Amazon Upgrades AI Image Generator

Amazon has unveiled an enhanced version of its in-house image-generating model, Titan Image Generator, for AWS customers utilizing its Bedrock generative AI platform. The updated model, now called Titan Image Generator v2, introduces several new capabilities, according to AWS principal developer advocate Channy Yun in a blog post.

Users can now guide the images they generate using reference images, edit existing visuals, remove backgrounds, and create variations of images, Yun explains. “Titan Image Generator v2 can intelligently detect and segment multiple foreground objects,” he writes. “With Titan Image Generator v2, you can generate color-conditioned images based on a color palette and use the image conditioning feature to shape your creations.”

Titan Image Generator v2 supports image conditioning, which allows the model to take in a reference image and focus on specific visual characteristics such as edges, object outlines, and structural elements. It can also be fine-tuned using reference images like a product or company logo to ensure that generated images maintain a consistent aesthetic.

AWS remains vague about the specific data used to train its Titan Image Generator models, only disclosing to TechCrunch that it is a combination of proprietary and licensed data. Many vendors keep such information confidential, viewing training data as a competitive advantage. Revealing these details could also lead to intellectual property-related lawsuits, another reason for the lack of transparency.

To address concerns over potential copyright issues, AWS offers an indemnification policy that protects customers if a Titan model, such as Titan Image Generator v2, outputs a mirror copy of a copyrighted training example.

Despite some hesitancy from enterprises and the high costs associated with training, fine-tuning, and serving models, Amazon CEO Andy Jassy remains optimistic about the future of generative AI technology like AWS’ Titan models. In the company’s recent second-quarter earnings call, Jassy stated, “In the generative AI space, it’s going to get big fast, and it’s largely all going to be built from the get-go in the cloud.”