Intellectual Property India Publishes Patent Application for Text Driven Cross Modal Semantic Alignment For Few Shot Gan Latent Space In Creative Ai Systems

MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641053837 A) filed by The Principal, Mepco Schlenk Engineering College on April 28, 2026, for Text Driven Cross Modal Semantic Alignment For Few Shot Gan Latent Space In Creative Ai Systems.

Inventors include Dr V S Benitha; J. B. Kevin Jonathan; and S. Praveen.

The application for the patent was published on June 26, 2026, under issue no. 26/2026.

Abstract: A system and method for text-guided latent space editing using disentangled direction learning (FEditNet++) are disclosed. In some embodiments, an input image is transformed into a latent representation using a latent inversion module (103). A textual prompt is encoded using a text encoder (201) to generate a semantic embedding (202) representing the desired modification. The system employs FEditNet++ (301) to learn disentangled latent directions and modify the latent code accordingly, enabling targeted attribute manipulation while preserving non-target features. The modified latent representation is processed by a StyleGAN generator (105) to produce an edited output image. A feedback mechanism refines the latent modification to improve semantic alignment with the textual input. The approach improves controllability and reduces unintended attribute coupling. The system is applicable to image editing, digital content creation, virtual character design, and creative artificial intelligence applications.C

Disclaimer: Curated by HT Syndication.