@hienlh/ppm 0.9.81 → 0.9.83

package/.opencode/skills/video/references/storyboard-format.md

@@ -1,255 +0,0 @@
-# Storyboard Format Guide
-
-Create storyboards with START/END frames for Imagen 4 to Veo 3.1 pipeline with AI-powered validation.
-
-## General Direction
-- Express clear visual narrative flow, concept, and general art direction of the video (pick one from the available art directions in `./video-art-directions.md`)
-- Describe the emotional arc and pacing of the video, including key emotional beats
-- Define the overall tone and mood of the video
-- Establish clear visual hierarchy and focal points
-- Maintain consistent art direction throughout the storyboard
-- Ensure clear visual storytelling with strong composition
-- Plan for smooth transitions between frames
-- Keep text annotations minimal but descriptive and focused on key elements
-- Use consistent scale and perspective across all frames
-- Maintain consistent character proportions and poses where applicable
-- Ensure lighting and mood are consistent across frames
-- Avoid major compositional changes between start and end frames
-
-## Core Concept: Frame-Based Generation
-
-Quality principle: Generate START and END frames with Imagen 4, then use Veo to create video between them.
-
-```
-START FRAME (Imagen 4) --> [REVIEW] --> VIDEO (Veo 3.1) --> [REVIEW] --> END FRAME (Imagen 4)
-```
-
-## Per-Scene Structure
-
-Each scene requires:
-- **Start Frame**: Imagen prompt (Art style: Colored Sketch + Handwriting Text Annotations) + validation criteria
-- **End Frame**: Imagen prompt (Art style: Colored Sketch + Handwriting Text Annotations) + validation criteria
-- **Audio**: Voiceover text (Include tone and pacing instructions), music cue, SFX (at exact timecodes)
-- **Motion**: Veo directive
-- **Review checklist**: Quality gates
-
-## Frame Art Style
-**Choose One Bold Aesthetic Direction (commit 100%)**
-| Style Category              | Core Keywords (copy-paste ready)                                      | Color Palette Ideas                                      | Signature Effects & Details                                                                      |
-|-----------------------------|----------------------------------------------------------------------|----------------------------------------------------------|--------------------------------------------------------------------------------------------------|
-| Colored Sketch + Notes      | colored pencil sketch, rough linework, storyboard, handwritten annotations, arrows, callouts, animatic style | warm neutrals + accent color (teal/red), off-white paper  | visible construction lines, margin notes, circled highlights, scribble arrows, paper texture     |
-| Clean Line Art (Flat Color) | clean line art, vector-like outlines, flat colors, minimal shading, modern storyboard | neutral base + 1-2 brand accents, soft grays             | uniform stroke weight, crisp silhouettes, simple shadows, clear readability                       |
-| Cinematic Film Still        | cinematic, film still, realistic lighting, depth of field, anamorphic bokeh, 35mm | moody teal & orange, deep blacks, subtle highlights      | lens flares, film grain, letterbox framing, dramatic key light + rim light                        |
-| High-Contrast Noir          | noir, high contrast, chiaroscuro, hard shadows, dramatic composition, gritty | black/white + single accent (red or yellow)              | venetian blind shadows, rain streaks, silhouettes, smoky haze                                     |
-| Comic Book Inked            | comic book, inked outlines, halftone, bold shadows, dynamic poses, panel-ready | primary colors, CMYK pop, off-white paper                | halftone dots, speed lines, speech bubble placeholders, thick inks                                |
-| Anime / Manga               | anime style, manga linework, expressive faces, cel shading, clean highlights | pastel brights, high-saturation accents, soft gradients  | speed lines, stylized eyes, simplified backgrounds, impact frames                                 |
-| Stylized 3D (Pixar-like)    | stylized 3d, soft global illumination, smooth shapes, friendly proportions, render | warm vibrant palette, soft bounce light                  | subsurface skin glow, soft rim light, clean materials, subtle depth fog                            |
