youtube-seo-thumbnail

name	youtube-seo-thumbnail
description	Advanced thumbnail CTR analysis using computer vision, face/emotion detection, CLIP-embedding SERP similarity, Gestalt composition rules, and YouTube native Test & Compare planning. Use when user says "thumbnail review", "improve CTR", "thumbnail design", or provides a thumbnail image/URL.
user-invokable	true
argument-hint	[video-url | thumbnail-image-path]
allowed-tools	["Read","Grep","Bash","WebFetch"]

Thumbnail Analysis (Advanced)

Diagnose CTR potential quantitatively when possible, qualitatively when not. Thumbnail is the single biggest pre-click lever after title.

Data Collection

1. Source thumbnail: download the maxres (maxresdefault.jpg) from https://i.ytimg.com/vi/{VIDEO_ID}/maxresdefault.jpg, fall back to hqdefault.jpg if missing
2. SERP grid: fetch the top 12 YouTube results for the primary keyword, download each thumbnail, build a 4x3 contact sheet
3. Computer-vision pass via scripts/analyze_thumbnail.py:
   • OpenCV face detection (count, bbox, size %)
   • MediaPipe face-mesh for emotion inference (or DeepFace fallback)
   • Color histogram (HSV) — dominant 3 colors
   • Contrast score (stddev of luminance)
   • Text region detection (EAST or Tesseract) — estimated word count and coverage %
   • CLIP embedding (ViT-B/32) for similarity scoring vs SERP grid
4. Mobile preview: resize to 120x68, 246x138, 480x270 and inspect legibility manually

Analysis Dimensions

1. Technical

• Resolution 1280x720, 16:9, max 2MB
• Format JPG / PNG
• Safe area: no critical content in bottom-right (duration overlay ~6% of area) or top-right (menu dots)

2. Legibility at small sizes (binding constraint)

At 120x68:

• Subject identifiable? (pass/fail)
• Text readable? (pass/fail)
• Emotion readable? (pass/fail)

If any fail → critical issue. Most thumbnails designed on a desktop monitor at 500px+ fail at 120px.

3. Composition — Gestalt principles

• Figure-ground separation: subject clearly separated from background (rim light, cutout, color contrast, background blur)
• Closure: viewer's brain completes an implied shape or story — works against scroll blindness
• Continuity: eye travels in a designed path (eye direction of face, arrow of object, gradient)
• Similarity/proximity: elements that belong together are grouped
• Rule of thirds: focal point at intersections, not dead-center
• Depth: foreground / midground / background create dimensionality
• Negative space: enough breathing room for the focal point to pop at small sizes

4. Face + emotion

• Face present? (niche-dependent: boost for most, neutral for gaming/ product review close-ups)
• Facial emotion (from MediaPipe/DeepFace): surprise, fear, joy, anger, disgust, sadness, neutral. Surprise and fear typically lift Browse CTR most.
• Eye contact direction: direct-gaze lifts trust; side-gaze creates curiosity
• Face size: 20-35% of thumbnail area is the sweet spot
• Single face outperforms multiple faces in most A/B tests

5. Text overlay

• ≤4 words (ideally ≤3)
• Font: bold sans-serif, 80-120pt effective
• Legibility: stroke or drop-shadow against variable backgrounds
• Complementary to title (not redundant): the thumbnail text should say what the title does NOT
• Placement: top-left or top-right (viewers scan F-pattern on feeds)

6. Color palette

• 2-3 dominant colors (from histogram)
• High-saturation accent against low-saturation base
• Contrast vs YouTube UI (white on light mode, dark gray on dark mode): ensure thumbnail pops in both modes
• Brand consistency across the channel grid (same accent color, same typography family)

7. SERP differentiation (CLIP embedding)

Compute cosine similarity between the target thumbnail and each of the top-12 SERP thumbnails. Scoring:

• Avg similarity <0.55: strong pattern-break (good)
• 0.55-0.70: typical
• >0.70: too similar to competitors, will blend in

Also report the 3 most similar SERP thumbnails as visual references for the user.

8. Channel brand consistency

Compare the target thumbnail to the last 11 channel thumbnails (4x3 grid). Measure:

• Color palette overlap
• Typography consistency
• Face presence consistency
• Style template adherence

Flag if the target is an outlier — brand recognition in the subscriber feed drops when styles drift.

9. Title-thumbnail synergy

• Zero redundant words between title and thumbnail text
• Thumbnail emotion matches title promise
• No bait-mismatch with video content (long-term CTR decay risk)

Output

Thumbnail Score Card

Overall: XX/100

Technical:         XX/100
Legibility (120px):XX/100   ← hard gate
Composition:       XX/100
Face + Emotion:    XX/100
Text Overlay:      XX/100
Color / Contrast:  XX/100
SERP Differentiation: XX/100  (cosine avg: 0.XX)
Brand Consistency: XX/100
Title Synergy:     XX/100

CV Report

Faces detected: N
Dominant emotion: surprise (0.XX confidence)
Face area: XX% of thumbnail
Dominant colors: #RRGGBB, #RRGGBB, #RRGGBB
Contrast score: X.XX
Text regions: N  (est. words: M)
SERP similarity (avg cosine): 0.XX
Most similar SERP entry: {URL}

Issues Found

Critical → High → Medium → Low.

Redesign Brief (paste-ready for designer or image generator)

Subject: {who/what is centered}
Expression: {facial emotion — cite research-backed CTR driver}
Composition: {rule of thirds position, focal point, leading line}
Background: {setting, depth layers, color wash, blur level}
Foreground element: {optional prop or graphic element}
Text overlay: "{≤3 words}" in {font family, weight}, {color} with
              {stroke/shadow}, {position}, {size %}
Palette: {2-3 hex colors + accent}
Style: {photographic / illustrated / hybrid / 3D}
Differentiator vs SERP: {specific pattern break — "all competitors use
                         red + face left, we use teal + face right"}
References: {2-3 high-performing thumbnails from outside the niche}
Safe area: {avoid bottom-right duration overlay}
Export: 1280x720 JPG, <2MB, sRGB

Native A/B Test Plan

YouTube Studio → Content → Edit → Thumbnail → Test & Compare:

• Variant A: current thumbnail (control)
• Variant B: redesigned with strongest recommended change (e.g., swap emotion surprise→fear)
• Variant C: second-strongest change (e.g., different color palette)
• Rotation: YouTube decides split based on impressions
• Window: 2 weeks minimum, 10k impressions minimum for significance
• Success metric: Browse CTR (not Search CTR — Search is keyword- bound). Require ≥0.5 absolute point lift for decision.

Generation (optional)

If the user wants an actual image generated, delegate to seo-image-gen with the redesign brief as the prompt. Set aspect 16:9 and size 1280x720.

Error Handling

Scenario	Action
No image provided	Ask for video URL or thumbnail file
Image <640px wide	Flag and ask for the source file
Channel niche unknown	Ask before running SERP differentiation
CLIP unavailable in environment	Fall back to manual grid inspection and description-based similarity
MFK channel	Emphasize legibility + brand; skip face-emotion lift (kids thumbnails have different rules)