AI FACE DECORATION TEXTURE GENERATION IN SOCIAL MEDIA PLATFORM

Smart overview of the Invention

The patent application describes a computing system designed to enhance social media platforms by allowing users to generate face decoration textures using artificial intelligence (AI). The system can process a base image of a human face, an image mask, and a user text prompt to produce customized effects. The image mask specifies areas where inpainting should occur, excluding sensitive regions like the eyes. This innovation aims to simplify the creation of personalized visual effects for users, leveraging AI technology to reduce the need for manual editing.

Background

Social media platforms have long offered tools for adding effects to images and videos, particularly those involving human faces. These effects often include filters and textures that modify or augment real-world objects. Traditionally, creating such effects required manual input, limiting accessibility to advanced users. The rise of AI has introduced new possibilities for image generation, but previous attempts have still needed manual adjustments. This invention seeks to overcome these limitations by enabling users of all skill levels to create sophisticated visual effects with minimal effort.

Technical Specifications

The described system includes a server device equipped with processors that execute stored instructions to manage the social media platform. It features modules for generating face decoration textures and storing them in a library for user access. Users interact with the platform through a client device, which could be a smartphone or computer. The client device displays a graphical user interface (GUI), allowing users to input their base image and text prompts. The server processes these inputs using AI models to generate the desired texture effects.

AI Model Functionality

The AI model utilized in this system is based on diffusion models, such as STABLE DIFFUSION, which are trained to generate images from input data. An image encoder processes the base image and mask, extracting visual features and creating embeddings. A text encoder then combines these embeddings with user prompts to form an input feature vector. This vector is used by the diffusion module to synthesize the final face decoration texture. The system is designed to ensure accurate placement of textures on faces by using the eyes as anchor points.

Practical Applications

Various examples illustrate how the image mask guides texture generation. Masks can define areas like eyes and mouth where inpainting should not occur, allowing for realistic overlays such as makeup or masks that integrate seamlessly with facial features. By adjusting these masks, users can create diverse effects ranging from Halloween masks to everyday makeup applications without obscuring essential facial elements. This flexibility highlights the potential for broad customization within social media environments, enhancing user creativity and engagement.