Prompting for images

Gemini can create and modify images for you. You can generate images from scratch, edit existing images, or combine multiple images.

Creating images

To ask Gemini for an image, either:

• Start your prompt with a request such as “Generate an image of…”, “Create a picture of…”, or “Show me…”
• Select +, then Create image

Then describe the image you want and submit your prompt.

What to include in your prompt

Gemini will generate the image it thinks best matches your description. It will make assumptions about any aspects you have not specified.

Consider these aspects of your image:

• Subject: what should be in the image? What should be in the background?
• Arrangement: does the position or layout of elements matter?
• Style: such as cartoon, photorealistic, watercolour, or black-and-white inked noir
• Perspective: such as side-on, wide-angle, macro, or bird’s eye view
• Mood and lighting: time of day, atmosphere, and colour palette, such as warm golden tones, cool blues, or dramatic shadows

Example: adding detail

For example, start with a short prompt for an image of a meadow:

Generate a picture of a meadow.

The result is a reasonable start, but it may not be the meadow you had in mind. With a request as short as “a meadow”, Gemini had to make a lot of assumptions.

Add some detail to fine-tune it:

Generate a watercolour picture looking out over a summer meadow, with cornflowers, cosmos and ox-eye daisies flowering. There is one ancient English oak growing in the meadow, situated to the left of the image. The tree’s trunk and lower branches are visible. On the horizon is the outline of hedgerows. It’s golden hour in the evening, and the sky is hazy. The sun is setting to the right of the scene, with deep reds, pinks, and purples. Against the sky is the silhouette of three swifts flying in the distance overhead.

Choosing the size

By default, Gemini will usually give you a square image.

If you would like your image to have different proportions (its aspect ratio), ask for them. You do not have complete flexibility. Gemini will respect landscape (good for slides and displays) and portrait (good for posters and flyers), as well as common ratios like 16:9 or 4:3.

Getting the best results

To get the best results:

• Be specific: the more detail you give, the more control you have. Instead of asking for “a tree”, ask for “an ancient English oak, with a gnarled trunk, and golden autumnal leaves”.
• Describe positively: instead of saying “a desk with no clutter”, describe what you want to see: “a tidy desk with a notebook, a pen, and a cup of coffee”.
• Use step-by-step instructions: for complex scenes with many elements, try breaking your prompt into steps. For example:

  First, create a background of a snowy village street at dusk, with warm golden light glowing from the windows. Then, add a large decorated Christmas tree in the foreground on the left, dusted with snow. Finally, place a small group of people in winter coats in the middle distance, gathered around a street lantern.

If your image is not quite right the first time, ask for refinements. For example:

Darken the sky. Keep everything else the same.

Editing and combining images

You can also give Gemini images as part of your prompt. This lets you make adjustments to an image or add a specific object or design. For edits, also consider what you want to keep—Gemini will make assumptions about aspects you do not specify.

To give Gemini an image, either:

• Drag an image onto the prompt area
• Select + to add files from your Google Drive or computer

Make adjustments

Give Gemini an image¹ and ask for changes. Think through what to keep, what to change, and how. Include those details in your prompt.

Change the background wall to be a light sage colour. Leave everything else the same.

Add or remove elements

Give Gemini an image and describe the elements you want to add or remove.

Add a tiny paper crown to the cat in this photo. The crown sits naturally on the cat’s head, positioned between its ears, and is made of dull, metallic red paper. It should be photorealistic, with appropriate lighting and lens blur to match its position in the scene.

Update style

Give Gemini an image and transform it into a different style.

Transform this photo into the watercolour style of Quentin Blake.

Combine images

You can give Gemini multiple images and combine elements from each.

Input 12	Input 2

Take the image of a man wearing a t-shirt and apply the graphic across the front, with the white swirl logo centred. The graphic should look like it’s screen-printed on the t-shirt fabric. Adjust the graphic to match the scene’s lighting and the fabric’s texture. Ensure the rest of the image remains unchanged.

Other tools

Other AI tools available at Outwood that can generate and edit images:

• Canva
• Adobe Firefly

Related resources

• Image prompting tips from Google