-| Watercolor Wash             | watercolor, loose brush, washed pigments, soft edges, hand-painted, artistic | muted pastels, earthy tones, cool blues                  | paint blooms, paper grain, imperfect edges, layered washes                                         |
-| Minimal Monochrome          | minimalist, monochrome, grayscale, simple shapes, negative space, clean composition | grayscale ramp (charcoal to light gray)                  | strong staging, simplified props, value-based readability, diagram-like clarity                   |
-| Retro 80s Synthwave         | synthwave, retro 80s, neon glow, vaporwave, grid horizon, futuristic | magenta/cyan/purple, neon highlights, dark navy          | chromatic aberration, glow bloom, scanlines, gradient skies                                        |
-
-## Scene Template
-
-```json
-{
-  "scene_num": 1,
-  "timing": "0:00-0:02",
-  "start_frame": {
-    "prompt": "Wide shot of..., professional lighting, clean background",
-    "style_tags": ["photorealistic", "4k", "professional"],
-    "validation": {
-      "required_elements": ["product", "background"],
-      "forbidden": ["text", "watermarks"],
-      "quality_threshold": 8
-    }
-  },
-  "end_frame": {
-    "prompt": "Same scene, product rotated 45 degrees...",
-    "style_tags": ["photorealistic", "4k", "professional"],
-    "validation": {
-      "required_elements": ["product", "background"],
-      "continuity_with_start": true,
-      "quality_threshold": 8
-    }
-  },
-  "motion": "smooth rotation, camera static",
-  "audio": {
-    "voiceover": "Introducing the all-new...",
-    "music": "upbeat corporate, 120bpm",
-    "sfx": []
-  },
-  "review_notes": "Check product visibility, no distortion"
-}
-```
-
-## Shot Types
-
-| Type | Framing | Start Shows | End Shows | Review Focus |
-|------|---------|-------------|-----------|--------------|
-| Wide | Full body + env | Context | Action result | Scene continuity |
-| Medium | Chest up | Character | Emotional shift | Face accuracy |
-| Close-up | Face/product | Detail | Impact | No artifacts |
-
-## Camera Movements
-
-| Movement | Start Frame | End Frame | Veo Directive | Review Focus |
-|----------|-------------|-----------|---------------|--------------|
-| Static | Scene A | Scene A (slight change) | static shot | Minimal drift |
-| Dolly In | Wide view | Close view | camera moves forward | Smooth zoom |
-| Dolly Out | Close view | Wide view | camera pulls back | No jitter |
-| Pan L/R | Left composition | Right composition | camera pans | Fluid motion |
-
-## Timing Template
-
-- Scene 1 Hook: 0:00-0:02 (grab attention)
-- Scene 2 Problem: 0:02-0:04 (pain point)
-- Scene 3 Solution: 0:04-0:06 (product intro)
-- Scene 4 CTA: 0:06-0:08 (call to action)
-
-## Frame Validation Requirements
-
-### Before Veo Generation
-
-Each start/end frame pair must pass AI review:
-
-| Check | Pass Criteria | Tool |
-|-------|---------------|------|
-| Quality | Score >= 8/10 | Gemini vision |
-| Prompt match | All required elements present | Gemini vision |
-| Abnormalities | No extra limbs, distorted faces | Gemini vision |
-| Text accuracy | Any text spelled correctly | Gemini vision |
-| Continuity | End frame N matches start frame N+1 | Gemini vision compare |
-| Style | Consistent visual language | Gemini vision |
-
-### Frame Review Prompt
-
-```python
-def review_frame(frame_path: str, prompt: dict) -> dict:
-    """AI review of generated frame"""
-    response = client.models.generate_content(
-        model='gemini-2.5-flash',
-        contents=[
-            f"""Review this generated frame.
-            Expected: {prompt['prompt']}
-            Required elements: {prompt['validation']['required_elements']}
-            Forbidden: {prompt['validation'].get('forbidden', [])}
-
-            Check for:
-            1. Quality (1-10)
-            2. All required elements present?
-            3. Any forbidden elements?
-            4. AI abnormalities (extra limbs, distorted features)?
-            5. Text accuracy if any visible text?
-
-            Return JSON with passed/failed and issues.""",
-            genai.types.Part.from_bytes(data=frame_bytes, mime_type='image/png')
-        ]
-    )
-    return json.loads(response.text)
-```
-
-### Video Clip Review Prompt
-
-```python
-def review_video_clip(clip_path: str, scene: dict) -> dict:
-    """AI review of generated video clip"""
-    myfile = client.files.upload(file=clip_path)
-
-    response = client.models.generate_content(
-        model='gemini-2.5-flash',
-        contents=[
-            f"""Review this AI-generated video clip.
