zenmux-image-gen_skill

Overview

This skill generates images from text prompts using the ZenMux API and Google Gemini image models. It supports text-to-image and image-to-image editing with control over aspect ratio and resolution. The script enforces input validation, timeout protection, and safe handling of API keys (requires ZENMUX_API_KEY).

How this skill works

You call the provided script with a text prompt and output filename; the script sends the prompt to ZenMux using the Vertex AI protocol and a selected Gemini model. Options let you specify model, temperature, aspect ratio, and image size (1K/2K/4K). For image-to-image edits, pass an input image path and a descriptive edit prompt. The script auto-creates output directories, validates inputs, and retries with clear error messages when applicable.

When to use it

• Create artwork, illustrations, or marketing visuals from descriptive prompts
• Perform image-to-image edits (e.g., change time of day or style of an existing photo)
• Programmatically generate batches or variations of images for testing or A/B
• Need aspect ratio or resolution control (1:1 up to 21:9, 1K/2K/4K)
• Experiment with different Gemini models for quality vs. speed trade-offs

Best practices

• Set ZENMUX_API_KEY in the environment; never pass it on the command line or commit it
• Use prompts of 10+ words with concrete details: subject, lighting, style, and mood
• Start with default fast model, then switch to higher-quality models for final renders
• Compress or resize input images to stay under the 20 MB upload limit
• Run multiple iterations to refine prompts and generate variations

Example use cases

• Generate a series of product mockups with consistent aspect ratios for an online store
• Transform daytime photos into nighttime scenes via image-to-image editing
• Create concept art or storyboard frames quickly from text briefs
• Produce social media assets at 16:9 or square formats for multichannel posting
• Batch-generate character poses or scene variations for game prototyping