Google launches new image and video generative AIs: Imagen 3 and Veo

04 December 2024

Google’s Vertex AI platform welcomes two new AI models – Imagen 3 and Veo. They can work on their own, but can also be used together. The first can generate an image based on a text prompt, the second takes in an image and a text prompt and animates it.

Imagen 3 surpasses previous versions in “detail, lighting and artifact reduction”. It will be available to all Vertex AI users starting next week. Here are some sample images that were generated with the model alongside the prompts that were used.

Additional features are available for users on the “allowlist” (you can apply here). These include inpainting (adding elements to an existing image), outpainting (extending an image beyond its original borders) and product background (replaces the background automatically).

This browser does not support the video element.

Businesses can supply reference images to be used alongside the prompt. Google says that Imagen 3 Customization allows users to “infuse your own brand, style, logo, subject or product features when generating new images”.

With Veo, now available in private preview, users can take a static image and give stage directions to animate it. Here are a couple of examples.

This browser does not support the video element.

Additionally, Veo can generate videos completely from scratch, taking in just a text prompt.

This browser does not support the video element.

The output of both Imagen and Veo is invisibly watermarked with DeepMind’s SynthID, preventing AI-generated images from being passed off as real. Safety filters ensure that the AI can’t create harmful content. Also, Google has not used customer data to train its AI and offers copyright indemnity. Follow the Source link for more details.

Source