-            Expected motion: {scene['motion']}
-            Start frame: {scene['start_frame']['prompt']}
-            End frame: {scene['end_frame']['prompt']}
-
-            Check for:
-            1. Motion smoothness (1-10)
-            2. Start/end frames match descriptions?
-            3. Morphing artifacts?
-            4. Temporal consistency (objects don't disappear)?
-            5. Physics errors?
-
-            Return JSON with passed/failed and timestamps of issues.""",
-            myfile
-        ]
-    )
-    return json.loads(response.text)
-```
-
-## Continuity Rules
-
-1. **Visual continuity**: End frame N = Start frame N+1 (same composition)
-2. **Style continuity**: Same tags across all prompts
-3. **Color continuity**: Consistent palette throughout
-4. **Lighting continuity**: Same lighting setup
-5. **Character continuity**: Same clothing, position, expression flow
-
-## Continuity Review
-
-```python
-def review_continuity(end_frame: str, next_start_frame: str) -> dict:
-    """Check frame pair for continuity"""
-    response = client.models.generate_content(
-        model='gemini-2.5-flash',
-        contents=[
-            """Compare these two frames for video continuity.
-            Frame 1 is the END of Scene N.
-            Frame 2 is the START of Scene N+1.
-
-            Check:
-            1. Same scene/location?
-            2. Objects in same position?
-            3. Lighting consistent?
-            4. Colors match?
-            5. Would a cut between these be jarring?
-
-            Return continuity_score (1-10) and issues.""",
-            genai.types.Part.from_bytes(data=end_bytes, mime_type='image/png'),
-            genai.types.Part.from_bytes(data=start_bytes, mime_type='image/png')
-        ]
-    )
-    return json.loads(response.text)
-```
-
-## Best Practices
-
-1. **Validate before proceeding**: Never skip review gates
-2. **Specificity wins**: Include expressions, colors, lighting in prompts
-3. **Style tags**: Apply consistently across ALL frame prompts
-4. **Simple motion**: One movement per 2s scene for best results
-5. **Audio sync**: Plan SFX timestamps during storyboard phase
-6. **Regeneration budget**: Allow 2 retries per frame/clip
-
-## Common Issues & Fixes
-
-| Issue | Detection | Fix |
-|-------|-----------|-----|
-| Extra limbs | Vision review | Regenerate with "anatomically correct" |
-| Face distortion | Vision review | Add "natural face, symmetrical" |
-| Object morphing | Video review | Simplify motion, shorter duration |
-| Style drift | Compare frames | Enforce identical style tags |
-| Text errors | OCR review | Regenerate or post-process |
-
-## Output Structure
-
-Create directory: `assets/storyboards/{date}-{slug}/` containing:
-- `storyboard.md` - Full storyboard document
-- `storyboard.json` - Machine-readable scene data
-- `scene-{N}-start.png` - Start frame images
-- `scene-{N}-end.png` - End frame images
-
-## `storyboard.md` Format
-
-```markdown
-# Storyboard: {Title}
-
-**Generated:** {date} | **Aspect:** {ratio} | **Duration:** {total}s
-
-## Resources
-
-- Quality Review Workflow: `./quality-review-workflow.md`
-- Script Templates: `./script-templates.md`
-- Veo Prompt Guide: `./veo-prompt-guide.md`
-- Vision Understanding: `../../ai-multimodal/references/vision-understanding.md`
-- Video Analysis: `../../ai-multimodal/references/video-analysis.md`

package/.opencode/skills/video/references/thumbnail-design-guide.md

@@ -1,367 +0,0 @@
-# Thumbnail Design Guide
-
-Create high-converting video thumbnails for maximum click-through rates.
-
-## Thumbnail Fundamentals
-
-### Psychology
-
-1. **Pattern Interrupt:** Stand out from surrounding content
-2. **Curiosity Gap:** Hint at value without revealing everything
-3. **Emotional Trigger:** Evoke response (surprise, curiosity, desire)
-4. **Trust Signal:** Show faces, quality, authenticity
-
-### Platform Specifications
-
-| Platform | Size | Ratio | Safe Zone | Format |
-|----------|------|-------|-----------|--------|
-| YouTube | 1280×720 | 16:9 | Center 60% | JPG, PNG |
-| TikTok | 1080×1920 | 9:16 | Top 30% | JPG, PNG |
-| Instagram | 1080×1080 | 1:1 | Center | JPG, PNG |
-| LinkedIn | 1200×627 | 1.91:1 | Full | JPG, PNG |
-
-## Design Principles
-
-### The 3-Second Test
-
-Thumbnail must communicate value in < 3 seconds at small size.
