<![CDATA[This research adopts a Practice-Led Research method by involving twenty-eight students to record fragments of urban reality in Yogyakarta from their respective perspectives. The heterogeneous collection of videos is briefly described to map the diversity of visual representations, and subsequently examined through a documentary film entitled “Yogyakarta Hari Ini.” Specifically, this study examines how ChatGPT was employed as a narrative structuring tool: textual descriptions of the 28 footage clips were submitted to the model in randomized order, and the AI-generated sequence recommendations were subsequently evaluated and refined by the director during the editing phase. The findings demonstrate that AI effectively accelerated the categorization and narrative structuring of heterogeneous footage, while also revealing clear limitations in aesthetic judgment, atmospheric continuity, and socio-cultural sensitivity or areas that required consistent human intervention and correction. The study confirms that AI can serve as a creative tool in transforming raw visual data into a coherent narrative, functioning as a “dramaturgical assistant” during the directing phase. However, the director’s aesthetic judgment remained indispensable: approximately one-third of the AI-generated sequence recommendations required revision due to mechanical transitions and disregard for atmospheric continuity. This human-machine collaborative model contributes to the discourse on Indonesian film aesthetics in the digital landscape, offering a replicable prototype for documentary practice confronting heterogeneous visual archives.]]>
