Youtube Thumbnail Analyzer

YouTube Thumbnail Analyzer

Download and visually analyze thumbnails from top-performing YouTube videos for any topic or channel.

Usage

/youtube-thumbnail-analyzer "air fryer recipes"
/youtube-thumbnail-analyzer @ChannelHandle
/youtube-thumbnail-analyzer --topic "Python tutorial" --max 20
/youtube-thumbnail-analyzer @ChannelHandle --top 15

Instructions

Step 1: Parse Arguments

Input mode (one of):

• Topic/keyword: search YouTube for top videos on this topic
• Channel (@handle, URL, or ID): use this channel's top videos

Also:

• --max N (optional): number of thumbnails to download (default: 20, max: 50)
• --scan-limit N (optional, channel mode): cap uploads scanned (default: whole channel)

Step 2: Get the API Key

Check Claude memory for a YouTube Data API v3 key. If not found, ask:

"I need a YouTube Data API v3 key to fetch video data. You can get one from the Google Cloud Console. Please paste your key."

Step 3: Run the Bundled Script

Run scripts/fetch_thumbnails.py — resolve the path relative to this skill's own directory:

# Topic search
YT_API_KEY=API_KEY python3 <skill-dir>/scripts/fetch_thumbnails.py --topic "TOPIC" --max 20

# Channel
YT_API_KEY=API_KEY python3 <skill-dir>/scripts/fetch_thumbnails.py --channel "@handle" --max 20

Dependencies: pip3 install google-api-python-client requests.

The script ranks videos by views, downloads the highest-resolution thumbnail available (maxres → sd → hq → mq), and writes an index.md alongside the images.

Step 4: Read the Data

reports/data/thumbnail-analysis-<slug>-<YYYY-MM-DD>.json

Thumbnails land in reports/thumbnails/<slug>-<YYYY-MM-DD>/ with an index.md in the same folder. The JSON's videos[].thumbnail_path gives each image's full path.

Step 5: Visually Analyze the Thumbnails

This is the step that makes this skill unique — Claude can see images.

Read each downloaded thumbnail with the Read tool (it supports images). For each:

1. Color scheme: dominant colors, warm vs cool tones, brightness, contrast
2. Text overlay: present? how much? font size, position, color, ALL CAPS?
3. Face/person: face present? expression? close-up or full body? position?
4. Composition: rule of thirds? split screen? before/after? product-focused?
5. Objects/props: what objects are visible? food? tech? books?
6. Background: solid color? gradient? real location? blurred?
7. Branding: logo? consistent style across the channel?
8. Emotion: curiosity? excitement? shock?
9. Clickbait signals: arrows? circles? red X marks? exaggerated expressions?

Read thumbnails in batches of 5. Don't describe every thumbnail individually — focus on identifying PATTERNS across all of them, and on what separates the top-viewed from the rest.

Step 6: Write the Report

Write to the path the script printed: reports/thumbnail-analysis-<slug>-<YYYY-MM-DD>.md

# Thumbnail Analysis Report: [Source]
*Analyzed [date] | [N] thumbnails from top-performing videos*

## Executive Summary
3-4 bullet points: Key visual patterns that drive clicks in this niche.

## Performance Context
| # | Title | Views | Engagement | Thumbnail File |
|---|-------|-------|------------|----------------|
[Top 10 videos with their metrics]

## Visual Pattern Analysis

### Color Schemes
- Dominant color palette across top thumbnails
- Warm vs cool color usage
- Brightness and contrast patterns
- Color combinations that correlate with high views

### Text Overlay Patterns
- % of thumbnails using text
- Average word count on thumbnail
- Font style patterns (bold, ALL CAPS, etc.)
- Text positioning (top, center, bottom)
- Text color against background
- What kind of text works (numbers, questions, statements)

### Face & Expression Patterns
- % of thumbnails featuring a face
- Common expressions (surprise, smile, serious, excited)
- Close-up vs medium shot vs full body
- Face position (center, left, right)
- Eye contact with viewer

### Composition Styles
- Most common layouts
- Use of before/after, split-screen, product focus
- Negative space usage
- Visual hierarchy

### Clickbait Elements
- Arrows, circles, and highlighting
- Red X or checkmarks
- Exaggerated reactions
- "vs" or comparison layouts

### Branding & Consistency
- Channels using consistent thumbnail style
- Logo/watermark usage
- Color brand consistency

## What Separates Top Performers
Analysis of what the highest-view thumbnails do differently from lower-view ones.

## Thumbnail Design Recommendations
5-8 specific, actionable recommendations:
1. **[Recommendation]** -- Backed by [data/observation]
...

## Thumbnail Templates
Based on patterns, suggest 3 thumbnail "formulas" that work in this niche:

### Template 1: [Name]
- Layout description
- Color scheme
- Text approach
- Example from data

### Template 2: [Name]
### Template 3: [Name]

## Quota Usage
| Operation | Calls | Units |
|-----------|-------|-------|
Use the `quota_used.breakdown` block from the JSON.

Step 7: Report Completion

Tell the user the report path, the thumbnails folder, how many were downloaded, the top 3 visual insights, the template recommendations, and the quota consumed. Suggest opening index.md in Obsidian to see all thumbnails at once.

Quota Estimate

Channel mode previously cost ~103 units via search.list(channelId=..., order=viewCount), which also only saw a partial slice of the channel; the uploads-playlist route ranks across every upload for ~22. Thumbnail image downloads use img.youtube.com and cost 0 quota. Daily default allowance is 10,000 units, resetting at midnight Pacific Time.