With the release of the Adobe MAX conference in Los Angeles, Adobe has announced a bunch of AI services and features. The major highlight of the conference is their upgrade to the Image to Text model, Adobe Firefly Image 2, putting it directly against the popular choices: Midjourney and DALL-E 3.
The date also marked the entry of new AI models: the Firefly Vector and Design Model. Let’s break down the key points to take away from the conference:
Adobe Firefly Image 2 Overview:
Additions:
Generative Match
So, Adobe Firefly Image Model 2 includes the Generative Match feature. In this feature, users can provide pictorial references to the image style they want to generate from a pre-selected list or upload it from the device. The AI then evaluates all the images and creates new art replicating the supplied files, which might help imitate brand guidelines and design elements.
Content Credentials
Another addition is the Content Credentials, FireFly’s way to credit the source. Through Adobe Creative Cloud, this labeling mechanism applies metadata to the generated portrait, classifying whether the image was AI-generated or based on a reference image and, in that case, the citation to that image.
- More AI related:
Improvements:
Text Prompt Capabilities
Improving on its previous generation, the new Image 2 Firefly model allows it to understand user prompts better. The prompter also receives suggestions or auto-completions based on their input, allowing for effective and efficient prompts.
Image Generation
The Image Generation part itself has been improved as a result of more extensive training. The 2nd generation of Firefly Image enables more photorealistic imagery, as they quote in their conference. Furthermore, enhanced control over motion blur, depth of field, and lens settings contribute to the higher-fidelity details of the Generative AI.
By the way, do you know that you can also play around with this version of the Adobe Firefly Image Model? Go to firefly.adobe.com and find its Beta version link there, and the thing is, it is free.
The fun part is that creators can manually tinker around with these settings, giving them more adjustment powers. Additionally, Auto Mode picks the image creation style automatically between ‘art’ and ‘photo’ and also determines a more suitable configuration for the picture based on the text input.
The rest of the Adobe MAX conference
As I mentioned, there are two new models released to Firefly.
Firefly Vector Model
One is a Firefly Vector Model for Adobe Illustrator, allowing for generative vector graphics to be a thing. Creating a vector illustration might be more straightforward, thanks to all the tooling that this model delivers. Some of them are Generative Match like Firefly Image 2, gradients, seamless pattern tiling, and precise geometry.
Firefly Design Model
The final one is the Design Model, which focuses on helping designers create templates using AI. Like any other model in Adobe Firefly, this design engine allows prompts for development templates in Adobe Express while still being fully automated, editable, comprehensive, and whatnot.
Another thing that was announced at the MAX conference was Project Primrose.
- Meanwhile, check out what the ChatGPT Android App brings to the table.
