AI image generators have now become fairly common, and there is a very high chance of seeing AI-generated images on various online platforms. Almost all major companies are either developing their own versions of such tools or collaborating with others. However, Google's newest Google Labs experiment, Whisk, takes a different approach to generating images using Artificial Intelligence by letting users use existing images as prompts.
What is Whisk AI? Whisk is an AI tool that, instead of asking users to write detailed prompts for generating images, lets you upload pictures and use them as prompts, generating new images from those. The tool utilizes Google's Imagen 3 text-to-image model along with Gemini to generate images. Gemini analyzes the images you add and writes detailed prompts, which Imagen-3 then uses for generating images.
To use Whisk, you will first need to select an image for the subject, another one for the scene, and then one for the style. Whisk AI will remix these images together and create something entirely new. If you want, you can then use text prompts to refine the output. Keep in mind that the AI will not create an exact replica of the images you use as input since it only extracts the key characteristics from those. Because of this, the generated image can be different in various aspects, which makes Whisk ideal for exploring new ideas rather than detailed image editing.
Accessing Whisk AI
- Open a web browser and sign in to your Google account.
- Head to the official Whisk AI Google Labs page and click on the 'Let's Whisk' button.
- If you want to receive marketing emails and research invitations, click on the respective checkboxes on the next page to select them on the next page. Then click on the 'Next' button.
- Scroll down on the Privacy Policy page to accept it and click on the 'Next' button again.
- Finally, click on the 'Agree and Continue' button.
Using Whisk AI to generate images
- You will reach the homepage of Whisk AI. The dropdown button in the center of the page will have the 'Plushie' style selected by default. Click on it to select something else.
- Right now you have three style options to choose from: 'Plushie', 'Sticker', and 'Enamel pin'. Click on the one you want to select.
- After selecting a style, you can drag and drop an image you want to use as a subject in the box in the center. Alternatively, click on the box and upload an image from your computer.
- Once you upload an image, Whisk will take a few seconds to analyze the image before generating a new image from your input. Once the image is generated, click on the 'Open in tool' button below the generated image.
- Now, you can choose from two variants of the generated image and refine it. Click on any of the two images to select it.
- Hover your cursor over the selected image and then click on the 'Refine' button that appears on the top left.
- You can then add details to the image by typing them in the search box at the bottom.
Generating a new image from scratch
The above method lets you select from a few different styles and upload your own reference image. However, you can also generate images using your own custom references for style and scene as well as the subject.
- Click on the 'Start from scratch' button.
- Then click on the 'Subject' box to upload the image you want to use as the subject.
- After uploading your 'subject' image, click on the 'Scene' box and do the same for the scene.
- Finally, upload an image in the 'Style' box. Keep in mind that Whisk AI will try to generate the new image in the same style as the image you upload here.
- If you want, you can add additional details by typing them in the text box at the bottom.
- Once you've uploaded all your images and typed in the details, click on the 'Generate' button at the bottom right.
- Wait for a few seconds and Whisk AI will generate a couple of images from your references. You can then select either image and refine it in the same way as explained above.
Additional features of Whisk AI
Whisk AI also lets you change the aspect ratio, use random images as references, and go to the Whisk AI Discord directly from the website.
- When uploading images that the AI will use as references, you can click on the 'Dice' icon to use randomly generated images for the subject, scene, and style.
- To view and edit the prompt created by Gemini, hover your cursor over it and click on the 'Refine' button that appears on the top left. Then click on the button with a clipboard and pen icon next to the yellow 'Refine' button.
- Now you can see the prompt that Gemini has created and edit it. Once you are done, click on the 'Generate' button to refine the image.
- After generating an image, you can add it to your favorites, download it, and delete it by hovering your cursor over the image and clicking on the buttons that appear at the top right corner.
- Whisk AI also lets you select a custom template when generating an image by clicking on the hamburger menu at the top left.
- Then click on the 'Load Template' dropdown menu and select a template from the available options.
- Similarly, you can change the aspect ratio of the image to be generated using the 'Aspect Ratio' dropdown. You can also download all generated images using the 'Download All Images' button.
- To see all your generated images, click on the 'My Library' button at the top.
- If you want to visit Whisk AI's Discord page, click on the Discord icon at the top.
Things to know
- Whisk is not perfect and can make mistakes, which is why you should not rely on it when you need to generate images for professional purposes.
- The ideal use case for the tool is when you already have reference images using which you want to generate new ones. If you want to generate images from scratch, other AI image generators may deliver better results.
- Currently, Whisk is only available for users in the United States and only supports the English language. Google has not stated when it will be available globally.
- While you can use the AI tool from other regions using a VPN, keep in mind that the website will crash and reload frequently, which can cause problems when generating images.
Member discussion