Google DeepMind
Build with Gemini
Try Gemini
Nano Banana

How to create effective image prompts with Gemini Image

Try in Gemini
Try in Google AI Studio
A cinematic, wide-angle shot of a vintage orange sedan parked on a dark, asphalt road at night. The scene is enveloped in a thick, teal-tinted fog that creates a moody, ethereal atmosphere. The car's headlights are turned on, casting powerful beams of warm light through the mist. To the left, the silhouettes of towering palm trees and a suburban house loom in the background, while a solitary streetlamp glows on the right, further illuminating the hazy green air.

What to think about while writing prompts

Use detailed prompts to take more control over the images you generate. Think about what you want to see – the characters, the setting, and the overall feel. The more detail you add, the closer the image will be to what you’ve imagined.

Editing with prompts

After creating an image, you can keep editing it until you’re happy. Tweak specific details, reframe the image, upscale the quality – or overhaul the entire aesthetic. It’s up to you.

General prompting

Even short prompts will generate high-quality images, but you can take better control of the output by adding details bit-by-bit. Keep playing, experimenting, and iterating, until you get the image you want to use.

Visualize real-world knowledge

Describe how you’d like the model to organize information by asking for a timeline or a flowchart. Prompt using real-world concepts, locations, or historical eras to draw from the model’s deep knowledge base. Mention the visual theme you’d like the design to follow.

A craft-style diagram of the water cycle featuring evaporation, condensation, precipitation, and runoff.
A greeting card with two hands forming a heart shape, resting on a brown knit sweater.

Render precise text

Enclose your desired words in quotes (e.g., "Happy Birthday"). And describe the typography style, like a "bold sans-serif font" or "neon cursive signage".

Translate and localize ideas

Write in one language, then specify the target language and regional cultural cues in your prompt. If you need specific foreign phrasing, provide the exact translated text in quote marks to guide the rendering.

A wooden sign in a jungle illustrating local Indian wildlife with text in Hindi.
A grid of photos in several aspect ratios showing a person in a vibrant blue and pink patterned suit and yellow sunglasses.

Prompt for production-ready specs

Prompt for your desired aspect ratio (like 4:3, 1:1, or 9:16) and describe the image format you need – for example, a "widescreen backdrop," or a "vertical social post". Look for the in-product resolution dropdowns to upscale to 2K or 4K.

Ask for multiple images at once

Ask for "three distinct variations of a product mock up" or "four different color palettes" to quickly compare ideas side-by-side. Or upload a single reference image and ask the model to generate a multi-image story based on it.

A prompt interface showing three fluffy monster characters and instructions to create a story.A prompt interface showing three fluffy monster characters and instructions to create a story.

Maintain subject consistency

Upload clear reference images, and assign a distinct name to each character or object in your prompt. That way, the model can follow along and maintain their look as you build out your scenes.

A profile portrait of a woman with short hair, illuminated by dramatic teal and magenta rim lighting against a dark, textured background. Her eyes are closed, and the high-contrast lighting emphasizes the contours of her face and the silhouette of her head.