
+```
+organic shapes, biomorphic, flowing forms, cellular structure,
+natural texture, moss, lichen, botanical elements, coral-like,
+fluid organic, living material, growth pattern, natural imperfection
+```
+
+**The Perplexity Approach:**
+Combine organic (mossy green textures) with technological (chrome asterisk) — creates tension that captures attention.
+
+---
+
+### Tier 3: Surrealism & Impossibility
+
+**What It Is:** Dream logic, impossible scenarios, scale violations
+**Communicates:** Creativity, boldness, memorability
+**Best For:** Fashion, luxury, brands that can take creative risks
+
+**Keywords:**
+```
+surreal, impossible scene, dream-like, scale distortion,
+floating objects, defying physics, melting forms, M.C. Escher,
+Magritte-inspired, visual paradox, uncanny valley (intentional)
+```
+
+**Brand Examples:**
+- Jacquemus: Giant handbags crashing into city streets
+- Tod's: Shoes as actual cars
+- Victoria Beckham: Floating bags from landmarks
+
+---
+
+### Tier 4: 3D Photorealistic Rendering
+
+**What It Is:** Product visualization indistinguishable from photography
+**Communicates:** Precision, quality, trust
+**Best For:** E-commerce, product marketing, catalogs
+
+**Keywords:**
+```
+photorealistic 3D render, product visualization, studio lighting,
+seamless environmental integration, physically accurate materials,
+ray traced, subsurface scattering, realistic reflections,
+product photography style, commercial product shot
+```
+
+**Critical Factors:**
+- Material specification is essential
+- Lighting must match environment
+- Include surface imperfections for realism
+
+---
+
+### Tier 5: Abstract & Neural
+
+**What It Is:** Data visualization, network forms, computational aesthetics
+**Communicates:** Intelligence, innovation, complexity
+**Best For:** Tech companies, AI, data products
+
+**Keywords:**
+```
+neural network visualization, abstract data flow, bioluminescent,
+interconnected nodes, glowing pathways, digital synapse,
+information architecture, generative pattern, procedural forms,
+electric blue glow, computational aesthetic
+```
+
+---
+
+### Tier 6: Retro-Futurism & Synthwave
+
+**What It Is:** 1980s aesthetics + futuristic elements
+**Communicates:** Nostalgia + innovation, playful sophistication
+**Best For:** Gen Z/Millennial targeting, gaming, entertainment
+
+**Keywords:**
+```
+synthwave, retro-futurism, neon grid, vaporwave,
+80s gradient, chrome text, sunset palette, outrun aesthetic,
+retrofuture, cyberpunk neon, digital sunset, cassette futurism
+```
+
+---
+
+### Tier 7: Minimalist Premium
+
+**What It Is:** Extreme restraint, negative space, typography-forward
+**Communicates:** Sophistication, confidence, luxury
+**Best For:** Premium positioning, design-forward brands
+
+**Keywords:**
+```
+minimal composition, negative space, typography dominant,
+restrained palette, single accent color, architectural composition,
+Swiss design, editorial minimalism, luxury restraint
+```
+
+**The Linear Approach:**
+Desaturated single color, extreme breathing room, confidence through restraint.
+
+---
+
+### Tier 8: Tactile & Material (Emerging 2025-2026)
+
+**What It Is:** Emphasis on physical material qualities — fabric, clay, ceramic, concrete, wood grain. Making digital feel physical.
+**Communicates:** Craft, quality, tangibility, anti-digital
+**Best For:** DTC products, premium goods, brands that want to feel "real" in a sea of digital
+
+**Keywords:**
+```
+tactile texture, hand-made quality, raw ceramic, linen texture,
+concrete surface, aged patina, natural wood grain, hand-thrown pottery,
+wabi-sabi, imperfect beauty, material honesty, craft aesthetic
+```
+
+**Why It Is Emerging:**
+As AI-generated content floods feeds with hyper-polished digital imagery, audiences crave the opposite: things that look and feel handmade, physical, imperfect. This is the counter-trend to AI slop.
