Adobe
On Tuesday, Adobe announced Firefly, a new AI image synthesis generator. Unlike other AI art models such as Stable Diffusion and DALL-E, Adobe says its Firefly engine, which can generate new images from text descriptions, is trained solely on legal and ethical sources. and its output is made available for use by commercial artists. It will be integrated directly into Creative Cloud, but is currently only available as a beta.
Since the image synthesis model became mainstream last year, the field has been plagued with ethical and copyright issues. For example, an AI art generator called Stable Diffusion gained the ability to generate images from text descriptions after a researcher trained his AI model to analyze hundreds of millions of images collected from the internet. Many (probably most) of these images were copyrighted and were obtained without the consent of the rights holders, leading to lawsuits and protests from the artists.
To avoid these legal and ethical issues, Adobe creates AI art generators that are trained only on Adobe Stock images, open license content, and public domain content, ensuring that the content generated is commercial. I have confirmed that it is safe to use. Adobe provides more details in a news release.
Adobe Firefly consists of multiple models, tailored to serve customers with a wide range of skill sets and technical backgrounds, and works across a variety of use cases. Our first model was trained on Adobe Stock images, open license content, and public domain content with expired copyrights, focusing on image and text effects to produce content that is safe for commercial use. Designed. Hundreds of millions of professionally licensed images on Adobe Stock are among the highest quality on the market and help prevent Adobe Firefly from generating content based on his IP of other people and brands . Future Adobe Firefly models will leverage a variety of assets, technologies and training data from Adobe and others. As other models are implemented, Adobe continues to prioritize countering potentially harmful biases.
In addition to our commitment to a more ethical form of AI generator, Adobe doubles down on our ethics by tagging our content as “don’t train” for creators who don’t want their content used to train models. According to Adobe, this tag “remains associated with the content regardless of where the content is used, published, or stored.”
In terms of performance, Adobe Firefly is similar to OpenAI’s DALL-E, but Ars has yet to evaluate its performance. Demos on the Firefly website show you “Text to Image” (create your own image from a text description), “Text Effects” (apply styles and textures with text prompts), and “Vector Recolor” (your own It introduces various functions such as creating variations of works from the text description).
Whether or not artists will adopt Firefly into their workflows remains to be seen, but the new AI model at first glance looks like a positive step for advocates of AI ethics. As always, we have to take Adobe’s claims with a grain of salt.
