(CTN News) – During the AWS re: Invent conference, Amazon announced the release of its Titan text-to-image AI model, which is their entry into the AI image generation market.
Titan Image Generator is designed to create “realistic, studio-quality images” and has built-in guardrails to prevent toxicity and bias. It is not a standalone app or website, but rather a tool that developers can use to create their own image generators powered by the model.
Developers need access to Amazon Bedrock in order to utilize it.
During his keynote, Swami Sivasubramanian, AWS vice president of database, analytics, and machine learning, demonstrated Titan Image Generator’s ability to not only create an image from a natural language prompt but also change out backgrounds.
The tool is aimed at an enterprise audience, rather than the more consumer-oriented focus of existing image generators like OpenAI’s DALL-E.
All images generated by Titan Image Generator will include invisible watermarks, as part of the voluntary commitments Amazon made with the White House in July. Vasi Philomin, vice president for generative AI at AWS, confirmed this to The Verge.
During an interview, Philomin emphasized the need for an AI-generated image watermark that is visually seamless, instant, and resistant to manipulation.
He clarified that this watermark should not be limited to metadata. Detecting this invisible watermark is a challenge, but Amazon has developed an API for users to verify the origin of an image.
Philomin explained that this approach allows developers to decide how to present this information to their users, as the Titan Image Generator is a model, not a final product.
Watermarking and identifying AI-generated content are important aspects of the Biden administration’s executive order on AI. To address this, companies like Microsoft and Adobe have adopted the Content Credentials system developed by C2PA. Adobe even incorporates an icon to indicate content credentials in images and videos.
Sivasubramanian has additionally declared the availability of other Titan models, namely Titan Text Lite and Text Express. Titan Text Lite is a smaller model that is suitable for lighter text generation tasks such as copywriting, while Text Express is designed for larger uses like conversational chat apps.
Furthermore, Amazon will extend copyright indemnity to customers who use its Titan foundation models, including text-to-image.
Additionally, Amazon will provide legal cover to individuals who use any Amazon-created AI application, even if the app used a different foundation model found in its Bedrock AI model repository, such as Meta’s Llama 2 or Anthropic’s Claude 2.
The apps include AWS HealthScribe, CodeWhisperer, Amazon Personalize, Amazon Lex, and Amazon Connect Contact Lens.