This inventor holds 1 USPTO granted patent and 2 published patent applications. Top assignee: Nvidia Corporation. Active years: 2026.
Data source: USPTO (United States Patent and Trademark Office) public patent records. Weekly synchronization. How IDiyas builds this profile
Company Filing History:
Years Active: 2026
Title: Yehonatan Kasten and Consistent Text-to-Image Generation.
Introduction
Yehonatan Kasten is an inventor based in Hinanit, Israel, and works at NVIDIA Corporation. His patent work addresses a challenge in generative artificial intelligence: creating images of the same subject across different scenes and text prompts.
Latest Patents
Kasten is associated with 1 patent, titled “Training-Free Consistent Text-to-Image Generation.” The invention uses a pretrained text-to-image diffusion model to generate images of a consistent subject in scenes described by different prompts. It shares intermediate model activations across image generation and incorporates subject-driven shared attention and correspondence-based feature injection to support consistency. The approach also encourages varied layouts while maintaining subject identity, and is designed to work with multiple subjects without requiring additional optimization.
Career Highlights
Kasten’s listed affiliation is NVIDIA Corporation. His patent reflects work at the intersection of generative models, image synthesis, and methods for improving consistency between generated images.
Collaborations
The patent information names Yoad Tewel and Rinon Pery Gal as Kasten’s coworkers.
Conclusion
Yehonatan Kasten’s patent focuses on generating diverse, text-aligned scenes while preserving the identity of depicted subjects, contributing to research in consistent text-to-image generation.
Data source: USPTO (United States Patent and Trademark Office) public patent records. Weekly synchronization. How IDiyas builds this profile