Banner Creator - AI Image Generation Banner Design Tool

Skill Overview

Banner Creator is a banner design skill powered by AI image generation (Gemini 3 Pro Image / nanobanana). Through an iterative design workflow of “requirements confirmation → batch variant generation → preview feedback → iterative refinement → cropping and export,” it helps you quickly create GitHub README banners, Twitter header images, website hero images, and banners for various platforms.

Use Cases

1. Open-source project README banners: Generate 2:1 (1280x640) GitHub README banners for new projects, incorporating the project name, tagline, and mascot logo to give the repository homepage a more professional appearance.
2. Social media header and cover images: Create Twitter/X header images (3:1, 1500x500), LinkedIn banners, YouTube channel banners, and other visual assets for social platforms, with one-click cropping to meet each platform’s specifications.
3. Website hero images and content covers: Generate 16:9 (1920x1080) hero images for landing pages, or style-consistent cover images for blogs and documentation.

Core Features

1. Batch generation of design variants: Generate 20 banner variants at once (using the widest 21:9 aspect ratio), automatically numbered and saved (banner-01.png, banner-02.png, …), with an accompanying HTML preview page for easy side-by-side comparison and selection—avoiding repeated guesswork with a single concept.
2. Iterative feedback-based refinement: Based on your preferences indicated through the preview page and feedback—including style, color palette, and composition—generate another 10–20 variants from the concepts you like (such as banner-03-v1.png). Continue iterating through multiple rounds until you are satisfied. You can also provide an existing logo image and have the AI naturally incorporate the brand mascot into the banner composition.
3. Precise cropping for platform-specific aspect ratios: Once the final concept is selected, automatically crop it to the required dimensions for the target platform—GitHub README (2:1, 1280x640), Twitter header (3:1, 1500x500), or website hero (16:9, 1920x1080). The final outputs and all working files are archived together in the .skill-archive/banner-creator/ directory.

Frequently Asked Questions

What prerequisites are required to use Banner Creator?

Two things are required: first, set GEMINI_API_KEY in your environment variables (available from Google AI Studio); second, install the nanobanana skill, which handles the actual AI image generation using the Gemini 3 Pro Image model. Once both are ready, you can begin creating banners.

What if I’m not satisfied with the generated banner?

That is exactly what Banner Creator is designed for—it is not a one-shot image generator, but an iterative workflow. First, it generates 20 variants for you to choose from. Then tell it which ones you like (for example, #3 and #7), why you like them, and what you would like to adjust. It will generate another batch of variants based on your feedback, repeating the process until you select the final concept.

What sizes are supported, and how can I ensure they meet platform requirements?

The generation stage consistently uses the widest 21:9 aspect ratio to provide sufficient room for cropping. Once the design is finalized, a cropping script precisely exports the image at the target aspect ratio and width: GitHub README uses 2:1 (1280x640), Twitter headers use 3:1 (1500x500), and website heroes use 16:9 (1920x1080). All intermediate files and final outputs are stored in .skill-archive/banner-creator/<date-name>/, including the uncropped originals and final cropped versions.

Quick Start

1. Confirm that GEMINI_API_KEY is set and the nanobanana skill is available.
2. Tell it the banner’s purpose, target aspect ratio, preferred style, text to include, and logo requirements.
3. Generate 20 variants in batches and select from them using the HTML preview page.
4. Iterate based on feedback until you find a satisfactory concept.
5. Automatically crop the selected design to the target dimensions and deliver the final file.