Google opens new image generator in four countries, English only: Write a prompt, get an image

Google announced a new AI tool, ImageFX, on February 1, 2024. Users type a description of the image they want, and the tool generates it.

ImageFX is powered by Imagen 2, which Google describes as its most advanced text-to-image model. The same model is also used in several other Google products: Bard, Search, Ads, Duet AI in Workspace and Vertex AI.

Keywords as a shortcut

The interface includes what Google calls "expressive chips." These are suggested keywords that make it quick to try variations of an image. Instead of writing a completely new prompt, users can swap out individual elements.

According to Google, ImageFX is built for experimentation and creativity. Users write a simple text description and can then adjust the image using the suggested keywords.

A limited start

ImageFX was first available through Google's experimental AI Test Kitchen. It launched only in English and only in the US, Kenya, New Zealand and Australia. Danish users were therefore not included at the outset.

Labeling and filters

The images receive an invisible watermark called SynthID and IPTC metadata. This is hidden information in the file that makes it possible to identify the images as AI-generated.

Google also says the tool has built-in filters against violent, offensive and sexually explicit content. There are also filters against images of named people. In addition, Google has carried out adversarial testing (red-teaming), in which testers try to make the system fail. The information about safety and the model's quality comes from Google itself and has not been independently verified.

Bard becomes more widely available

At the same time, Google expanded Gemini Pro in Bard to more than 40 languages and more than 230 countries and territories. Free image generation in Bard was opened in most countries.