At the AWS re:Invent 2023 event, Amazon unveiled Titan Image Generator, which is currently in preview for AWS customers on Bedrock, Amazon's AI development platform. As a member of Amazon's Titan family of generative AI models, Titan Image Generator can generate new images based on a text description or customize existing images.
Amazon Titan models offer a variety of high-performing image, multimodal, and text model options through a fully managed API. These strong, all-purpose models were pre-trained by AWS on extensive datasets, enabling the safe use of AI while supporting a wide range of use cases.
Titan Image Generator offers a range of image editing capabilities, including automatically editing images based on a provided text prompt through a built-in segmentation model.
This model facilitates inpainting using an image mask and outpainting to either extend or modify the image background. Users can configure image dimensions and specify the desired number of generated image variations.
Moreover, the model allows for customization by incorporating proprietary data, ensuring the generated images align with specific brand guidelines or adopt a particular style.
For instance, users can fine-tune the model using images from a previous marketing campaign to achieve desired results. Notably, Titan Image Generator prioritizes responsible AI use by incorporating measures to mitigate the generation of harmful content.
You can use the model to seamlessly swap out backgrounds to generate lifestyle images, all while retaining the main subject of the image and to create a few more options, said Swami Sivasubramanian, VP for data and machine learning services at AWS.
Every picture created by Amazon Titan has an indiscernible watermark to improve accountability further. In an effort to lessen the spread of false information, this watermark acts as a covert method of identifying photos created using artificial intelligence. A watermark that cannot be removed has been created by other companies, such as Google DeepMind, that identifies AI-generated images with SynthID, a system that embeds a watermark inside the pixels of AI-produced images, which means that while a human eye can’t spot the difference, a detection tool can spot these tell-tale identifiers and be able to say an image was created with AI.