Google releases AI tool that creates images from other images

Google has unveiled Whisk, a new AI tool that generates images using other images as prompts. Unlike older tools that rely heavily on long, detailed text descriptions, Whisk reportedly allows users to upload images to define the subject, scene, and style they want to generate.

The tool also supports multiple images for each of these inputs, giving users significant flexibility to experiment. If you don’t have an image to start with, Google has a dice icon that auto-fills the prompts with visuals.

However, these placeholder images appear to be AI-generated as well. Users can also add a bit of text to fine-tune the output, but the text box is optional. Whisk still churns out results without it, according to the company.

The tool then generates an image and a corresponding text prompt that was used in its creation. If you like the result, you can favorite it or download it. If you don’t, you can refine it by editing the underlying text prompt or generating something entirely new.

Imagen 3 powers Google’s Whisk AI

Google says Whisk is built for “rapid visual exploration, not pixel-perfect edits.” So while it’s flexible and fun to use, it won’t deliver flawless results every time. The company admits the tool can “miss the mark,” which is why they allow quick edits for better accuracy.

Whisk runs on Google’s latest Imagen 3 image generation model, which the company also announced alongside the tool. Imagen 3 powers Whisk to produce its AI-generated results, and the underlying technology is said to improve the overall quality of the outputs.

Alongside Imagen 3, Google also introduced Veo 2, its upgraded video generation model. Veo 2 reportedly has a better grasp of cinematography and fixes common issues like hallucinated features—for example, the classic AI error of extra fingers.

The model is first launching on Google’s VideoFX platform, which is still in limited testing through Google Labs. The company says Veo 2 will eventually expand to YouTube Shorts and other Google products sometime in 2025.

For now, the company positions Whisk as an exploration tool rather than a professional editing solution. Whether it holds up against competitors like OpenAI’s DALL-E remains to be seen.

From Zero to Web3 Pro: Your 90-Day Career Launch Plan