Amazon has unveiled an enhanced version of its proprietary image-generating model, Titan Image Generator, now available to AWS customers via the Bedrock generative AI platform.
This upgraded model, named Titan Image Generator v2, introduces several advanced features. As outlined by AWS principal developer advocate Channy Yun in a blog post, users can now "guide" the creation of images with reference images, make edits to existing visuals, remove backgrounds, and produce image variations.
"With Titan Image Generator v2, you can intelligently detect and segment multiple foreground objects," Yun elaborates. "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 Image Credits: Amazon The model supports image conditioning, allowing users to incorporate a reference image and emphasize specific visual characteristics such as edges, object outlines, and structural elements. It can also be fine-tuned using reference images like product logos or company branding, ensuring consistency in the generated images' aesthetic.
Titan Image Generator v2 Image Credits: Amazon However, AWS remains reticent about the specific data used to train its Titan Image Generator models. Previously, the company informed TechCrunch that it employs a mix of proprietary and licensed data, but did not provide further details.
Such secrecy is common among vendors, who view training data as a competitive asset. This lack of transparency also mitigates the risk of intellectual property disputes, which could arise if a model inadvertently reproduces a copyrighted training example.
To address these concerns, AWS offers an indemnification policy to protect customers if a Titan model, such as Titan Image Generator v2, unintentionally replicates a potentially copyrighted image.
