Whisk is Google Labs' answer to a problem text prompts handle awkwardly: describing a specific look. Instead of writing paragraphs of description, you feed Whisk three reference images, a subject, a scene, and a style, and it blends them into something new.

Drop in a photo of a person or object, a separate image for the setting, and a third for the visual style, and Whisk composes them into a single generated image using Google's Imagen models, remixing rather than literally collaging. Because it reasons over images instead of text, it captures nuances, a specific pose, a particular lighting mood, that are painful to type out, and its image-to-video extension animates the result into a short clip. Results feed back in as new inputs, so you can iteratively remix toward a final look.

Who is it for?

Creators who think visually rather than verbally, marketers assembling mood-board-style concepts fast, and anyone frustrated by how much prompt-writing skill affects image quality elsewhere.

How much does Whisk cost?

Free to use with a Google account, part of Google Labs' experimental tools, with usage subject to Google's standard rate limits for free experiments.

Our verdict

Whisk's image-in, image-out paradigm is a genuinely different way to direct generation, and it solves the "how do I describe this exact vibe" problem better than prompting does. As a free Google Labs experiment, it is worth having in your toolkit even if it does not replace a primary generator.