Adobe introduces Firefly-AI tool for image-via-text generation

Adobe has announced Firefly, an AI tool that can generate images and text effects based on text descriptions. According to Adobe, it is a “new family of creative generative AI models” that continues to expand over time.

David Wadwani from Adobe explains that the former Firefly model, for which the beta is now available, can be especially valuable for those who want to create content for commercial use. He points to the fact that the machine learning models are trained on hundreds of millions of licensed images in Adobe Stock, combined with open-licensed and copyright-expired content. According to Adobe users can therefore be sure that no content is generated that is based on the rights of others.

Adobe says it has added a style engine to the model, allowing users to experiment with creating photo art, graphics, and text effects with a variety of styles, colors, lighting, and variations in composition. The company also plans to create models where makers can train on their own style.

Adobe says everything is designed to be integrated into other applications such as Adobe Creative Cloud, Adobe Expresss, and Document Cloud. The company plans to make Firefly available through an API as well, so it can appear on other platforms as well. When that API will be ready is unknown.