Adobe announced today that it has upgraded the models that power Firefly , its artificial intelligence- powered image creation service .
According to Adobe, the Firefly Image 2 Model will be improved in rendering human features such as the face, skin, body, and hands. These improvements have effectively addressed previous issues that have arisen with similar models.
Adobe today announced that Firefly users have created over three billion images since the service launched about half a year ago, with one billion images created in the past month alone. It also found that 90 percent of Firefly users are new to Adobe products. The majority of these users use the app , which explains why the company recently decided to turn the Firefly demo site into a full Creative Cloud service.
Alexandru Costin, Adobe’s vice president of genetic AI and Sensei, said the new model was not only trained on more recent images from Adobe Stock and other commercially safe sources, but it’s also significantly larger. He said Firefly is made up of a set of multiple models and has grown in size by three times. So it’s like a brain that’s three times bigger and has the ability to make those connections and visualize images in more detail, delivering more beautiful pixels and details for the user. In addition, the company has nearly doubled the data set, giving the model a better understanding of user needs.
This larger model requires more resources, but Costin noted that it should operate at the same speed as the first model. “We continue to explore and invest in cuts and optimization. There’s a lot of work to be done to make sure customers have a consistent experience, but we don’t make the cost of the cloud.” Right now, Adobe is focused on quality, not optimization.
See also: Adobe: Critical zero-days threaten Acrobat and Reader
For now, the new model will be available through the Firefly app. However, it will soon be available in Creative Cloudlike Photoshop, enhancing popular features like Generative Fill, which allows you to add and remove content from an image. This integrated approach is also highlighted by Costin. Adobe is focusing on genetic AI not only for content creation, but also for genetic editing, as mentioned.
What we’ve seen from our customers, and the key to the success of Photoshop genetic editing, is not just creating new assets, but also taking existing assets – a photo, a product shot – and using the production capabilities to significantly improve the work. That’s why we call genetics the most advanced form of editing, rather than simply converting text to image, as we believe that’s more important.
With this new model, Adobe is also introducing some new features to the Firefly website, allowing users to adjust the depth of field of their images, as well as motion blur and field of view settings. There’s also the ability to upload an existing image and then link it to the Firefly app’s style, as well as a new auto-complete feature as you type (which Adobe says has been improved to help you achieve better results).
Source: techcrunch.com
