Photoshop has integrated more AI, allowing users to easily edit photos as they want
- Tram Ho
Adobe software company recently updated a new tool for the beta version of the famous photo editing application Photoshop called “Generative Fill”, which uses AI and cloud technology to assist users in editing photos. This is considered a very noticeable update of Photoshop, when for the first time users can select and edit part of an image by entering text.
Generative Fill can be considered as a usable version integrated into Photoshop of Adobe Firefly – Adobe’s custom image compositing model, similar to AI image creation models (based on user commands). like DALL-E, Stable Diffusion. As a deep learning AI model, Firefly has been trained on millions of images in Adobe’s photo library, thereby generating outputs based on text prompts from users.
Original image (left) and image edited with Generative Fill (right) after using text commands to add a car, a similarly shaped cloud in the sky, and a puddle . Photo: Engadget
To use a Generative Fill, the user selects an area of an existing image that they want to modify. Once selected, a “Contextual Task Bar” will pop up, allowing the user to enter a description of what they would like to see created in the selected area.
For example, if you want the sky in the photo you’ve taken to be more ‘surreal’, select the area and type something like ‘surreal sky with weird colors’ in the taskbar. Or, if the angle of your shot is a bit narrow, you can select the area outside the photo and prompt the AI to expand the scene and change the aspect ratio.
Photoshop will then send this data to Adobe’s servers for processing, which then returns the results in the application. The user can of course choose between the outputs, or can enter another description to create the desired output image.
Here, Generative Fill uses a well-known AI technique called “inpainting”, which allows the AI to recognize and seamlessly combine the composite image into an existing image. Accordingly, these changes are automatically adjusted to the perspective, lighting, and style of the selected image. When used, the Generative Fill tool creates a “Creative Layer”, allowing the user to make changes without affecting the image content. Generative Fill can then create and insert completely new objects.
The user localizes the object to be edited here is a deer, then enters the command “A wet alley at night”. AI will immediately insert the deer into the background scene according to the command, with a very high degree of realism and naturalness. Photo: Engadget
Besides, Generative Fill also allows users to remove part or completely parts of an image, or even expand the size of the image by creating new content around the existing image. This is an AI technique known as “outpainting,” where AI-generated content goes beyond the frame.
It must also be added that the above features have been available in famous AI imaging models such as DALL-E since August last year. Until now, Adobe has updated these features for its famous photo editing software.
Reportedly, Generative Fill will be available in beta version of Photoshop for PC starting today. Adobe adds that the feature will be “widely available in the second half of 2023.” In parallel, Generative Fill is also now available on the web as a module in Firefly’s beta.
Synthetic
Source : Genk