At its MAX conference, Adobe revealed substantial upgrades to the AI models that enable Firefly, its generative image creation service.
According to Adobe, the new Adobe Firefly Image 2 model delivers enhanced image quality, especially for rendering detailed elements like human faces, skin, hands, and bodies.
The announcement comes as Firefly usage has skyrocketed, with 3 billion images created in total and 1 billion last month alone. Over 90% of Firefly users are new to Adobe products.
New Model Boasts 3X Larger Size for More Realistic Results
Adobe’s Alexandru Costin said the new model is significantly larger, about 3 times the size of the original models. This expanded “brain” enables stronger connections to produce more lifelike pixels and details based on user prompts.
The model was also trained on nearly twice as much data from Adobe Stock and other sources to improve its understanding of desired outputs.
Despite its greater resource demands, Adobe focused on maintaining similar generation speeds to the original model through optimization techniques.
Rollout Across Firefly Web App and Creative Cloud
Initially, the new model will be available through the Firefly web application. It will then come to Creative Cloud apps like Photoshop for features such as generative fill in the near future.
Costin stressed Adobe sees generative AI as enabling creative editing workflows rather than just generating standalone images.
New Controls for Depth, Motion Blur, Style Matching
The Firefly web app gains new controls to adjust image depth of field, motion blur, and field of view. Users can also upload an existing image for Firefly to match its style.
An auto-complete feature for prompts helps users refine wording to get better results faster.
Adobe’s Focus on ‘Generative Editing’ Over Text-to-Image
According to Costin, Adobe defines its generative AI approach as “generative editing” rather than just text-to-image. This better serves customer creative workflows.
For example, Photoshop’s popular Generative Fill feature allows enhancing existing assets like photoshoots versus creating wholly new images.
While enabling content creation, Adobe believes applying generative AI to complement and augment creative projects is more impactful for its users.
Surging Interest in AI Image Generation
Adobe’s upgrades come amid booming interest in AI image generation following the launch of tools like DALL-E 2 and Stable Diffusion in 2022.
The field’s rapid advances have compelled creative software giants like Adobe to invest heavily in new generative capabilities.
Democratizing Access While Addressing Ethical Concerns
Adobe aims to make sophisticated generative AI accessible to mainstream users through Firefly’s ease of use. But it recognizes open issues around intellectual property and artistic integrity.
That’s why Adobe trained its models on licensed content and instituted policies requiring attribution. It seeks to democratize access responsibly.
Stepping Up Competition in Generative AI Space
With the new models and surge in adoption, Adobe is asserting itself as a leader in bringing enterprise-grade generative AI to market.
Competitors like OpenAI with DALL-E and Stability AI will need to keep pace with Adobe’s aggressive rollout of new capabilities and tools.
The improvements also widen Adobe’s edge over smaller startups lacking its resources and community. Adobe appears determined to dominate the new generative frontier.
Transforming Creative Workflows with AI Assistance
Adobe is guiding customers into an emerging era where AI collaborates with humans in creative projects, rather than replaces them fully.
The new Firefly updates move Adobe closer toward this vision, providing creators powerful new tools to enhance their unique skills rather than replace them.
With the new Firefly model’s increased realism and control, creative professionals get an AI partner to build upon their talents even further.
The Bottom Line
Adobe’s major upgrades to Firefly deliver substantially more capable AI image generation to its wide community of artists and designers.
By combining expansive datasets, enlarged AI models, and a focus on augmenting human creativity, Adobe is striving to set the standard for responsible and beneficial generative AI.
