Conversational video editing method, device and equipment capable of customizing story drama

The invention discloses a dialogue type video editing method, device and equipment capable of customizing story drama, which introduces a multi-modal large language model to carry out deep video understanding, and maps movie content into text description. Specifically, a cue word template is designed for the understanding of video content to guide the MLLM to understand character interaction, relationship and story plots and generate detailed script description. Secondly, a video creator can have a dialogue with the large language model to generate a target script; a cue word template is designed for generation of a target script, and a target script about story content and a timestamp is generated according to a customized story plot input by a creator and the corresponding timestamp. Through multiple rounds of conversations, the target script can be refined. And finally, combining the edited video according to the timestamps in the target script. According to the method, video content understanding and edit