A Comparative Study on the Practicality of AI Video Generation Tools in Short Video Creation

Abstract

With the rapid advancement of artificial intelligence, AI-generated content (AIGC) is experiencing explosive growth on short-form video platforms. It is increasingly permeating diverse fields, including lifestyle entertainment, gaming, and anime. Currently, the mainstream AI video-generation technologies primarily include three paradigms: text-to-video (T2V), image-to-video (I2V), and video-to-video (V2V). These three types of technologies differ significantly in their generation mechanisms, underlying prin ciples, and output quality. Empirical studies reveal that the quality of AI-generated videos exhibits considerable heterogeneity. Some works achieve a remarkable level of realism, closely resembling real-world scenarios, while other low- quality content dis plays obvious flaws that make it easily identifiable. In-depth analysis shows that T2V technology excels in following user instructions and generating high-quality videos, while V2V technology also delivers impressive visual transformation results. By contrast, I2V technology is constrained by the limited information in the source images and insufficient guidance from prompts, which often makes it difficult to produce satisfactory dynamic videos. Moreover, influenced by the algorithmic recommendation mechan isms of short-video platforms, different platforms adopt varying strategies for distributing AI-generated content. Based on these observations, this paper aims to explore how to optimize the AI video creation process and, in conjunction with platform rules, develop effective publishing strategies to enhance the dissemination and impact of AI-generated content within the short-video ecosystem.