+
+---
+
+## Part 6: Leading Brand Visual Identities
+
+### What They Do and Why It Works
+
+### Perplexity AI
+
+**Visual Identity:**
+- Warm earth tones + techy dark blue
+- 3D metallic asterisk (sculptural, tactile)
+- Intellectual positioning through quotes
+- Texture contrast (digital meets organic)
+
+**Why It Works:**
+- Unexpected for an AI company (not cold/clinical)
+- The asterisk doubles as metaphor (sources, citations)
+- Creates tension that captures attention
+
+**Style Keywords:** `techy dark blue, warm earth tones, metallic sculpture, organic texture, intellectual minimal`
+
+---
+
+### Anthropic Claude
+
+**Visual Identity:**
+- Rust orange (#C15F3C) — notably NOT tech blue
+- Extreme restraint and simplicity
+- Rounded, humanistic typography
+- Starburst icon suggesting ideas radiating
+
+**Why It Works:**
+- Differentiates from every other AI company
+- Warmth communicates safety/trust
+- Restraint = confidence
+
+**Style Keywords:** `rust orange, warm neutral, humanistic minimal, restrained sophistication, trust through warmth`
+
+---
+
+### Figma
+
+**Visual Identity:**
+- Dynamic primitives (jumbo shapes)
+- Motion as core expression
+- Expanded color palette (primaries, neons, earth tones)
+- Figma Sans custom typeface
+
+**Motion Principles:**
+- **Shift:** Diverse perspectives converging
+- **Transform:** Ongoing change
+- **Amplify:** Tools scaling vision
+- **Activate:** Collaborative energy
+
+**Style Keywords:** `vibrant primitives, dynamic motion, jumbo scale, collaborative energy, shape-based language`
+
+---
+
+### Notion
+
+**Visual Identity:**
+- Profile illustrations (showing thinking, not working)
+- Bold primary colors (yellow, blue, red) for OOH impact
+- Cubism influences (Picasso, Saul Steinberg)
+- "Think it. Make it." positioning
+
+**Why It Works:**
+- Moved from showing product to showing mindset
+- Profile view = intimacy, interiority
+- Colors punch through urban visual noise
+
+**Style Keywords:** `profile illustration, primary colors, cubist influence, interior thinking, gestural linework`
+
+---
+
+### Linear
+
+**Visual Identity:**
+- Single desaturated blue
+- Extreme minimalism
+- Dark mode native
+- Typography does the work
+
+**Why It Works:**
+- Restraint = sophistication for developer audience
+- No noise = clarity
+- Confidence through limitation
+
+**Style Keywords:** `desaturated blue, extreme restraint, dark mode, typographic hierarchy, developer aesthetic`
+
+---
+
+### Vercel
+
+**Visual Identity:**
+- Grid-based precision
+- Geist typeface family (custom)
+- High contrast, systematic color
+- Triangle logo (V = Vercel)
+
+**Why It Works:**
+- Design system IS the brand
+- Precision communicates engineering excellence
+- Grid = structure = reliability
+
+**Style Keywords:** `grid system, geometric precision, high contrast, systematic, engineering aesthetic`
+
+---
+
+## Part 7: Platform-Specific Visual Strategy
+
+### Instagram Feed (2025-2026 Data)
+
+**What Performs:**
+- Carousels: 0.55% engagement rate (highest format)
+- Reels shared to feed outperform static images 2-3x
+- Multi-image storytelling continues to dominate
+- Educational sequences with bold text on first slide
+
+**Visual Strategy:**
+- Consistent color palette across carousel slides
+- Rule of thirds composition
+- High contrast for feed visibility
+- Hook on first slide — value promise or pattern interrupt
+- 4:5 aspect ratio occupies maximum feed real estate
+
+**Aspect Ratio:** `4:5` for feed posts (maximum height in feed), `1:1` for cross-platform reuse
+
+**Anti-Pattern:** Single static images without story arc. Carousels outperform by 2-3x.
+
+---
+
+### Instagram Stories & Reels
+
+**What Performs:**
+- 60-90 second Reels (not shorter — algorithm rewards completion of longer content)
+- Trending sounds increase discovery 30-40%
+- Pattern interrupt opening (first 0.5 seconds)
+- UGC aesthetic outperforms polished production
+
+**Visual Strategy:**
+- Movement in first frame (no static opens)
+- Close-up faces with emotion
+- Bold text overlays for silent viewing
+- High contrast color, especially in the first second
+
+**Aspect Ratio:** `9:16` always
+
+**The Hook:** You have 0.5-1 second for stories, 3 seconds for Reels.
