MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641025600 A) filed by Oligen Studios Private Limited on March 04, 2026, for System And Method For Generating Multilingual And Emotion-Preserving Dubbed Speech Through A Unified Architecture.
Inventors include Goutham Ravisankar; Sindhuja Kandamaran; Dhinesh Kumar R; Gopikrishnan Janakaraja; and Ramkumar Rajendraraja.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: SYSTEM AND METHOD FOR GENERATING MULTILINGUAL AND EMOTION-PRESERVING DUBBED SPEECH THROUGH A UNIFIED ARCHITECTURE An embodiment herein provides a system (100) for generating multilingual and emotion-preserving dubbed speech through a unified architecture configured to prevent cascading errors. The system includes (i) a preprocessing unit (104) configured to transform a multimodal input data stream into normalised multimodal input data segments within a shared emotional-prosodic and linguistic latent feature space, (ii) a dual-path adaptive speech transformation subsystem (108) configured to use the normalised multimodal input data segments to perform multilingual neural voice generation using the normalised multimodal input data segments, including (a) performing voice conversion, and (b) performing text-to-speech synthesis, (iii) a synchronisation unit (116) configured to align generated speech from the dual-path adaptive speech transformation subsystem with visual content from the video processing unit, and (iv) an output unit (118) configured to generate a synchronized audio-video output with embedded neural spectral watermarks implanted into mel-spectrogram regions for identity governance. FIG. 1
Disclaimer: Curated by HT Syndication.
