Pre-Publication V2. A deep learning technique that is intended to automatically replaces one person's face with another's in a video or image. It uses a neural network (like a GAN or a Diffusion Model) to analyze thousands of facial landmarks from a source image, mapping the new face's identity onto the original subject while preserving natural movements, expressions, and lighting.
Deliberation Summary:
(1) Note that the term names the outcome of a task that can be accomplished either through an AI workflow or with conventional photo editing software and thus the technical detail may not be necessary, since the underlying concept will outlast any particular technology.(2) Consider edits as the description of the preserved result is subjective.
Pre-Publication V1. A deep learning technique that automatically replaces one person's face with another's in a video or image. It uses a neural network (like a GAN or a Diffusion Model) to analyze thousands of facial landmarks from a source image, mapping the new face's identity onto the original subject while preserving natural movements, expressions, and lighting.