Customizable AI-supported text-based visual content creation for increasing user’s visual-creation creativity
Co-creation with AI is trending and AI-generated images from textual descriptions has shown advanced and attractive capabilities. However, commonly trained machine-learning models or built AI-based systems may have deficient points to generate satisfied results for personal usage or novice users on AI co-creation drawing, which may be because of deficient understanding of personal textual expressions or low customization of trained text-to-image machine learning models.
Therefore, we aim to assist users in creating personalized visual contents from their own descriptions. First, we develop advanced text-to-image machine-learning models, especially capable of efficient re-training models. Second, we collected and revised datasets containing more diversity in textual or visual expressions. Third, we design flexible interfaces in creating user-intended AI-generated visual results. Note that we especially design functions for novice users who are not familiar with creating visual contents and focus on more AI co-creation visual results of Japan.
Jun-Li Lu, Naruya Kondo, Hiroyuki Osone, Yoichi Ochiai