Google has been quite vocal about the changes it is bringing across its verticals using generative AI. The company is taking measures to introduce the tech across its products to stay relevant in the AI dynamics and compete with players like Microsoft and OpenAI. Among the latest interesting additions to the company’s AI catalogue is Lumiere for video generation and Art Selfie 2 for generative AI tweaks for selfies. Let’s explore more about these two tools.
Google Lumiere AI
According to the research note, the text-to-video diffusion model utilises a diffusion architecture – Space-Time-U-Net (or STUNet) to synthesise videos to portray realistic and coherent motion. The model creates 5-second videos in a single process rather than piecing together frames.
It can locate things in a frame and learn how they move or change their positions over time. Its design can facilitate a range of tasks, such as video generation using images and text, video inpainting, stylised generation and more. “The Lumiere model is able to animate the content of an image within a specific user-provided region,” the Google Research website noted.
Lumiere can also edit videos based on text-based prompts among other things. It can output 1024 × 1024p 5-second videos and generate 80 frames (compared to 25) via Stable Video Diffusion. The website noted the risk associated with “creating fake or harmful content” and it will “develop and apply tools for detecting biases and malicious use cases.”
Art Selfie 2
The revamped Art Selfie tool allows users to click a selfie, choose a style and wait for generative AI to process the output in the stylised scenario. “See how you’d look as an explorer scaling Everest, painted as a muse in Monet’s garden, or sporting armour as a courageous medieval knight,” Google said.
It will have 25 styles, and more shall be added soon. While the image is being generated, the tool will “deliver topical facts” to learn more about the context. Users can download the ‘Google Arts & Culture’ app on Android or iOS and find Art Selfie 2 within the “Play” tab to start experimenting with selfies and learn about world history and culture.
Moreover, Google also expanded ‘Art Remix’ to more countries. “This feature puts your AI prompting to the test, presenting artworks alongside a prompt that describes them, challenging you to edit the prompt and imagine the work with your own twist,” the company said. It will also be available in the Play Tab of Google Arts and Culture.