-
-**Test:** View at 150×84px (YouTube mobile size)
-- Can you identify subject?
-- Is text readable?
-- Is emotion clear?
-
-### Visual Hierarchy
-
-```
-1. FACE/EXPRESSION (if applicable)
-    ↓
-2. KEY OBJECT/PRODUCT
-    ↓
-3. TEXT (3-4 words max)
-    ↓
-4. BRAND ELEMENT
-    ↓
-5. BACKGROUND
-```
-
-### Color Strategy
-
-**High Contrast:**
-- Use complementary colors
-- Avoid matching platform colors (avoid red on YouTube)
-- Bright colors attract attention
-
-**Recommended Palettes:**
-- Yellow + Black (high visibility)
-- Blue + Orange (complementary)
-- White + Bold color (clean)
-- Gradient backgrounds (depth)
-
-## Thumbnail Types
-
-### Face-Forward
-
-Best for: Vlogs, tutorials, testimonials
-
-```
-Layout:
-┌─────────────────────────┐
-│  TEXT       [FACE]      │
-│  HERE    (expression)   │
-│           [Product]     │
-└─────────────────────────┘
-```
-
-**Requirements:**
-- Clear facial expression
-- Eyes visible and engaging
-- Emotion matches content
-- Face in right 2/3
-
-### Product Focus
-
-Best for: Reviews, demos, unboxings
-
-```
-Layout:
-┌─────────────────────────┐
-│     BIG TEXT            │
-│   [PRODUCT IMAGE]       │
-│    hero placement       │
-└─────────────────────────┘
-```
-
-**Requirements:**
-- Product clearly visible
-- High-quality image
-- Contrast with background
-- Supporting text
-
-### Before/After
-
-Best for: Transformations, tutorials, results
-
-```
-Layout:
-┌────────────┬────────────┐
-│   BEFORE   │   AFTER    │
-│   (dull)   │  (bright)  │
-│            │            │
-└────────────┴────────────┘
-```
-
-**Requirements:**
-- Clear visual difference
-- Arrow or divider
-- Dramatic contrast
-- Transformation obvious
-
-### List/Number
-
-Best for: Tips, rankings, compilations
-
-```
-Layout:
-┌─────────────────────────┐
-│    5 WAYS TO...         │
-│        [icons]          │
-│    [visual elements]    │
-└─────────────────────────┘
-```
-
-**Requirements:**
-- Large, bold number
-- Clear topic indicator
-- Visual representations
-- Numbered badge
-
-### Curiosity Gap
-
-Best for: Revealing content, mysteries, surprises
-
-```
-Layout:
-┌─────────────────────────┐
-│    THIS IS WHY...       │
-│    [blurred/hidden]     │
-│         ?               │
-└─────────────────────────┘
-```
-
-**Requirements:**
-- Something hidden/obscured
-- Question element
-- Intrigue without clickbait
-- Promise of reveal
-
-## Text Best Practices
-
-### Typography Rules
-
-| Element | Size | Style |
-|---------|------|-------|
-| Main text | 72-96pt | Bold, sans-serif |
-| Secondary | 48-60pt | Semi-bold |
-| Brand/logo | 24-36pt | Consistent |
-
-### Word Count
-
-- **Ideal:** 3-4 words
-- **Maximum:** 6 words
-- **No text:** Also valid (visual-only)
-
-### Effective Text
-
-**Do:**
-- Use power words (FREE, NEW, SECRET, HACK)
-- Create curiosity ("I tried..." "What happens when...")
