IMAGE GENERATION METHOD USING GENERATIVE MODEL, AND COMPUTING DEVICE FOR PERFORMING SAME

Smart overview of the Invention

The described method leverages a generative model to create images based on dialogues. It involves processing a dialogue through a prompt generative model to generate a contextual prompt, which is then used to produce an image. This approach aims to address the mismatch often found between dialogues and the images generated from them by ensuring the image reflects the dialogue's context accurately.

Generative AI Technology

Generative AI technology underpins the method, capable of learning from extensive training data to produce new data similar to the input. This technology is particularly useful for generating images from text, a feature increasingly relevant due to the rise in online communication. The challenge addressed is the inconvenience of manually inputting image descriptions to ensure accurate image generation from dialogues.

Method and Device Implementation

The method involves a computing device equipped with a memory and processor to execute the image generation process. The device receives a dialogue, processes it through the prompt generative model to derive a prompt, and then generates an image based on this prompt. The prompt generative model is trained with numerous dialogue-prompt pairs, ensuring it captures the dialogue's context effectively.

Technical Details

Technical descriptions are streamlined to focus on the core innovation without unnecessary complexity. The method's implementation involves software and hardware components, potentially utilizing field programmable gate arrays (FPGAs) or application-specific integrated circuits (ASICs). These components execute instructions stored in memory, directing processors to perform the image generation functions.

Definitions and Models

Key terms are defined to clarify their use in the context of the method. 'Generative AI' refers to AI capable of generating new content, while a 'generative model' is a neural network that learns data patterns. The 'prompt generative model' is a specific type of language model that transforms dialogue into prompts, ensuring the generated image accurately reflects the dialogue's context.