[2O4-GS-7] Vision, speech media processing

[2O4-GS-7-02] Automatic converter to portrait video for news videos

〇Nanoka Mizukado¹, Takuto Yamauchi¹, Kenji Tei¹ (1. Waseda University)

Keywords:video processing

As video watching with smartphones becomes more common, vertical videos are becoming more popular on social networking service such as YouTube and TikTok. Since people use smartphone in vertical basically, vertical videos can be more appealing than horizontal videos, depending on your target market. In this paper, we focus on news videos that require quick and reliable information, and propose a tool that automatically converts news videos to vertical (9:16). This converter consists of three steps. First, we identify the area of most interest of the video and crop it. Next, it acquires textual information from the cropped content. Finally, the textual information is re-layout of the vertical video. As a result, we cropped video with minimizing the loss of semantic important content (textual information) . In this paper, we experimented with inputting 20 actual broadcasted news videos into the proposed tool to confirm its sufficient usability.