Scene one video production
The first video serves as a trailer for our entire performance, succinctly showcasing the content of the show within two minutes to heighten the audience’s anticipation. This video will be displayed on a TV screen at the front end of the corridor passage and will also be projected on the left side of the passage using a shower curtain as the screen. The shower curtain is thin, lightweight, and translucent, making it ideal for depicting underwater scenes.
I created the first scene of the video using AI. The overall process involved initially generating images through ChatGPT and then using these images to create videos in Runway. When generating images with ChatGPT, it is crucial to precisely describe the needed content and style to produce effective images. Based on our theme, I needed several images related to fish and dragon gates in different styles. For instance, describing the color and texture of carp, such as golden scales shimmering in the sunlight; depicting their vigorous leaping motion with splashing water; key environmental descriptions like turbulent rivers and distant dragon gates; and capturing an atmosphere filled with challenge and hope. These detailed descriptions help AI more accurately understand and achieve the visual style I want.
After confirming the images, I imported them into Runway and used the Gen-2 model to generate videos. A single image can develop into a video clip of about four seconds, which can then be extended. Tests show that to maintain the original content and style, a video generated from one image can be extended up to 16 seconds at most. Subsequently, I used Runway’s animation tools in the video editing interface, such as panning, zooming, and rotating, along with camera-related settings, to manually adjust the video montage based on the generated video.
