Adobe finally introduced Firefly, an AI-powered tool that allows users to quickly generate and modify images using typed commands. The initial Firefly model focuses on creating images and text effects.
In a product demo, Adobe showcased a “Generate Variations” option that enables the AI to produce different versions of a highlighted element in a multilayered artwork, such as a lighthouse.
Another example demonstrated how a summer scene photograph could be transformed into a winter scene with a simple typed command of “change scene to a winter day” without any editing required from the user. Firefly’s generative AI can also automatically alter the photo that a user is working on by creating a paintbrush based on an existing element in the image.
Adobe’s generative AI tools are differentiated by a significant aspect: the company is transparent about the data its models are trained on.
According to Adobe, the input data is either out of copyright, licensed for training, or sourced from the Adobe Stock library, which gives the system the benefits of avoiding artist backlash and being more brand-safe.
Costin, a representative for Adobe, stated, “We can generate high-quality content and not random brands’ and others’ IP because our model has never seen that brand content or trademark.”
Costin also shared that Adobe intends to compensate artists who contribute training data, although this will not be implemented at the tool’s launch. Before the system is out of beta, the company aims to establish a “compensation strategy,” which is currently being explored through various options.
Adobe’s latest offering comes at a crucial moment for the company and the broader field of AI. In September, Adobe spent $20 billion to acquire Figma, a design tool and announced plans to integrate Figma’s features with other Adobe products.
Related Stories: