YouTube Thumbnails — AI-generated YouTube thumbnails in proven styles

Tools / YouTube Thumbnails

Generate Thumbnail

Generate YouTube thumbnails from a video topic. Supports 10 proven styles, text overlays, reference images for brand/face consistency, and batch generation for A/B testing.

Returns: Thumbnail image(s) at 1280x720, downloadable via asset system, prompt used, YouTube specs, and best practice tips

Parameters

topic *stringVideo topic or thumbnail concept (e.g. "unboxing the new iPhone", "how I lost 50 pounds in 6 months")

stylestringThumbnail style: reaction, before_after, versus, tutorial, listicle, product, storytime, travel, gaming, food

text_overlaystringBold text to render on the thumbnail (3-5 words max, e.g. "I QUIT MY JOB", "$10K IN 30 DAYS"). Rendered in thick sans-serif with contrasting outline.

moodstringEmotional tone (e.g. "shocking", "inspiring", "mysterious", "exciting", "cozy", "dramatic")

color_schemestringColor guidance (e.g. "red and yellow", "dark with neon accents", "warm golden tones", "blue gradient")

reference_imagesarrayURLs of reference images — face photos for consistency, brand assets, or style references. First image is primary.

additional_detailsstringExtra prompt details for fine-tuning the composition, subject, or scene (e.g. "person wearing red hoodie on left side", "luxury car in background")

num_imagesnumberNumber of thumbnail variations to generate (1-12). Use 4-6 for A/B testing selection. Defaults to 1.

modelstringImage model to use. Defaults to Nano Banana 2 (best text rendering + quality). Use generate-image list_models for alternatives.

seednumberRandom seed for reproducible results.

output_formatstringOutput format. PNG (default) for text-heavy thumbnails, JPEG for photo-heavy. YouTube accepts JPG, PNG, GIF, BMP under 2MB.

negative_promptstringThings to avoid in the thumbnail (appended to built-in exclusions for blurry, watermarks, etc.)

guidance_scalenumberHow closely to follow the prompt (1-20). Higher = more literal. Model-dependent.

enhance_promptbooleanLet the model expand and improve the prompt automatically.

Example

Generate a reaction-style thumbnail for a tech video

curl -H "Authorization: Bearer $TOOLROUTER_API_KEY" \
  -d '{
  "tool": "youtube-thumbnails",
  "skill": "generate_thumbnail",
  "input": {
    "topic": "Unboxing the most expensive laptop ever made",
    "style": "reaction",
    "text_overlay": "$15,000 LAPTOP",
    "num_images": 4
  }
}' \
  https://api.toolrouter.com/v1/tools/call

Quick Start

MCP (Claude Code)

claude mcp add --transport stdio \
  --env TOOLROUTER_API_KEY=YOUR_API_KEY \
  toolrouter -- npx -y toolrouter-mcp

REST API

curl -H "Authorization: Bearer $TOOLROUTER_API_KEY" \
  -d '{"tool":"youtube-thumbnails","skill":"generate_thumbnail","input":{}}' \
  https://api.toolrouter.com/v1/tools/call

Frequently Asked Questions

What thumbnail styles can I choose from?

`generate_thumbnail` supports reaction, before_after, versus, tutorial, listicle, product, storytime, travel, gaming, and food styles.

Can I add bold text and brand or face references?

Yes. Use `text_overlay` for the headline and `reference_images` for face consistency or brand assets.

How many versions can I generate for testing?

Set `num_images` from 1 to 12; 4 to 6 is a good range when you want A/B options.

What size and format will I get?

Thumbnails come out at 1280x720, and you can choose PNG, JPEG, or WebP. The tool also returns a shareable `image_page` link.