Classical continuity without explicit guidance? An exploratory comparison of minimal and guided prompts in Kling
Year: 2026
Type: article
Abstract: Classical continuity consists of the spatial, temporal, and causal relationships that contribute to the comprehensibility of a scene. Shots are connected through these relationships, allowing viewers to follow actions even as viewpoints change within classical narrative cinema. This exploratory study aims to identify classical continuity relationships that are maintained or disrupted in multi-shot videos generated with Kling and to compare outputs produced using minimal and guided prompts. While previous research on AI-generated video has generally evaluated aspects such as visual quality, temporal consistency, and prompt adherence, this study focuses on continuity relationships between shots. To this end, six micro-scenes were developed, and each was generated once per prompt condition using Kling Video 3.0 Omni. The two conditions used a minimal prompt and a guided prompt that more explicitly specified continuity relationships. The Multi-Shot feature was enabled in both conditions, resulting in twelve ten-second videos. Descriptive coding and paired formal analysis examined shot structure, spatial continuity, gaze, action, visual identity, and narrative continuity. An independent second coder achieved exact agreement on 172 of 192 coding decisions. The findings indicate that many classical continuity relationships were maintained in videos generated with minimal prompts. Although outputs produced with guided prompts contained a higher total number of shots, no consistent pattern of improvement in continuity relationships was observed. Some guided outputs also showed disruptions in specific continuity dimensions. Without repeated generations, the observed differences cannot be directly attributed to prompt conditions. The findings suggest that shot count and continuity relationships between shots should be considered separately when evaluating AI-generated videos
Author: Seda Aktaş
Institution: İstanbul Esenyurt Üniversitesi
Language: English
Topic: Digital Media and Visual Art
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7237353201