+
+---
+
+### YouTube Thumbnails
+
+**What Performs:**
+- Click-through rate is the primary discovery metric
+- Faces with exaggerated expressions: 25-40% higher CTR than no face
+- 3-5 words maximum text
+- Complementary colors to title (not matching)
+
+**Visual Strategy:**
+- Background blur/darken to isolate subject
+- Complementary (not matching) colors to video title
+- Curiosity gap (show result, hide method)
+- Arrows/circles sparingly — overuse has reduced their effectiveness
+- Before/after split compositions perform well for tutorial content
+
+**Aspect Ratio:** `16:9` (1280x720 minimum, 1920x1080 preferred)
+
+**3-Second Test:** Can a stranger understand the value from the thumbnail alone?
+
+---
+
+### LinkedIn (2025-2026 Data)
+
+**What Performs:**
+- Multi-image carousels: 6.6% engagement rate
+- Document posts (PDF carousels): 6.1%
+- Native video: 5.6%
+- Poll posts: 5.8%
+- Text + single image: 3.2%
+
+**Visual Strategy:**
+- Professional but not corporate (the line has shifted — authenticity wins)
+- Data visualization and infographics
+- Minimal, clean composition
+- Thought leadership aesthetic — "executive with ideas" not "corporate with stock photos"
+- Dark backgrounds perform well in dark mode (LinkedIn's growing default)
+
+**Aspect Ratio:** `1:1` for posts, `16:9` for article headers and link previews
+
+**Color Psychology:**
+- Dark blues = trust, professionalism
+- White space = sophistication
+- Minimal accent colors (one is enough)
+
+---
+
+### TikTok
+
+**What Performs:**
+- Hook in first 0.3 seconds (even faster than Instagram)
+- Authenticity dramatically outperforms polish
+- Text-on-screen is mandatory (80%+ watch with sound off at some point)
+- Pattern interrupts: physical movement, unexpected visuals, direct-to-camera address
+
+**Proven Hook Types:**
+1. **Pattern interrupt** — unexpected visual/audio in first frame
+2. **List format** — "3 ways to..."
+3. **Curiosity gap** — incomplete information that demands completion
+4. **Question** — voices viewer's internal thought
+5. **Pain point** — "tired of..."
+6. **Bold claim** — provocative assertion that demands attention
+7. **Story tease** — narrative setup that promises payoff
+
+**Visual Strategy:**
+- Vertical native (9:16) — never letterboxed
+- Movement in first frame
+- Face + emotion (UGC feel)
+- Text overlays: large, high contrast, positioned in safe zone (center 80%)
+
+**Aspect Ratio:** `9:16` exclusively
+
+---
+
+### X / Twitter
+
+**What Performs:**
+- Native video: 10x engagement vs text-only
+- Videos under 2:20 that achieve high completion rate = algorithm reward
+- Replies valued 13-27x higher than likes for distribution
+- Quote tweets with added visual context outperform retweets
+- Image posts with 1-2 images outperform 3-4 image layouts
+
+**Visual Strategy:**
+- High contrast for dark mode (now the majority of users)
+- Text legibility at small sizes — test at mobile preview scale
+- Quick visual communication — Twitter is a speed medium
+- Bold, graphic imagery over photographic subtlety
+
+**Aspect Ratio:** `16:9` for images and video, `1:1` also supported
+
+---
+
+### Pinterest (2025-2026 Data)
+
+**What Performs:**
+- Tall pins (2:3 ratio) occupy maximum visual space in the feed
+- Step-by-step instructional images
+- Text overlay with clear benefit statement
+- Warm, aspirational photography (Pinterest audience skews toward planning/aspiration)
+
+**Visual Strategy:**
+- Tall vertical composition with clear visual hierarchy
+- Product in lifestyle context (not isolated on white)
+- Warm, bright, inviting color palettes
+- Text at top: benefit/title. Image fills remaining space.
+- Seasonal and trend-aligned content gets amplified
+
+**Aspect Ratio:** `2:3` (1000x1500 optimal)
+
+---
+
+## Part 8: Prompt Construction for the configured image model
+
+the configured image model (see brand/stack.md) is the default image model for the creative engine. These guidelines are specific to how this model interprets prompts.
