Adobe has made an exciting announcement at its MAX conference for creatives. The tech giant has unveiled major updates to Firefly, its generative AI image creation service. The newly introduced Firefly Image 2 Model is set to offer enhanced realism in image rendering, particularly when it comes to human elements such as facial features, skin, body, and hands.
Adobe has introduced the Firefly Image 2 Model, which offers improved image rendering capabilities, particularly for human features. Firefly users have generated an impressive three billion images to date.
Since its launch approximately six months ago, Firefly’s users have already generated a staggering three billion images, with one billion alone in the past month. The majority of these users are new customers to Adobe, with 90% of Firefly users being completely new to Adobe’s suite of products. This surge in usage is likely a result of the transformation of the Firefly demo site into a full-fledged Creative Cloud service.
According to Alexandru Costin, Adobe’s VP for generative AI and Sensei, the new model has undergone significant enhancements. It has been trained using more recent images from Adobe Stock and other commercially safe sources, making it more proficient in creating realistic and detailed images. The size of the model has been increased three-fold, essentially creating a larger and more capable “brain” that can efficiently render high-quality pixels and details.
While the larger model requires more resources, Adobe has worked on optimizing its performance to ensure it runs at the same speed as the previous model. The company has invested in distillation, pruning, optimization, and quantization techniques to strike a balance between quality and cost efficiency. Despite this focus on quality, Adobe also plans to bring the new Firefly Image 2 Model to Creative Cloud apps like Photoshop, expanding its benefits beyond the Firefly web app.
Adobe’s approach to generative AI centers around generative editing rather than content creation. Its customers have leveraged generative capabilities to enhance existing workflows by optimizing and improving images rather than creating entirely new assets. This perspective has yielded successful features like generative fill in Photoshop, enabling users to enhance photoshoots and product images.
To further empower users, Adobe has introduced new controls in the Firefly web app. Users can now customize settings like depth of field, motion blur, and field of view for their images. Moreover, users have the option to upload an existing image and have Firefly match its style. An auto-complete feature has also been added to prompt writing, facilitating quicker and more accurate image generation.
With this significant update to Firefly, Adobe is solidifying its position in the generative AI game, providing users with a powerful and versatile tool for creative image editing and enhancement.