-- Include numbers ("5 Ways", "$1000")
-- Use ALL CAPS sparingly
-
-**Don't:**
-- Full sentences
-- Small text
-- Script/decorative fonts
-- Text over busy areas
-
-### Text Placement
-
-```
-Safe zones:
-┌─────────────────────────┐
-│ [TEXT]              [X] │  ← Avoid corners (UI elements)
-│                         │
-│    BEST PLACEMENT       │  ← Center-left works well
-│                         │
-│                [TIME]   │  ← Avoid bottom-right
-└─────────────────────────┘
-```
-
-## Element Guidelines
-
-### Faces
-
-**Expression Types:**
-- Surprised (mouth open, wide eyes)
-- Excited (big smile)
-- Thoughtful (raised eyebrow)
-- Curious (slight head tilt)
-
-**Technical:**
-- High resolution
-- Good lighting on face
-- Eyes visible
-- Clean background behind face
-
-### Products
-
-**Photography:**
-- Hero angle
-- Studio lighting or bright natural
-- Shadow for depth
-- Multiple angles for context
-
-**Composition:**
-- Product fills 40-60% of frame
-- Clear space around product
-- Consistent style with brand
-
-### Graphics/Icons
-
-**Usage:**
-- Arrows (pointing to key element)
-- Circles/highlights (drawing attention)
-- Badges (NEW, FREE, etc.)
-- Emoji (sparingly, if brand-appropriate)
-
-## Platform-Specific Tips
-
-### YouTube
-
-- Time stamp visible in corner (avoid)
-- Mobile-first design
-- A/B test different versions
-- Update underperforming thumbnails
-
-### TikTok
-
-- First frame often becomes thumbnail
-- Can upload custom cover
-- Vertical format priority
-- Bold, simple designs
-
-### Instagram Reels
-
-- Taken from video or custom
-- Matches feed aesthetic
-- Profile image overlay consideration
-- Caption may appear
-
-### LinkedIn
-
-- Professional aesthetic
-- Less flashy, more credible
-- Face photos work well
-- Corporate color schemes
-
-## Imagen 4 Thumbnail Prompts
-
-### Face-Forward Thumbnail
-
-```
-Professional YouTube thumbnail:
-Person: {description}, {expression}
-Composition: Face prominent right side
-Background: Clean gradient, {color}
-Text overlay space: Left third
-Style: High contrast, professional
-Lighting: Studio quality, flattering
-```
-
-### Product Thumbnail
-
-```
-Product marketing thumbnail:
-Product: {product_name}, hero angle
-Background: Clean {color}, subtle gradient
-Lighting: Professional studio
-Composition: Product centered, text space above
-Style: Premium, high-end
-Quality: Sharp, high resolution
-```
-
-### Before/After Thumbnail
-
-```
-Split comparison thumbnail:
-Left side: "Before" {condition}
-Right side: "After" {result}
-Divider: Arrow or slash in center
-Colors: Muted left, vibrant right
-Text: BEFORE/AFTER labels
-Style: Clear transformation visual
-```
-
-## Checklist
-
-### Pre-Design
-
-- [ ] Understand video content
-- [ ] Identify key emotion/message
-- [ ] Choose thumbnail type
-- [ ] Select color scheme
-- [ ] Plan text (3-4 words)
-
-### Design
-
-- [ ] Create at correct size
-- [ ] Apply visual hierarchy
-- [ ] Add face/product prominently
-- [ ] Include minimal, bold text
-- [ ] Ensure high contrast
-
-### Quality Check
-
-- [ ] Test at mobile size (150×84px)
-- [ ] Verify text readability
-- [ ] Check face/emotion clarity
-- [ ] Confirm no corner conflicts
-- [ ] Match brand guidelines
-
-### A/B Testing
-
-- [ ] Create 2-3 variations
-- [ ] Test different expressions
-- [ ] Try different text
-- [ ] Compare color schemes
-- [ ] Track CTR by version
-
-## Tools Integration
-
-### Imagen 4 Generation
-
-Use `ai-multimodal` skill:
-```bash
-python scripts/gemini_batch_process.py \
-  --task generate \
-  --prompt "YouTube thumbnail: {description}" \
-  --aspect-ratio 16:9 \
-  --output thumbnail.png
-```
-
-### FFmpeg Thumbnail Extraction
-
-```bash
-# Extract frame from video
-ffmpeg -i video.mp4 -ss 00:00:02 -vframes 1 thumbnail.png
-
-# Resize to YouTube specs
-ffmpeg -i thumbnail.png -vf "scale=1280:720" youtube_thumb.png
-```
-
-## Resources
-
-- [Video Optimization](./video-optimization.md)
-- [Platform Specs](./video-types-specs.md)
-- [Image Generation](../../ai-multimodal/references/image-generation.md)