---
name: thumbnailcreator
description: Generate YouTube, blog and social thumbnails from a text prompt with multiple AI image models (GPT Image 2, Nano Banana Pro, FLUX.2, Seedream, Ideogram, Grok) using ThumbnailCreator. Use when the user asks for a thumbnail, cover image or social preview image, or wants to compare image models on one prompt.
---

# ThumbnailCreator

## How to call it

1. **MCP (preferred).** If the `thumbnailcreator` MCP server is connected, call `generate_thumbnails` with `prompt`, optional `models`, `n` and `format`. It waits and returns image URLs (valid 1 hour) and the amount charged. If it is not connected, tell the user to add it:
   `claude mcp add --transport http thumbnailcreator https://thumbnailcreator.ai/api/mcp --header "Authorization: Bearer <key>"` with a key from https://thumbnailcreator.ai (account menu > MCP keys).
2. **x402 (no account).** If you can make x402 payments (a wallet-enabled fetch), POST to `https://thumbnailcreator.ai/api/x402/generations?models=gpt-image-2&n=1&format=16:9` with `{"prompt": "..."}` and pay the quoted USDC amount. The response contains `images[]` with URLs.

Download images the user wants to keep; URLs expire.

## Choosing models
- gpt-image-2 (GPT Image 2, OpenAI): $0.167/image, up to 4 reference images. Best prompt following and in-image text.
- nano-banana-pro (Nano Banana Pro, Google): $0.195/image, up to 4 reference images. Photoreal scenes, faces and reference fidelity.
- nano-banana-2 (Nano Banana 2, Google): $0.132/image, up to 4 reference images. Fast Gemini model with strong composition.
- flux-2-max (FLUX.2 Max, Black Forest Labs): $0.130/image, up to 0 reference images. Highest-fidelity photographic detail.
- seedream-5-lite (Seedream 5 Lite, ByteDance): $0.046/image, up to 4 reference images. Reasoning-driven layouts at a low price.
- ideogram-v3 (Ideogram 3, Ideogram): $0.117/image, up to 3 reference images. Bold typography and graphic design.
- grok-imagine (Grok Imagine, xAI): $0.026/image, up to 0 reference images. Cheapest and quickest for exploring ideas.

Defaults: `gpt-image-2` for text-heavy thumbnails, add `nano-banana-pro` for photoreal faces, `grok-imagine` for cheap exploration. To compare, pass 2-3 models with `n: 1`. Max 8 images per request.

## Prompting
- Describe subject, emotion, composition and background. Put exact on-image text in double quotes and keep it to 3-5 words.
- Formats: 16:9, 1:1, 9:16, 4:5. Use 16:9 for YouTube.
- Tell the user what it cost after generating.

## Errors
Free allowance used up: the user must add a card at https://thumbnailcreator.ai. Monthly spend limit reached: ask the user. Too many requests: wait and retry.