Adobe launches three new AI models, comprehensive upgrade of image editing features

Adobe is fully developing AI and today announced the launch of three new generative AI models, adding powerful features to Illustrator and Adobe Express, and significantly improving Photoshop's text-to-image capabilities. At the Adobe Max event on Tuesday, Adobe released the Firefly Image 2 model, the latest version of the original Firefly AI image generator, which powers popular features such as generative filling in Photoshop, and also introduced two new Firefly models for generating vector images and design templates.

Adobe states that compared to its predecessor, the new Firefly Image 2 model can generate higher quality images, especially when rendering realistic characters, with high-frequency details such as leaves, skin texture, hair, hands, and facial features. Images generated using the Firefly Image 2 model have higher resolution, more vibrant colors, and greater color contrast.

The Image 2 model also introduces new AI editing capabilities to help users customize the results. Users can manually or automatically apply photo settings to adjust the depth of field, motion blur, and field of view of the generated image, just like manual camera controls. The "Prompt Guidance" feature has also been added to help users improve the wording of text descriptions and automatically complete prompts to improve efficiency.

In addition, Adobe has also introduced a new "Generate Match" feature that allows the generated content to match the style of specific images. Users can choose from a pre-selected list of images or upload their own reference images to replicate the style and control similarity using sliders. The final output will automatically include content authentication as a digital "nutrition label" generated by AI, with additional attribution metadata and image identification.

In a dedicated blog post, Adobe's Head of Design, Scott Belsky, stated that the company has developed "new policies and safeguards" to protect against misuse of the Generate Match feature. The feature will prompt users to agree to Adobe's terms of use and confirm that they have the right to use the uploaded images, while storing thumbnails of uploaded content on Adobe's servers (not used for training AI models) to provide some accountability. During the feedback solicitation period, the Generate Match feature will remain in beta and users will not be allowed to use it for commercial purposes.

However, it seems difficult to truly prevent users from imitating protected content, which may deepen the divide between Adobe and creators who oppose having their style replicated by AI. Currently, this system seems to limit Adobe's liability rather than prevent imitation.

Firefly Image 2 is now available for testing through the web version of Firefly and will soon be available for Creative Cloud applications. This means that it cannot be used in Photoshop (standard, beta, or web version) yet, but at least you can compare it with the original Firefly image model before its official release.

Adobe has also introduced a new Firefly vector model for Adobe Illustrator, claiming it to be the "world's first generative AI model for vector graphics." Adobe's Firefly vector model is now available in the Firefly beta, allowing users to create editable vector images using text prompts and automatically segmenting each element of the graphic into "logical" groups and layers. Unlike traditional JPEG and PNG files, vector graphics (also known as SVG files) are ideal for creators such as logo designers because they can be scaled to any size without affecting overall image quality.

Like the original Firefly text-to-image model, Adobe states that its Firefly vector model is safe for commercial use (after the beta period) as it was trained on licensed content such as Adobe Stock and expired copyright public domain content. The Firefly vector model can be tested through the Adobe Illustrator beta and also offers other beta features such as Mockup (realistically showcasing designs on 3D models) and Retype (recognizing and editing vector fonts).

Finally, Adobe has launched the Firefly Design model for generating customizable print materials, social media posts, online ads, videos, and other templates. Adobe's Firefly Design model uses the new text-to-template beta feature in Adobe Express to generate fully editable templates based on text prompts, suitable for "all common aspect ratios." It has some similarities to Canva's Magic Design feature (another all-in-one design platform competing with Adobe Express), where users can describe content like "beach vacation flyers" to generate unique templates instead of placing individual text and image assets on a blank canvas.

These new models may currently only be available in beta, but if we consider the release schedule of Photoshop's generative filling feature, it may not be long before they are officially launched. According to Adobe, the original Firefly model has been used to generate over 3 billion images so far. In the past year, other companies such as Canva and Microsoft have also launched various AI-based creative tools, so perhaps releasing their own AI innovations is the best way for Adobe to stay competitive.