Adobe has announced three new generative AI models that are set to introduce powerful new features and experiences across its various products, including Photoshop, Illustrator, and Express. On Tuesday, the computer software company unveiled Firefly 2, Firefly Design, and Firefly Vector models at the annual Adobe Max event.
In what has been a busy year for artificial intelligence technologies, Adobe crashed the party back in March with the debut of its text-to-image generating large language model (LLM) dubbed “Firefly”.
The graphic design software giant called its generative AI an “ethical alternative” to rivals Stable Diffusion and Midjourney and said the tool does not infringe any copyrights as it is being trained using stock images derived from Adobe Stock.
Following a successful beta testing phase, Adobe brought Firely as an added feature in Photoshop, Illustrator, Experience Manager, and Express. It is available as “Generative Fill” and “Generative Expand” in Photoshop, “Generative Recolor” in Illustrator, and “Text to Image” and “Text Effects” in Adobe Express.
Firefly 2 For Photoshop Now Renders Photorealistic Humans With Detailed Features
Now, Adobe has introduced the second generation of its AI image generator called the Firefly 2, which the company says is capable of generating high-quality images with an added focus on fine details like foliage, skin texture, hair, hands, and facial features when rendering photorealistic humans. The Photoshop developer also said images generated by the Firefly 2 image model will be of a higher resolution and feature more vivid colors and contrasts.
The revamped model also introduces new AI-powered image editing features that allow users to customize various aspects of a photo, either automatically or manually. Users can adjust the depth of field, motion blur, and field of view of a Firefly-generated image just like toggling manual controls in cameras.
There is a new “Prompt Guidance” function that streamlines the image generation process by optimizing users’ text prompts to improve the wording of their descriptions, alongside automatically completing prompts to boost the system’s efficiency.
Adobe also introduced another feature called “Generative Match” which allows users to match the style of generated content to that of a specific image. Users can pick a style from a preselected list of images or upload their own to replicate the style. They can also control the resemblance of a style to the generated image with a slider.
The feature is mainly aimed at companies that want to implement their brand style across various products. For the time being, Generative Match will remain in beta, and users won’t be permitted to use the feature commercially until the end of the trial period.
Text-to-Vector Model Is Coming To Adobe Illustrator
Adobe Max also saw the rollout of the “Firefly Vector” model that is specifically designed for its vector graphic software Illustrator. The text-to-vector algorithm, currently in beta, allows users to create editable vector images based on text prompts. Adobe says the AI can automatically split each element of the vector graphic into “logical” groups and layers.
Vector graphics are ideal for creatives like logo designers as they can be scaled to any size and edited without losing image quality, unlike bitmapped graphics such as JPEG and PNG image files.
Similar to the Firefly text-to-image model for Photoshop, Firefly Vector is also being trained on licensed content featured on Adobe Stocks and public domain content that has expired copyrights. Users can test out the AI on Adobe Illustrator beta, alongside additional features like “Mockup”, which displays vector designs on realistic 3D models, and “Retype” which identifies and edits vector-based fonts.
For more information : Chipmaker AMD Acquires AI Software Startup Nod.ai
Firefly Design Can Create Web And Online Ad Templates From User Prompts
Last but not least, Adobe introduced the Firefly Design, which is a text-to-template model that generates customizable templates for print media, social media posts, online advertising, video, and web on Adobe Express. Firefly Design uses text prompts to create editable templates of all aspect ratios.
Firefly 2, Firefly Vector, and Firefly Design models are currently in beta and Adobe is yet to reveal when the features will be fully rolled out commercially. The company is using the training period to collect valuable feedback from its creative customers to improve the generative AI tools before releasing them alongside new versions of its Creative Cloud applications.
Adobe Updates Policy Guidelines For AI-generated Content
The company also developed new policies and safeguards to protect AI-produced images from being abused. Before using the content, users will have to agree to Adobe’s terms and conditions and also confirm that they have the right to use an uploaded image to create a certain style or customize it.
Additionally, Adobe will also store a thumbnail of the uploaded image on its servers for further accountability, which won’t be used to train its AI models.
For more information : China Unveils New Quantum Computer Prototype In Computing Power Fight Against U.S.