Amazon announced its own AI image generator, officially joining other big tech companies in the race to generative AI.
Read Also : Amazon Develops Workplace Assistant "Q"
Amazon Titan Image Generator
During the Amazon Web Services re: Invent 2023 conference, the company debuted its Titan Image Generator, allowing users to create or modify existing images through text descriptions.
According to Amazon, the new AI tool aims to help companies in advertising, e-commerce, and entertainment, by creating studio-quality yet realistic images in large volumes at a cheap price. The Titan Image Generator can produce multiple image options through textual prompts.
Moreover, the company boasted that since the model is trained on high-quality and diverse data, it can generate accurate and realistic outputs with minimal distortions. Users can easily change backgrounds by simply masking and outpainting part of the image.
Meanwhile, Amazon assured that its new tool supports the responsible use of AI. "All images generated by Amazon Titan contain an invisible watermark, designed to help reduce the spread of misinformation by providing a discreet mechanism to identify AI-generated images," the company announced.
Amazon Titan
The Titan Image Generator is part of Amazon's Titan models, which are equipped with 25 years of AI and machine learning innovation from the company. "AWS pre-trained these models on large datasets, making them powerful, general-purpose models built to support a variety of use cases while also supporting the responsible use of AI." the company wrote.
Aside from the AI image generator, the company also launched the Titan Multimodal Embeddings which allows users to submit text, images, or a combination of the two as input. Both Titan tools are exclusively available to Amazon Bedrock.
Currently, the Titan Multimodal Embeddings can convert images and short English texts up to 128 tokens and can handle images up to 25 MB. Meanwhile, the Titan Image Generator is open for preview and can be used for up to 77 tokens with a maximum input image size support of 25 MB.
