Thumbnail Generator creates high-CTR video thumbnails by analyzing video content, extracting the most visually compelling frame, adding optimized text overlays, and applying design principles proven to drive clicks. It uses Claude Code to understand content context and generate designs that match your channel's brand.
Thumbnails are the single biggest factor in YouTube click-through rates. Yet most creators spend either too little time (screenshot with text) or too much time (Photoshop from scratch) on them. Thumbnail Generator finds the sweet spot — AI-designed thumbnails that follow proven CTR principles.
The skill analyzes your video to find the most engaging frame (facial expressions, high contrast, dynamic composition), adds text with optimal placement and size, applies your brand colors and style, and generates multiple variants for A/B testing.
It understands YouTube thumbnail psychology: large faces with emotion perform best, contrasting colors stand out in feeds, and minimal text (3-5 words max) drives the highest CTR.
# Generate thumbnail from video
thumb-gen --video my-video.mp4 --text "This Changes Everything"
# With brand guidelines
thumb-gen --video my-video.mp4 --text "AI Agents 2026" --brand brand-kit.json
# Batch mode
thumb-gen --dir ./videos/ --style energetic --output ./thumbnails/
Analyzing: how-to-build-agents.mp4
Frame Analysis:
Best frame: 04:23 (speaker with surprised expression, high contrast)
Face detected: Yes (center-left composition)
Emotion: Excited/Surprised (score: 0.87)
Generated Variants:
1. Bold white text, dark gradient overlay — CTR prediction: High
2. Yellow text, minimal overlay — CTR prediction: Medium-High
3. Split design with comparison — CTR prediction: Medium
Saved: thumbnail-v1.png, thumbnail-v2.png, thumbnail-v3.png (1280x720)
AI agents that work well with Thumbnail Generator.
Full podcast transcription with speaker diarization, timestamps, and exportable show notes.
Transcribe and summarize YouTube videos with timestamps, key points, and chapter-based navigation.
AI-assisted video editing with scene detection, auto-cuts, transitions, and caption generation.