Abstract
Popular Micro-videos, dominant on platforms like TikTok and YouTube, holdsignificant commercial value. The rise of high-quality AI-generated content hasspurred interest in AI-driven micro-video creation. However, despite theadvanced capabilities of large language models (LLMs) like ChatGPT and DeepSeekin text generation and reasoning, their potential to assist the creation ofpopular micro-videos remains largely unexplored. In this paper, we conduct an empirical study on LLM-assisted popularmicro-video generation (LLMPopcorn). Specifically, we investigate the followingresearch questions: (i) How can LLMs be effectively utilized to assist popularmicro-video generation? (ii) To what extent can prompt-based enhancementsoptimize the LLM-generated content for higher popularity? (iii) How well dovarious LLMs and video generators perform in the popular micro-video generationtask? By exploring these questions, we show that advanced LLMs like DeepSeek-V3enable micro-video generation to achieve popularity comparable to human-createdcontent. Prompt enhancements further boost popularity, and benchmarkinghighlights DeepSeek-V3 and DeepSeek-R1 among LLMs, while LTX-Video andHunyuanVideo lead in video generation. This pioneering work advancesAI-assisted micro-video creation, uncovering new research opportunities. Wewill release the code and datasets to support future studies.