Amazon launches its text-to-image generative AI, Titan

All images created with Titan will include what Amazon has called invisible watermarks, to certify the non-human origin of the works

Amazon is entering the world of AI image generation with the release of its Titan model based on text-to-image technology. Presented at the AWS re:Invent conference, the Titan image generator is capable of creating “studio quality” images and stands out for its security measures. Titan is not an independent app or website, but a tool that programmers can use to develop their own image generators based on this model: To use it, you need to log in to Amazon Bedrock. Swami Sivasubramanian, AWS vice president of database, analytics and machine learning, previewed the Titan Image Generator during his keynote speech, highlighting the model’s ability to do more than just create an image from a natural language prompt , but also to change the backgrounds. This tool is mainly aimed at a corporate audienceunlike more consumer-oriented image generators, such as OpenAI’s DALL-E.

All images created with Titan will include what Amazon has called invisible watermarks. Vasi Philomin, vice president for generative AI at AWS, said this was part of the voluntary commitments Amazon signed with the White House in July. “We wanted a way to mark an image as being created with AI, and specifically with the Titan Image Generator, that did not interfere with the visual, without latency, and that could not be cropped or compressed,” Philomin said in a ‘interview. He added that the watermark is not confined to the file’s metadata. However, invisible watermark detection presents some challenges. Amazon has created an API that people can connect to and then submit the image to verify its provenance. Philomin explained that this was intentionally designed: after all, Titan is not a finished product but a model, so programmers working with the image generator will choose how to provide this information to users.

Sivasubramanian also announced the general availability of other Titan models: Titan Text Lite, a smaller model used for lighter text generation tasks such as writing advertising copy, and Text Express, which is intended for broader uses such as apps of conversational chats. Amazon will also extend copyright relief to customers who use its core Titan templates, including text-to-image templates. It will also provide legal cover to people who use any AI applications created by Amazon, even if the app used a different fundamental model found in its Bedrock AI model repository, such as Meta’s Llama 2 or Anthropic’s Claude 2. These apps include AWS HealthScribe, CodeWhisperer, Amazon Personalize, Amazon Lex, and Amazon Connect Contact Lens.