
For the video “A Feeling of Anger and Rebellion,” I had to remove part of the background. Due to limited space during filming, the video footage always included a portion of the room that was not supposed to appear in the final shot.
In general, AI-based background removal was already quite effective at the time of the experiment (as of March 2025), particularly for static video footage. In my case as well, the camera was stationary. However, a significant limitation arose due to the non-static light source, which caused the lighting conditions to change from frame to frame.
An initial approach involved enhancing the video using generative video AI, specifically Runway AI’s Inpaint tool, which was one of the most popular tools for video-based AI editing at the time. However, due to the changing lighting conditions, this approach did not produce consistent results, making it unsuitable for the project.
So I decided to do it manually, frame by frame, in Photoshop. To do this, I selected the area of the image I wanted to change and generated the new background using a prompt. Although this workflow worked in principle, it was extremely time-consuming.




Frame mit störendem Hintergrund.
Frame mit KI-Überarbeitung.

