Amazon Introduces Its Own AI Image Generator, And It Can Edit Pics Too

Image via Amazon

Amazon is hopping aboard the artificial intelligence image generator train, unveiling its own tool that transforms text prompts into striking visuals.

The all-new Titan Image Generator turns English language text prompts into detailed images, and is said to be handy asset for those in advertising, e-commerce, and entertainment, enabling them to whip up “studio-quality” images quickly and without breaking the bank.

Video screenshot via Amazon Web Services

Prompt: “green iguana.” Image via Amazon

Prompts: “iguana as a cartoon” and “iguana as a sketch.” Image via Amazon

What makes this tool stand out is its knack for handling tricky prompts. Whether it’s juggling multiple objects or zeroing in on specific attributes, Amazon says the generator can churn out results that are both spot-on and lifelike. It’s also equipped to handle some smart editing tricks like filling in parts of an image (inpainting) or tweaking the background (outpainting).

Automatic editing input: Image + “iguana in a rain forest” as a prompt. Image via Amazon

Inpainting input: Image + a mask area of the iguana + “toucan” as a prompt. Image via Amazon

Outpainting input: Image + a mask area of the background + “beach” as a prompt. Image via Amazon

For those who are particular about visual attributes, the tool offers options to customize image dimensions, the number of variations, and even the ability to feed in proprietary data to keep things in line with specific brand styles.

As responsible AI use becomes increasingly crucial, Amazon notes that it has included safeguards to prevent misuse, like measures to prevent harmful content generation. Plus, every image comes with an invisible watermark, subtly identifying AI-generated images and combat misinformation.

Images via Amazon

Part of Amazon’s broader AI and machine learning initiatives, the Titan Image Generator is accessible through Amazon Bedrock. This platform aims to serve as a one-stop shop for a variety of top-tier image, multimodal, and text model options.

[via Engadget and TechCrunch, images via Amazon]