American technological company Google has presented a new online tool called ImageFX that allows users to create images based on text descriptions. The tool is based on the Imagen 2 image generation model from Google DeepMind.
ImageFX provides an intuitive user interface that allows users to experiment with image creation. It includes “expressive chips” which enable users to change the style, mood, color, and other parameters of the generated content.
In addition to ImageFX, Google also announced improvements for existing AI-based tools such as MusicFX and TextFX. MusicFX now allows users to create musical compositions up to 70 seconds long and use “expressive chips” to change the style and mood. Results can also be shared on social networks. TextFX, a tool for creating texts, has been made more convenient for users with improved navigation.
The developers have ensured that all new tools are built with the principles of responsible AI use in mind. ImageFX, in particular, is trained using safe and ethical data from the Imagen 2 model. The tool also has built-in filters to prevent the generation of offensive, violent, or explicit content, as well as a filter to create images of famous people.
To enhance security, all generated images and musical works are automatically marked with invisible digital watermarks using Google DeepMind’s Synthid technology. This allows for the tracking of content created using AI-based tools. Additionally, images generated by ImageFX have special IPTC metadata added for additional information about their origin.
The new AI-based tools are currently available to users from the USA, New Zealand, Kenya, and Australia on the Google Labs website.