The race to dominate the AI space is getting tougher, and Amazon has just clocked in with the launch of its artificial intelligence service that can generate images from text captions and prompts.
The AI, which is called Amazon Titan Image Generator, was made to provide developers and enterprises advanced generative capabilities to build custom solutions.
The AI generator was unveiled during a keynote presentation at Amazon’s annual AWS re:Invent conference. According to the announcement, the service will be available in preview to Amazon Web Services cloud customers through the company’s Bedrock machine learning development platform.
According to the presentation, Titan Image Generator can create original images based on text descriptions provided by users. It can also customize existing images by intelligently modifying aspects like backgrounds.
Swami Sivasubramanian, Vice President of Amazon AI services, demonstrated on stage how the technology can seamlessly swap image backgrounds while retaining main subjects. He positioned Titan Image Generator’s creative functions as useful for generating lifestyle product images to enhance web content.
“We wanted a way to mark an image as created with AI, and specifically made with Titan Image Generator that will not interfere with the visual, have no latency, and cannot be cropped or compressed out,” Vasi Philomin, VP for generative AI at AWS, said.
The newly announced generator fits into Amazon’s broader family of Titan generative AI offerings including speech, text, and video generation tools. With Titan Image Generator’s launch, Amazon now provides services across the major categories of generative media powered by its artificial intelligence capabilities.
Amazon stated that the model was trained across diverse data sets covering a wide range of topics and applications. The company claimed built-in safeguards help Titan Image Generator mitigate issues like biases and toxic content.
However, specific details were not provided regarding Amazon’s training data sources and procedures. This could raise questions given recent controversies about other tech giants’ image generator training processes involving copyrighted material or sensitive content.
Nonetheless, Amazon said the underlying model will enable customizable fine-tuning to best suit customers’ needs. The company also outlined plans to legally indemnify clients against certain misuse claims regarding Titan-generated images.
Additionally, images created by Titan Image Generator will carry invisible forensic watermarking by default, according to Amazon. The company indicated this tamper-resistant tracking helps curb potential misinformation. Technical specifics about the form of watermarking were not included though.
As generative image models continue to advance rapidly, Amazon is positioning itself to provide its customers across sectors like ecommerce, publishing and advertising with sophisticated AI tools to create and adapt visual content.