+
+### The Prompt Formula
+
+```
+[SUBJECT with specific details] + [SETTING with environmental context] +
+[STYLE reference] + [LIGHTING with direction and quality] +
+[COLOR PALETTE with specific names] + [MOOD in precise emotional terms] +
+[TECHNICAL reference for texture/processing]
+```
+
+### How the configured image model Reads Prompts
+
+**What it does well:**
+- Follows detailed natural language descriptions faithfully
+- Handles multiple style references without confusion
+- Renders text and typography better than most models
+- Understands camera and lens references (translates to appropriate depth of field, perspective, etc.)
+- Responds well to specific color names and palettes
+- Handles complex multi-element compositions
+
+**What it struggles with:**
+- Very long prompts (500+ words) — keep under 300 words for reliability
+- Abstract concepts without visual anchors — always give it something concrete
+- Exact spatial relationships between many objects — keep compositions to 2-3 key elements
+- Consistent character appearance across multiple generations (use brand kit techniques below)
+- Hands in complex poses — keep hand references simple or crop them out
+
+### Subject Specificity Guide
+
+The more specific the subject, the less the model guesses. Guessing = generic.
+
+```
+WEAK ──────────────────────────────────── STRONG
+
+"a coffee cup"                            "handmade ceramic coffee cup with
+                                           matte sage green glaze, slight
+                                           irregularity in the rim, latte
+                                           art rosetta visible, wisps of
+                                           steam catching the light"
+
+"a laptop"                                "silver MacBook Pro on aged oak
+                                           desk, screen showing a code editor
+                                           with dark theme, desk lamp casting
+                                           warm pool of light from the left"
+
+"a person working"                        "woman in her mid-30s, hair loosely
+                                           tied back, reading glasses pushed
+                                           up on forehead, mid-thought with
+                                           pen to lips, afternoon window
+                                           light catching dust particles"
+```
+
+### Lighting as Mood Control
+
+Lighting is the single most impactful prompt element after the subject. the configured image model responds particularly well to specific lighting descriptions.
+
+| Lighting Description | Mood It Creates | When to Use |
+|---------------------|-----------------|-------------|
+| "soft natural daylight from a window" | Calm, authentic, approachable | Lifestyle, editorial |
+| "golden hour light, long shadows" | Warm, nostalgic, aspirational | Brand storytelling, lifestyle |
+| "single dramatic spotlight from upper left" | Premium, focused, confident | Product hero shots |
+| "overcast diffused light, no harsh shadows" | Even, clean, professional | E-commerce, product catalog |
+| "backlit with rim lighting, deep shadows" | Dramatic, mysterious, high-end | Fashion, luxury, tech |
+| "neon accent lighting, cyan and magenta" | Energetic, modern, nightlife | Entertainment, youth brands |
+| "chiaroscuro, dramatic light-dark contrast" | Artistic, powerful, serious | Editorial, thought leadership |
+| "flat even studio lighting" | Clean, commercial, no drama | Product documentation, UI mockups |
+
+### Color Palette Integration
+
+Never say "colorful" or "vibrant colors." Always specify the palette.
+
+**Effective patterns:**
+```
+"warm palette of terracotta (#C4704A), cream (#F5F0E8), and deep forest (#1B3A2D)"
+"monochromatic blue palette ranging from #0A1628 to #4A9BD9"
+"complementary contrast of rust orange and cool slate blue"
+"muted earth tones: sand, warm grey, dried sage, soft clay"
+```
+
+**the configured image model understands hex codes** in prompts. Using them produces more precise color results than generic color names alone.
+
+### Quality and Realism Modifiers
+
+**For photographic realism:**
+```
+"shot on Fujifilm X-T5 with 56mm f/1.2 lens"
+"medium format film photography, slight grain"
+"editorial photography, Condé Nast quality"
+"commercial product photography, Phase One"
+```
+
+**For illustration/stylized:**
+```
+"digital illustration, clean vector style"
+"watercolor with visible brush strokes and paper texture"
+"risograph print, slight misregistration, textured paper"
+"Japanese woodblock print aesthetic, limited color palette"
+```
+
+**Avoid these overused quality boosters** (they push toward AI slop):
+- "ultra-detailed" — pushes toward over-rendered
+- "8K resolution" — does nothing useful, triggers hyper-processing
+- "hyper-realistic" — uncanny valley territory
+- "masterpiece" — generic booster with no specific meaning
+- "best quality" — same as "masterpiece"
+
+**Use instead:**
+- "professional photography" — grounded in real-world standard
+- "editorial quality" — implies taste and curation
+- "commercial campaign quality" — implies real-world usage standard
+- "[specific photographer] style" — references a real aesthetic
+
+---
+
+## Part 9: Text Rendering & Typography in AI Images
+
+the configured image model has significantly better text rendering capabilities than previous generation models. However, text in AI-generated images still requires specific techniques to get clean results.
