Nano Banana Pro

• You want to generate a new image from a text description, with control over style, composition, and resolution.
• You need to edit an existing image — change colors, remove elements, adjust style, or apply effects — by describing the change in plain language.
• You want a fast draft → iterate → final workflow: generate a cheap 1K draft, refine the prompt, then produce a 4K final only when the prompt is locked.
• You have a GEMINI_API_KEY and are working in a CLI/agent environment where you can pass --input-image paths.

• You need a non-Google image model (e.g. DALL·E, Stable Diffusion, Flux) — this skill exclusively uses the Gemini image API.
• You do not have a valid GEMINI_API_KEY with image generation access enabled on your account.

Describe any subject, style, composition, lighting, and color palette in a prompt and the skill will produce a PNG image saved to your current working directory. Prompt templates are included for high hit-rate results even from vague input.

Pass an existing image via --input-image and describe only what should change. A built-in editing prompt template helps preserve everything else — subject, pose, lighting, background, and existing text — while applying a single precise modification.

Choose 1K (~1024px) for rapid iteration, 2K (~2048px) for medium-quality output, or 4K (~4096px) for high-resolution finals. The skill maps common user phrases like "hi-res" or "ultra" to the correct API parameter automatically.

The recommended workflow starts with a 1K draft for fast feedback, allows prompt refinement over multiple iterations (keeping the same --input-image for edits), and only upgrades to 4K once the prompt is confirmed — saving time and API quota.

Output files are automatically named with a yyyy-mm-dd-hh-mm-ss-descriptive-name.png pattern derived from the prompt context, keeping your working directory organized across multiple generation runs.

Quickly generate visual concepts — characters, environments, product mockups — by describing them in natural language. Iterate on style and composition at 1K before exporting a 4K final for presentation.

Take an existing photo or design asset and apply changes like "remove the person in the background", "change to a watercolor painting style", or "make the sky more dramatic" without opening a graphics editor.

Integrate into a CLI-driven AI agent pipeline to generate images on demand as part of a larger task — for example, producing illustration assets for a document or website being built programmatically.

Supply a source image and instruct the skill to reinterpret it in a different visual style — cartoon, oil painting, cinematic — while keeping the original composition and subject intact.