After Q AI, Amazon announces AI-powered image generator tool Titan

Following the revelation of their latest Q AI chatbot, Amazon is now making significant strides into the realm of image generation AI. The announcement comes as part of the extensive showcase at the AWS re:Invent conference wherein the company announced its first image generation AI model dubbed Titan.

Unveiled during the keynote by Swami Sivasubramanian, AWS Vice President of Database, Analytics, and Machine Learning, the Titan Image Generator is positioned as a tool for developers to create their own image generators using the model, requiring access to Amazon Bedrock.

It offers a streamlined and efficient way to create and refine images through natural language prompts in English or by uploading an image to edit. Targeting industries such as advertising, e-commerce, and media and entertainment, this model will help create “studio-quality, realistic images” in large volumes and at a low cost, says company.
Key features of Titan Image Generator

The new AI image generator Titan is not a standalone app. It is rather a tool that developers can use to create their own image generators.

It has the ability to iterate on image concepts by producing multiple options based on text descriptions. The model is capable of understanding prompts involving multiple objects, delivering relevant and accurate images. It is trained on diverse and high-quality data.

The image editing capabilities include automatic editing with text prompts through a built-in segmentation model. The model supports various editing functions, such as inpainting with an image mask and outpainting to modify or extend the background. Users can also customise image dimensions and specify the desired number of image variations to be generated.

Furthermore, the Titan Image Generator allows for customisation with proprietary data, enabling users to align generated images with brand guidelines or achieve a specific style.

Amazon informs that all images created by Titan will have an invisible watermark. This watermark serves as a discreet mechanism to identify AI-generated images and mitigate the potential spread of misinformation.