+
+### When to Render Text in the Image vs. Overlay Later
+
+| Scenario | Approach | Why |
+|----------|----------|-----|
+| Text is integrated into the scene (signs, labels, packaging) | Render in image | Needs to interact with lighting, perspective, materials |
+| Text is a UI element or overlay (headline, CTA, caption) | Add in post-production | More control over font, sizing, positioning |
+| Short text (1-4 words) | Render in image | Model handles short text well |
+| Long text (5+ words, paragraphs) | Add in post-production | Error rate increases with text length |
+| Stylized/decorative text (neon signs, graffiti, embossed) | Render in image | Style integration requires generation |
+
+### Best Practices for Text Rendering
+
+**Keep it short.** 1-4 words render cleanly. Each additional word increases the chance of errors.
+
+**Specify the text exactly in quotes:**
+```
+Good: 'large neon sign reading "OPEN" in red cursive'
+Bad:  'a sign that says something about being open'
+```
+
+**Describe the typography style:**
+```
+'bold sans-serif text reading "LAUNCH" in white, condensed uppercase'
+'elegant serif text reading "Artisan" in gold foil on dark leather'
+'hand-painted vintage sign reading "CAFE" in cream on weathered wood'
+```
+
+**Describe the text material and surface:**
+```
+'chrome 3D letters spelling "PRO" with metallic reflections'
+'embossed text reading "PREMIUM" pressed into thick cotton paper'
+'frosted glass text reading "HELLO" backlit with soft blue light'
+```
+
+### Common Text Rendering Issues and Fixes
+
+| Issue | Fix |
+|-------|-----|
+| Letters are swapped or duplicated | Regenerate. This is stochastic — same prompt may work on next attempt. |
+| Text is partially legible | Shorten the text. Remove any word that is not essential. |
+| Wrong font style | Be more specific: "bold geometric sans-serif" not just "modern font" |
+| Text does not interact with lighting | Add: "text catching the same [lighting description] as the scene" |
+| Text floats unnaturally | Specify the surface: "text painted on the wall", "text engraved in the surface" |
+
+### Typography Style Keywords for Image Generation
+
+```
+Serif:        "elegant serif", "classic roman", "editorial serif", "high-contrast serif"
+Sans-serif:   "clean sans-serif", "geometric sans", "humanist sans", "condensed gothic"
+Display:      "bold display type", "compressed headline", "ultra-black", "fat face"
+Script:       "flowing script", "casual handwriting", "formal calligraphy", "brush script"
+Decorative:   "vintage lettering", "art deco type", "psychedelic lettering", "stencil"
+Technical:    "monospace", "terminal font", "code editor typeface", "engineering stencil"
+```
+
+---
+
+## Part 10: Brand Consistency Across Multiple Generations
+
+The biggest challenge in AI creative production is maintaining visual consistency across multiple generations. Each API call is independent — the model has no memory of previous outputs. These techniques create consistency without model memory.
+
+### The Brand Prompt Prefix
+
+Create a standard prefix that is prepended to every prompt for a project:
+
+```
+Brand prefix (example):
+"In the [Brand Name] visual style: warm terracotta (#C4704A) and cream (#F5F0E8)
+color palette, soft natural lighting, minimal composition with generous negative
+space, shot on medium format film with subtle grain, editorial quality. "
+```
+
+This prefix goes at the START of every prompt. It anchors the model to a consistent aesthetic before the specific subject description begins.
