Adobe's creative generative AI toolset, Firefly, is entering public beta, promising results that are "safe for commercial use."

With AI-generated imaging tools still emerging, Adobe is throwing its metaphorical hat into the ring with an open beta of its Firefly AI models. Available through and compatible with Adobe’s Creative Cloud suite, Firefly claims to be able to generate images, video, audio, brushes, and more using text prompts.

Commercial viability is a key focus for Adobe, which says that Firefly's initial models were built using the company's stock media and public domain content. It's a decision intended to avoid some of the recent problems with AI imagery, where many training models borrow from artists without warning (or permission). Future models, however, will use images, videos, etc. "from Adobe and others" as training data.

Users who do not want their Cloud work to be used to train future Firefly models can add a “Do Not Train” tag to a project’s Content Credentials. Adobe claims that this will prevent the project from ending up in the general pool. Many artists find this approach inadequate and feel that it should be opt-in rather than having to manually opt out on a case-by-case basis.