Style

Style helps define the specific aesthetic of the image you want to generate. There are various ways to apply a style depending on the model used.

Available styles

To access the Style functionality and its options, click or tap the prompt box to reveal the options at the bottom. Then, click the fifth tab to open a popup with the available style choices.

The styles shown will vary depending on the model you have selected.

Styles in v3.0

• Auto: Selecting Auto lets the AI decide which style—General, Realistic, or Design—to apply to the generated image. It analyzes your prompt and chooses the style that seems most appropriate for the best result.

• Random: Selecting Random tells the AI to randomly select a style from its vast image database and apply that visual style to the image that will be generated with your prompt. This is a powerful option when you’re looking for fresh, unexpected visual ideas.

• General: A versatile style that can handle just about anything. Ideal for creating artistic works like abstract paintings, pencil sketches, digitally manipulated images, and more.

• Realistic: Best used when aiming for lifelike, photographic images. This style emphasizes realism and detail.

• Design: Perfect for creating logos, print-on-demand products, promotional materials, flyers, menus, and anything related to graphic design. It’s also suitable for certain artistic styles.

• Reference: Allows you to use up to three existing images as the visual style basis. The AI blends those styles with your prompt to generate a new image.

• Style code: Lets you reuse the visual style of any image generated with the Random option. Just enter its 8-character Style code to apply the same look to a new prompt.

Random style examples

The four images below were generated using version 3.0 with Random style. Magic Prompt was OFF.

Prompt: Painting of a woman holding her baby.

Whenever you use Random to generate a set of images, each image will receive a Style code that corresponds to a particular style used in conjunction with the prompt. This code can then be reused with a new prompt to generate images with a similar visual style. More details below.

What is a Reference style?

Available in v3.0, the Reference style feature allows you to select up to three images that the AI will use as a visual style basis when generating new images based on your prompt.

How to add Reference images:

There are two ways to add images as references:

1. Upload from your computer:

Click the Reference icon and select an image from your device.

2. Use an existing Ideogram image:

• While viewing a generated image, you can choose Reference Style from the detailed view on the right-side panel.

• Alternatively, click the More icon on any image, then select Style from the menu.

There are two ways to add images as references:

1. Upload from your computer: Click on the Reference icon and select an image from your computer.

2. Use an existing Ideogram image: While viewing a generated image, you can choose Reference style from the detailed view on the right-side panel. Alternatively, click the More icon from any image, on any image, then select Style reference from the menu.

Reference style examples

Here are some comparisons of images generated using the same prompt, with and without a Reference style:

Prompt: Close-up photo of a wild west anthropomorphic fox cowboy.

Prompt: A dynamic digital abstract splash painting of a man playing the saxophone.

Prompt: A retro-style poster for a hair salon named "Nancy Hair Studio", featuring a woman's head with he hair being dried with a hand hairdryer. Right below the salon's name a banner wit the text: "Opens at noon". Other text: "Coloring 20$" and "Haircut 10$".

Sometimes, using the General, Realistic, or Design styles gets you close to the image you’re aiming for — but a reference image can provide that extra something that makes all the difference.

Prompt: A black-and-white, silhouette photograph of a toddler in diapers, standing in a carpeted hallway at night, facing the camera and holding a teddy bear.

These are simple examples using only one reference image. You can use up to three, unlocking a wide range of creative possibilities and experimentations.

What is a Style code?

Whenever Random is used as a style when generating an image, a Style code is automatically associated with the result. This code represents the unique visual style applied, that contributed with you prompt to the generated image’s visual aspect of the final image.

You can find the Style code at the bottom of the right-side panel when viewing an image in detail.

To use that style again in another image generation, click the Plus icon next to the Style code. It will automatically be copied into the Style code field, ready for your next generation.

Style code example

An image was generated using the Random style with Magic Prompt turned off.

Prompt: A teddy bear playing with a huge ball. Magic Prompt: Off

The Style Code for the teddy bear image was 07175460. It was reused to generate the two other images, each with a different prompt shown in the captions.

Styles in version 2.0 and 2a

The styles available in version 2.0 and 2a are specialized AI models and will not add anything extra to your prompt, unlike version 1.0.

Here are the available styles:

• The Auto, General, Realistic and Design styles just like their counterpart in v3.0.

• 3D: This style is perfect for generating images of characters, various objects, and other creations with a 3D visual appeal.

• Anime: This style is ideal for creating images and characters in the likeness of your favorite anime.

Example styles in 2.0

Here are some examples of the 2.0 styles with the prompt: Portrait of a fireman :

Styles in version 1.0

Desktop

Mobile

In version 1.0, using a Style effectively adds the corresponding keywords to the end of your prompt just before generating the image, ensuring that the AI understands the type of image you are looking for.

Click on a style to apply it. With Version 1.0, you can select and apply multiple styles at once.

Example styles in 1.0