Implement video production with artificial intelligence
Video producers quickly get bogged down by the many different disciplines they have to master before the finished film is ready for the audience: viewing and arranging footage, and recording a suitably timed audio track. Finally, subtitles want to accompany the work, also synchronously. When working through the long to-do list, the stressed filmmaker asks himself: Is there one for me among the many artificial intelligences?
To get straight to the point: AIs cannot currently produce continuous video content from pure text. Short scenes succeed – but you also have to spend a lot of time yourself to generate them. The AIs hire themselves out more as assistants: They help to sift through sample recordings (stock videography), simulate authentic speaker personalities and analyze spoken content to create synchronous subtitles.
In our experiment, we tried to shoot a suitable video clip for an already existing text and burden the AIs with as much work as possible. As a basis we chose the public domain classic “On Two Planets” by Kurd Laßwitz. The book, first published in 1897, is considered one of the first science fiction novels. It begins with three men crossing the Arctic in a high-tech hot-air balloon – perfect material for exciting film scenes. Our starting material consists only of a text file and a ten-minute audio book version of it, which we recorded in Audacity. This means that there is a defined timeline, i.e. the length of the project and the (acoustic) content distributed over it. In the end, a video should come out in landscape format (16:9) that fits the text.
