Adobe launches Firefly generative A.I., which lets users type to edit images

Adobe on Tuesday launched an artificial intelligence tool called Firefly that will let users type commands to quickly modify images.

The first Firefly model is focused on the creation of images and text effects. A sample video from Adobe included a product demo for a “Generate Variations” option, for example. By highlighting an element in a multilayer work of art — a lighthouse, in the demo video — Adobe Firefly uses AI to generate different versions of the lighthouse.

In another example, Adobe shows how one might take a picture of a summer scene and type “change scene to winter day” to alter the image without any editing by the user.

Firefly can also automatically change the photo or image a user is working on by using generative AI to create a paintbrush based on something already in an image.

Adobe’s new product comes at a key inflection point both for Adobe and AI more broadly. The company spent $20 billion to acquire the design tool Figma in September and said at the time it would integrate features from other products into Figma.

Adobe said Firefly will launch first as a private beta.

AI investment has accelerated since the explosive debut of OpenAI’s ChatGPT. OpenAI and Stable Diffusion, another AI organization, both offer generative AI image products.

New AI tools have struggled with ethical concerns. For ChatGPT and similar products, those concerns deal with the propensity of the models to “hallucinate” plausible-sounding but inaccurate information.

For imaging technology such as DALL-E or Stable Diffusion, concerns have been raised about the models stealing artist content, regenerating it, and presenting it without credit or permission from the creator.

Adobe said Firefly will place an emphasis on giving creators “opportunities to benefit from your skills and creativity and protect your work.” The company already offers non-AI platforms that do that.