+
+### Style Lock Document
+
+After the 5-direction exploration process (described in SKILL.md), create a style lock document:
+
+```markdown
+## Style Lock: [Project Name]
+
+### Color Anchors
+- Primary surface: [color + hex]
+- Accent: [color + hex]
+- Background: [color + hex]
+- Mood modifier: [color + hex]
+
+### Lighting Lock
+- Direction: [e.g., "upper left, 45 degrees"]
+- Quality: [e.g., "soft diffused, no harsh shadows"]
+- Temperature: [e.g., "warm, 4000K equivalent"]
+
+### Composition Rules
+- Framing: [e.g., "rule of thirds, subject on left intersection"]
+- Negative space: [e.g., "30% minimum, right side for text overlay"]
+- Depth: [e.g., "shallow depth of field, f/2.8 equivalent"]
+
+### Texture/Processing
+- Base: [e.g., "slight film grain, natural skin texture"]
+- Color grade: [e.g., "warm shadows, cool highlights, low contrast"]
+- Avoid: [e.g., "HDR look, over-saturation, plastic smoothing"]
+
+### Prompt Template
+[The exact prompt structure with {{variables}} for subject matter]
+```
+
+### Consistency Techniques
+
+**1. Anchor Colors with Hex Codes**
+
+Modern image models understand hex codes. Using them produces more consistent color results than descriptive names.
+
+```
+Less consistent: "warm orange and cream"
+More consistent: "terracotta #C4704A and cream #F5F0E8"
+```
+
+**2. Consistent Lighting Description**
+
+Use the exact same lighting phrase across all prompts. Do not paraphrase.
+
+```
+Locked phrase: "soft natural daylight from a large window on the left, warm tone"
+
+Prompt 1: "Product bottle [locked lighting phrase], on marble countertop..."
+Prompt 2: "Model wearing jacket [locked lighting phrase], in studio space..."
+Prompt 3: "Close-up of texture [locked lighting phrase], on linen surface..."
+```
+
+**3. Consistent Camera/Technical Reference**
+
+The same camera reference produces similar depth of field, grain, and tonal characteristics.
+
+```
+Locked technical: "shot on Hasselblad X2D, 90mm f/3.2, slight film emulation"
+```
+
+Use this exact phrase in every prompt for the project.
+
+**4. Seed Reuse for Variations**
+
+When you find a generation you like, note the seed. Use that seed with modified prompts to get variations that share a similar "feel" (though this is not guaranteed — it increases consistency probability, not certainty).
+
+**5. Reference Image Workflow**
+
+For maximum consistency, use the image editing capability of your configured image model:
+1. Generate the first image and approve it
+2. For subsequent images, provide the approved image as `image_input` with a prompt describing the variation
+3. This anchors the model to the established visual style
+
+### What Breaks Consistency
+
+- Changing the lighting description between prompts
+- Using different color names for the same color
+- Mixing camera references ("shot on iPhone" in one, "Hasselblad" in another)
+- Adding conflicting style keywords ("warm" in one prompt, "cool-toned" in another)
+- Omitting the brand prefix from any prompt
+
+---
+
+## Part 11: Prompt Templates
+
+### Template 1: Anti-Generic Product Shot
+
+```
+A [product with specific material/color/detail] positioned [composition],
+[lighting with specific direction and quality],
+in [specific environment, not generic studio],
+[color palette with specific named colors or hex codes],
+[mood in specific emotional terms],
+shot on [camera reference for texture/depth],
+natural color grading, not over-saturated,
+professional commercial photography
+```
+
+### Template 2: Editorial Portrait
+
+```
+[Subject with specific details — age, expression, posture],
+[lighting technique — Rembrandt, butterfly, split, etc.],
+[composition — tight crop, medium, environmental],
+[color palette — specific names],
+[artistic reference — photographer or painter style],
+[texture — film grain, natural skin texture],
+[mood — specific emotional quality],
+natural skin texture, visible pores, asymmetrical composition
+```
+
+### Template 3: Brand-Aligned Lifestyle
+
+```
+[Specific scene with narrative details],
+featuring [product naturally integrated into the moment],
+[time of day and lighting condition],
+[brand color palette with hex codes],
+[composition style],
+[atmosphere and mood],
+candid, authentic feel, not posed or staged,
+editorial quality, [reference photographer or publication style]
+```
+
+### Template 4: Surrealist Campaign
+
+```
+[Impossible scenario described with precise visual detail],
+photorealistic rendering of [specific impossibility],
+[lighting maintains physical consistency despite the impossibility],
+[brand colors and product integrated naturally],
+[scale or physics violation clearly specified],
+dream-like clarity, sharp focus throughout,
+[artistic reference — Magritte, contemporary surrealist, etc.],
+commercial campaign quality
+```
+
+### Template 5: Social Graphic with Text
+
+```
+[Background scene or abstract composition],
+bold [typography style] text reading "[EXACT TEXT]" in [color],
+[text placement: centered, upper third, lower third],
+[brand color palette],
+[composition leaving appropriate space for text],
+clean, legible, designed for [platform] at mobile viewing size,
+graphic design quality, not photographic
+```
+
+### Template 6: Product E-Commerce Flat
+
+```
+[Product with specific material and color details],
+on [clean surface: pure white, light grey gradient, marble, etc.],
+[overhead/slightly elevated angle],
+soft diffused studio lighting, no harsh shadows,
+clean minimal composition, product fills 60-70% of frame,
+[specific detail callouts visible],
+commercial product photography, e-commerce standard,
+natural color accuracy, not enhanced
+```
+
+---
+
+## Part 12: Quality Checklist
+
+Before approving any visual:
+
+### Distinctiveness
+- [ ] Would this stop my scrolling?
