Skip to main content
Category: Visual & Media · Version: 0.0.1 · Author: opendirectory · Compatibility: All targets (claude, codex, opencode, gemini, anti-gravity, openclaw, hermes)
A modern, AI-powered CLI tool designed to automatically generate high-converting, minimalist blog cover images and thumbnails using Gemini 3.1 Flash Image Preview. blog-cover-image-cli-cover It handles everything from fetching company logos to pixel-perfect typography integration, all from your terminal or directly via an AI Agent using the included OpenCode Skill.
BannerFormatPlatform

​
Install

​
Features

  • Full AI Generation: Uses gemini-3.1-flash-image-preview to generate the entire image.
  • Smart Logo Fetching: Pass a domain (like vercel.com) and the CLI automatically fetches the logo using Brandfetch, normalizes it to PNG via sharp, and injects it into the AI context.
  • Aesthetic Control: Bundled with examples/ that automatically guide the model to produce clean, white-background, heavy-typography styles.
  • Google Search Grounding: The image generation is hooked into Google Search to pull real-time data if your title involves current events.
  • Agent Ready: Includes an OpenCode SKILL.md so your favorite AI agents can use this CLI autonomously.
  • Self-Healing AI Generator: Automatically validates generated images using Gemini Pro Vision to detect typos or layout issues, retrying up to 3 times with corrective feedback.

​
Setup

You can install this globally via npm:
(Note: Ensure you are using Node.js v18+)

​
Configuration

The CLI securely stores your API key on your local machine using the conf package so you don’t have to export it every time.
If you run the generate command without a key, a secure, interactive prompt will ask you for it.

​
Usage

Generate a 16:9 cover image by providing a title and a domain name for the logo.

​
Options

If you omit the --output flag, the CLI automatically creates an output/ directory in your current path and names the file intelligently based on the logo domain or title (e.g., output/cursor-cover.png).

​
For AI Agents (OpenCode Skill)

This skill folder includes a structured SKILL.md for AI agents. The skill automatically calls the published blog-cover-image-cli npm package to generate cover images for users autonomously. Workflow for Agents:
  1. Execute npx -p blog-cover-image-cli blog-cover-cli config set-key $KEY
  2. Execute npx -p blog-cover-image-cli blog-cover-cli generate -t "Title" -l "domain.com"
  3. Return the generated image to the user.

​
Self-Healing AI Generator

The CLI features a built-in Automated QA (Critic) loop to ensure high-quality results.
  1. Generation: The tool generates an image based on your title and logo.
  2. Validation: It uses gemini-3.1-pro-preview to OCR the generated image and check for typos, layout issues, or missing elements.
  3. Self-Correction: If the validation fails, the CLI automatically retries (up to 3 times), passing the specific “critical feedback” back to the generator to fix the errors.
This ensures that common AI image generation issues, like misspelled words in typography, are caught and corrected before you even see the file.

​
How it works under the hood

  1. Logo Fetcher: Hits Brandfetch, parses WebP/SVGs/AVIFs, and converts to strict PNGs.
  2. Context Assembly: Loads aesthetic examples from the ./examples folder to ground the style.
  3. Multimodal Prompting: Assembles the exact text instructions, the visual examples, and the fetched logo into a single unified payload.
  4. Google GenAI SDK: Sends the payload with tools: [{ googleSearch: {} }] to the Gemini 3.1 Flash Image model.