ABSTRACT
The capabilities of neural networks such as diffusion models, generative adversarial networks (GANs), and transformers to produce quality image and video content have the potential to disrupt established cinematographic practices and workflows. The traditional understanding of cinematography may need to expand beyond physical production to include image content curated through prompting. By examining how neural networks emulate both the creative and technical functions associated with traditional cameras, how curatorial text prompting enables greater control over cinematographic parameters, and how the current limitations of image/video generation pose practical obstacles to their widespread adoption, a clearer understanding emerges of where these tools are today and where they need to develop.
