How to create AI images with Google Gemini, ImageFX, SGE

You can now create AI images with Google Gemini, ImageFX, and Search, and here's how.

Google Bard and ImageFX
Google Bard and ImageFX / Image: Mauro Huculak

Google Gemini (formerly Bard) can now create images with AI, and in this guide, I’ll outline how to complete this task. Starting with the update that Google rolled out on February 1, 2024, makes it possible to compose a text prompt to convert it into an image. (Currently, only in English.)

Previously, the company added the image generator technology to its Search Generative Experience (SGE), but it was not available through Gemini. In addition, Google is also adding a dedicated tool known as “ImageFX” to create images with AI.

In this guide, I will outline the steps to create images with Google’s AI technology through the Chrome browser on Windows 11. However, these instructions also apply to any supported platform.

Create AI images with Google Gemini

To create an image with AI with Gemini, use these steps:

  1. Open Gemini (web).

    Quick tip: You can also install the chatbot AI as an app on Windows 11.
  2. Sign in with your Google account (if applicable).

  3. Create a very descriptive prompt to create an image. For example, “Act as a professional artist and create an image of a realistic bear drinking coffee in the forests.”

    Google Bard create image

    Quick note: The more information and details you add, the more accurate the image will be.
  4. Click the Generate more option to create two additional variants.

  5. Click the image to enlarge.

  6. Click the Download button to download the full resolution.

    Bard download AI image

Once you complete the steps, the chatbot will generate two variants of the prompt. You can create another group of images, but if you want more images, you will have to enter a new prompt.

Also, while the preview shows a small image, the download is a “.png” image of 1536 x 1536 pixels. On the other hand, Microsoft Copilot’s default “.jpeg” image size is 1024 x 1014 pixels.

Create AI images with ImageFx

To create images with AI through Google’s ImageFx, use these steps:

  1. Open AI Test Kitchen (web).

  2. Click the Sign in with Google button.

    AI Test Kitchen

  3. Click the Sign in button.

  4. Click the Next button.

    Quick note: You can skip the email marketing options.
  5. Click the Next button to agree to the terms.

  6. Click the Agree and continue button.

  7. Choose the ImageFx from the “Home” menu in the top-left corner.

    Google ImageFX

  8. Click the Got it button to continue.

  9. Create a very descriptive prompt to create an image. For example, “Act as a professional artist and create an image of a realistic bear drinking coffee in the forests.”

    ImageFX create AI image prompt

  10. (Optional) Choose the image type. For example, photorealistic, 35mm film, minimal, etc.

    Quick tip: You can click the “More” button to generate more types of image options.
  11. (Optional) Click the Tune button from the right and choose a seed number to fine-tune the creation of the image. (The greater the number, the more tuned the creative process will be.)

  12. Click the Generate button.

  13. Review the generated images.

  14. Click the Download button to save the image on your computer.

    ImageFX download AI image

After you complete the steps, the ImageFX tool will create an image using the context of the prompt. The tool will generate four variants of the image at 1536 x 1536 pixels.

If you want to create additional images, click the “Start over” button and compose a new prompt. If you want the AI to create an image automatically without a prompt, click the “I’m feeling lucky” button.

Create AI images with Google Search

To create an image with AI on Google Search (SGE), use these steps:

  1. Open a web browser.

  2. Open the Google Search page.

  3. Create a very descriptive prompt to create an image. For example, “Create an image of a realistic bear drinking coffee in the forests.”

    Google SGE cerate AI image prompt

  4. Choose the best version of the image that matches your intent.

  5. Click the Edit button.

    Google Search AI edit image

  6. Update your prompt to refine the image (as necessary).

  7. Click the Generate images button.

  8. Select the AI image that Google Search has generated.

  9. Open the Export menu and choose the Download option to save the image on the computer.

    Google Search download AI image

Once you complete the steps, the image will download as a “.png” file that you can use anywhere.

When using Google Search AI to create images, for the best results, consider these tips:

  • Start your prompt with the intent keywords like “draw a picture” or “create an image.”
  • Offer as many details as possible for a more accurate result.
  • Use the “Edit” option to refine your prompt.
  • Provide the dimensions of the image, as the AI model will always create a small rendering by default. Use keywords like “create a large image” or specify the width and height in pixels (the chatbot may ignore the specific size, but the end result will be a larger image).

You may also see the option to create images with AI from the “Google Images” page while searching for different pictures. If the feature is available, Google will show a “Create something new” block in the page results with the “Generate” button to create a new image with AI.

If you can generate images from the search page, you can also try signing in with your Google account and, from the Labs page, opt into the Search Generative Experience (SGE).

Google Gemini vs. Microsoft Copilot AI image creation

When comparing the AI images from Gemini and Copilot, both model languages (Google’s Imagen 2 and OpenAI’s DALL-E 3 that Microsoft uses) produce great results. However, in my opinion, the Microsoft Image Creator produces more details on the image with a simple prompt.

Now, let’s talk about the user interface and features. Google Gemini seems faster at rendering AI images, but this could also be because it only generates two image variants at a time. Also, I found no limitations in using the technology that Google offers.

Cat riding a dragon
Cat riding a dragon / Image: Gemini

On the other hand, if you use the free version of Copilot, image creation now takes several moments, and it produces three variants of the image. Also, you only get 15 tokens per day to create images. However, if you also count the image creator feature built on the Paint app, then you have an additional 50 tokens.

Cat riding a dragon
Cat riding a dragon / Image: Copilot

While looking at the standalone image creator for Google and Microsoft, the one from Google seems more straightforward and, at the time, more robust, with different options to help you generate the best image possible. The interface even offers an option to fine-tune the image.

The one from Microsoft only includes a prompt box with samples to help you spark creativity and compose a good prompt.

It’s important to note that Gemini is now officially powered by Gemini Pro. This more advanced language model is better at understanding and summarizing content, reasoning, brainstorming, writing, and planning, but it’s not as good as the Gemini Ultra.

Update February 13, 2024: This guide has been updated to reflect that Google Bard is now Gemini.

About the author

Mauro Huculak is a Windows How-To Expert who started Pureinfotech in 2010 as an independent online publication. He has also been a Windows Central contributor for nearly a decade. Mauro has over 15 years of experience writing comprehensive guides and creating professional videos about Windows and software, including Android and Linux. Before becoming a technology writer, he was an IT administrator for seven years. In total, Mauro has over 21 years of combined experience in technology. Throughout his career, he achieved different professional certifications from Microsoft (MSCA), Cisco (CCNP), VMware (VCP), and CompTIA (A+ and Network+), and he has been recognized as a Microsoft MVP for many years. You can follow him on X (Twitter), YouTube, LinkedIn and About.me. Email him at [email protected].