MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641111085 A) filed by Rajalakshmi Engineering College on September 16, 2026, for System And Method For Branched Multi-Agent Generative Advertisement Synthesis With Deterministic Layered Typography And Localized Non-Destructive Image Refinement.
Inventors include Dr. N. Duraimurugan; Lithan G; and Logesh D.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: ABSTRACT The present invention discloses a system and method for branched multi-agent generative advertisement synthesis with deterministic layered typography and localized non-destructive image refinement. The system receives a natural-language advertisement requirement and optionally a product or reference image, and performs semantic analysis to identify visual attributes, advertising-copy requirements, spatial constraints, negative constraints, and other generation pallaTeters. A structured representation of the analyzed requirement is generated and distributed through a branched orchestration mechanism to separate viSual-generation and copy- generation pathways. A visual-generation agent generates a raster visual background using a generative image model with text-suppression constraints, while a copygeneration agent independently generates advertising text as structured data. A deterministic composition engine combines the generated visual background with the independently generated text as editable information layers, enabling controlled modification of typography, text position, size, alignment, and layer ordering without regenerating the complete visual background. The system further provides a localized refinement module configured to receive a spatial mask and perform generative recomputation within a selected region while preserving designated unmasked regions and separately maintained information layers. Optional image-to- image conditioning may be employed for product or reference-image preservation. The system may additionally perform model selection, image resizing, aspectratio conversion, resampling, resolution adjustment, and output-format conversion. The disclosed architecture enables controlled generation, deterministic composition, editable layered content, and localized visual refinement while reducing unnecessary complete-image regeneration during iterative advertisement creation and modification.
Disclaimer: Curated by HT Syndication.
