Technology

Google Research launched LUMIERE, a new AI model that creates videos from text and prototype images.

2024-01-26 02:23:10

Last Tuesday, Thai time Google Research launched Lumiere, an AI model with the ability to create videos from text prompts and prototype images, it is another AI model developed by Google after the launch of Gemini late last year.

This AI model can create videos that are consistent, relatively smooth, and realistic in some videos, which Google Research has been an example of the processing of this model for viewing. This shows the ability of AI to create animations from text prompts. and create a video that can be created at many angles Another ability is to create a video from an image by entering the desired prompt. Another interesting ability is to create a video from just one reference image. The resulting video will be in the same style as the reference image and will Be movement that is appropriate for the image slide.

In addition to this basic capability, it is also possible to synthesize videos that display realistic, diverse, and consistent movements. This is a major challenge in video synthesis. For this reason, we introduce a Space-Time U-Net architecture that generates the entire temporal duration of a video at once. One pass through the model This is in contrast to existing video models that synthesize remote keyframes followed by temporally ultra-high resolution. such as commercially available text image editing for consistent video editing. Animate the content of the image. Adding missing video and adjusting elements within the video

Currently, the status of this AI model is still only in the early stages of research. The developer is not yet ready to be able to try it out due to some limitations, causing improvements and corrections to be made before it can be used in the future.