+- [ ] Does this pass the "Is this AI?" test? (it should not look obviously generated)
+- [ ] Is there a distinctive visual element I can describe?
+- [ ] Would I know this brand from the visual alone?
+- [ ] Does this avoid every item in the Anti-AI-Slop list?
+
+### Technical Quality
+- [ ] Is composition intentional (not centered by default)?
+- [ ] Is lighting directional and motivated (not flat AI glow)?
+- [ ] Are there natural imperfections (texture, grain, variation)?
+- [ ] Do materials look physically accurate?
+- [ ] Are colors natural (not over-saturated)?
+- [ ] Is text (if present) legible and correctly spelled?
+
+### Strategic Alignment
+- [ ] Does this match the brand's creative-kit.md?
+- [ ] Is the emotional register appropriate for the message?
+- [ ] Does this serve the communication goal?
+- [ ] Is this differentiated from competitors' visuals?
+
+### Platform Optimization
+- [ ] Right aspect ratio for target platform?
+- [ ] Will this work at thumbnail/preview size?
+- [ ] Does it hook within 3 seconds (video) or 0.5 seconds (image in feed)?
+- [ ] Is text legible at mobile size?
+- [ ] Does it work in both light and dark mode (if platform supports both)?
+
+### Brand Consistency
+- [ ] Colors match the brand palette (check hex codes)?
+- [ ] Lighting matches the style lock?
+- [ ] Camera/technical treatment is consistent with other assets?
+- [ ] This could sit alongside other project assets without looking like a different brand?
+
+---
+
+## Quick Reference: Style Keywords by Vibe
+
+### "Premium & Sophisticated"
+```
+editorial, luxury restraint, negative space, architectural composition,
+muted palette, natural materials, subtle texture, confident minimalism
+```
+
+### "Warm & Approachable"
+```
+golden hour, warm tones, natural light, organic texture, authentic,
+lifestyle context, imperfect beauty, human moments, candid
+```
+
+### "Bold & Disruptive"
+```
+high contrast, unexpected color, surreal element, scale violation,
+provocative, pattern interrupt, visual tension, audacious
+```
+
+### "Technical & Innovative"
+```
+chrome, metallic, neural, data flow, precision, systematic,
+high-tech material, engineered, computational aesthetic
+```
+
+### "Nostalgic & Emotional"
+```
+film grain, vintage color grade, soft focus, warm memory,
+analog texture, timeless, gentle light, emotional resonance
+```
+
+### "Tactile & Handmade"
+```
+raw ceramic, linen texture, natural wood grain, hand-thrown,
+wabi-sabi, craft aesthetic, material honesty, visible process marks,
+imperfect edges, artisan quality, analog warmth
+```
+
+---
+
+## Implementation Notes
+
+This document is a reference, not a rigid system. The best creative work happens when you:
+
+1. **Start with the communication goal** — What needs to be felt/understood?
+2. **Choose aesthetic direction intentionally** — Match style to strategy
+3. **Specify across multiple dimensions** — Do not rely on single keywords
+4. **Apply anti-slop techniques by default** — Every prompt should include at least 2-3 anti-slop modifiers
+5. **Iterate aggressively** — Generate, critique, refine
+6. **Question your defaults** — The first idea is rarely the most distinctive
+7. **Maintain brand consistency** — Use the brand prefix, style lock, and consistent technical references
+
+The goal is not to follow these rules. It is to understand the principles deeply enough to break them intentionally.
+
+---
+
+*This reference is part of the Vibe Marketing Skills v2 creative engine. Use alongside SKILL.md for creative production and MODEL_REGISTRY.md for API calls.*