Google Whisk has shut down: our old how-to, and how to use Flow now

What happened to Google Whisk

Whisk, Google's AI image tool, shut down on April 30, 2026. Its main features moved to Flow, Google's image and video creation tool. If you open Whisk's old address (labs.google/fx/tools/whisk) today, you are sent to Flow (flow.google.com). We checked the redirect on October 26, 2026.

Here are the main events in order.

HIKARI'S TAKEWhisk is gone as a separate tool, and Flow has taken over its image-combining features. If you're searching for how to use Whisk, what you want now is how to use Flow.

Why Whisk moved to Flow

Google describes this as a move, not a shutdown. In its February 25, 2026 blog post, Google said it was bringing the best features of its image experiments, Whisk and ImageFX, directly into Flow. The reason it gave: to let you create, edit and animate images in one workspace.

Flow started out as a tool for making and editing AI video. With this update, image generation moved to the center of the screen, and the images you make can go straight into a video as an ingredient or a first frame. Flow also gained a way to select part of an image and change only that area with a prompt.

Whisk could already turn images into short videos from April 2025. But making the image and animating it happened in separate places. In Flow, you move between the two on one screen.

HIKARI'S TAKEThe official reason is to let you make images and video in one place. I find it helps to think of it this way: Whisk's idea didn't disappear, it moved into a bigger toolbox.

Checking our 2025 Whisk how-to

In February 2025, right after Whisk became available in Japan, we published a step-by-step guide to it, with screenshots, in Japanese. Here is how what we wrote then holds up today.

Our biggest miss was not writing a single line about the rights to the images you put in.

We gave the example of using your own photo as the model, but said nothing about what happens if you put in someone else's photo, a famous character, or another company's product shot. In a tool that combines images, the images you put in are where rights problems start. We cover this in section 6.

HIKARI'S TAKEOur feature descriptions were mostly right for the time. What we got wrong was how we described availability in Japan, and the outlook. What we left out was rights.

If you made images in Whisk

The official Google Workspace blog (March 12, 2026) gave these instructions ahead of the shutdown.

1. A notice inside Whisk let you move your work to Flow. Nothing moved automatically; only people who opted in were covered.
2. Until the deadline, you could also download images directly from Whisk.
3. Images still in the library after April 30, 2026 are permanently deleted and cannot be recovered.

If you opted in, check your library in Flow. Whisk and Flow use the same AI credit system, so Google says any credits you had left carry over to Flow.

In countries and regions where Flow isn't offered, Whisk simply stopped working on April 30, with nowhere to move to. The US is on the list of countries where Flow is available.

HIKARI'S TAKEIf you didn't opt in, your Whisk images are gone. If you downloaded any, those files are the only copies left, so keep them somewhere safe.

How to use Flow to combine images, the way Whisk did

Making a new image from photos you already have is something you now do in Flow. Based on Flow's help pages, here is how to get started and how to make an image (checked October 26, 2026).

1. Open Flow on a computer. Go to flow.google.com in a Chromium-based browser such as Chrome or Edge, and sign in with your Google Account. Google recommends a desktop computer for the detailed editing tools.
2. Open a project and choose "Image". In the prompt box, click the model name and choose "Image". This is also where you pick the aspect ratio, the number of outputs and the model.
3. Add the images you want to combine as ingredients. Hover over an image in your project, open its menu and choose "Add to Prompt". It then serves as a reference for your prompt.
4. Describe each image's role and click "Generate". Whisk's "model", "background" and "style" slots are gone. Instead, write which image is for what, for example: "Put the dog from the first image in the meadow from the second, painted in the style of the third."
5. Use "Agent" to make changes. Turn on Agent, drag the image you want to edit into the prompt box, and describe what to change. The original image is kept.

You can choose from three image models: Nano Banana 2 Lite, the default model available at no charge; Nano Banana 2.1, which balances speed and quality; and Nano Banana Pro, which handles detailed instructions best. Nano Banana Pro is the default model on Google AI Ultra, the top plan.

Prices are from Google's Gemini subscriptions page for the US and credit amounts from Flow's help pages, both checked on October 26, 2026. On the same page, Google AI Ultra starts at $99.99 a month. Unused credits do not roll over to the next day or the next month. In the US, if you run out, Google AI Pro and Ultra subscribers can buy extra AI credits (Google excludes Japan from this); Google's help pages disagree on whether AI Plus subscribers can. Nano Banana 2 is also available in the Gemini app.

HIKARI'S TAKEWhisk's three slots are gone, but if you add images as ingredients and describe their roles, you can make the same kind of image. Prices and credit costs change often, so check the screen before you generate.

Rights when you combine images, and what Whisk's end teaches us

With a tool like Whisk, where you put images in, you need to think about the rights to those images before you think about the result.

Flow's help pages say your use must follow Google's Terms of Service, that Google won't claim ownership of original content you generate, and that there are policies against generating harmful or illegal content. Images and videos made in Flow also carry SynthID, an invisible digital watermark. Google not claiming ownership does not mean you are free to infringe someone else's rights.

In the US, the Copyright Office's report on copyrightability (Part 2, January 2025) says prompts alone do not make you the author of an AI output. But when you put in your own copyrightable work, such as a photo you took or a drawing you made, and it is still perceptible in the output, you remain the author of at least that part. The other side follows from general copyright principles, AI or not: an output that copies someone else's work and is substantially similar to it can infringe. When you use that work as a reference image, the copying is easy to show. Using real people's faces raises a separate issue, the right of publicity under state law. The Copyright Office's Part 1 report (July 2024) looked at AI digital replicas and called for a federal law.

With that in mind, we settle these four things first.

1. Only put in images you took yourself or have permission to use. No photos of other people's faces, no anime or game characters, no other companies' logos or product shots.
2. Have a person check the result before using it for work. Look for anything too close to the source images or to existing works.
3. Keep what you make. In Whisk, images that weren't moved disappeared after the deadline. Save yours to your company drive.
4. Keep the recipe, not the tool. Which photo, in which scene, in what style. That note carries over to the next tool as it is.

HIKARI'S TAKEIn a tool you feed images into, what you put in matters most. And save what you make somewhere outside the service, too.