US20250157109
2025-05-15
Physics
G06T13/00
The application outlines a method for generating real-time interactive videos using advanced machine learning models. Initially, a user's facial image is captured and processed by a first machine learning model to create a source image. This source image, along with additional facial images depicting various expressions, is then used by a second machine learning model to dynamically generate interactive video content.
Machine learning is widely used in content creation across industries, yet current methods often lack interactivity and fail to produce high-quality real-time video. The disclosed techniques aim to enhance content generation by enabling dynamic interaction with user expressions, thereby improving the quality and engagement of video content.
The system comprises two main components: a first machine learning model that captures and processes the user's facial features to generate a source image, and a second model that uses this source image to create interactive videos. The second model applies the user's real-time facial expressions to the source image, resulting in an animated character or avatar that mirrors the user's expressions.
Users interact with the system through a device interface that guides them in capturing their facial expressions. These expressions are continuously recorded and processed to update the source image in real time. The second machine learning model uses this data to animate the character or avatar, ensuring that it reflects the user's current expressions in the video output.
The system is designed for ease of use on devices such as mobile phones or tablets. Users are prompted through an interface to position their face correctly for optimal capture. The system then processes these inputs to generate real-time interactive videos where users can see their expressions animated on a digital character, enhancing user engagement and providing a dynamic content creation experience.