Adobe’s image generation AI supports prompts in 100 languages. To use it, users must be activated for Firefly’s beta and log in to their browser. Among the languages are German, French, Hindi and Japanese, but also Klingon. The prompts are translated by Microsoft Translator. Until recently, the image generation AI could only be written to in English. Firefly’s user interface is to be translated into 20 languages. German, French, Japanese, Spanish and Brazilian Portuguese are already available.
Advertisement
Firefly works like text-to-image AIs like Midjourney. However, the two differ in how they are used: Midjourney expects prompt instructions as text input, in which all image properties and content are described in detail. Firefly has a graphical interface and only the image content needs to be written – and that’s enough in everyday language.
Firefly’s versions
Adobe has currently released four functions for Firefly’s web version: The text-to-image generation outputs four images. Generative Recolor generates color palettes based on the description text for vector graphics and assigns the colors to the vectors. The Text Effects feature renders typefaces in different materialities. Generative Fill can replace, add, or remove image content.
In addition, the image generation AI beta has moved into three of Adobe’s programs. In the beta version of Photoshop, Adobe’s photo editing software, you can select areas of an image and edit them using text you type. “Generative filling” has also arrived as a function in Adobe’s graphics program Illustartor. The Retype function recognizes and applies fonts in documents. Beginners can use Adobe’s Software Express to lay out flyers or social media posts, for example. With the addition of Firefly, users in the beta program can create layouts, images and animations with a text input.
Licensed content for training
In addition, Firefly differs from its competitors by the database on which the AI was trained. Adobe has only used images for Firefly that are available under license to them. This includes their own stock collection of 100 million images, openly licensed and public domain content where copyright has expired. In addition, Firefly embeds so-called content credentials via a tag to identify AI-generated content. The content credentials contain the name, date and the programs used.
Advertisement
These four AI generators are available in Firefly’s web service: Text to Image, Generative Fill, Text Effects, and Generative Recolor.
Firefly is accessible globally in countries where Adobe is available. Russia, Belarus and China, among others, are excluded from this. The languages and regions are listed in Adobe’s FAQ.
(str)