machine for film

machine for film is an experimental AI project focusing on the automation of film and video production.

The project is founded on the idea that machines - which we define as systems that combine both AI and traditional programming - will ultimately be better than humans at using other machines (e.g. generative AI models and systems) at generating films. This moves away from predominantly prompt-based approaches that require human input and iteration.

We are interested in developing machines that carry out end-to-end video production (e.g. automatically generating entire films from single inputs) as well as more specific and / or experimental aspects of video generation. These include:

• Automating specific stages of video production (e.g. generating full scripts from agentic AI interactions, generating stop motion videos from rough sketches etc.)
• Generating novel styles of video using AI systems (e.g. continuously autogenerating video feeds, videos that evolve in response to real-time live data inputs etc.)

The key unifying feature is full automation, or human-out-the-loop. We're interested in creating systems that work without human involvement for a specific task, including both broad and more narrowly defined aspects of filmmaking.

We also care a lot about creating high quality outputs that are genuinely interesting and compelling to watch. We are filmmakers ourselves, and this is an important part of the project - it's not just about technical experimentation, but also creating systems that generate videos that are in themselves interesting and watchable.

We are posting (almost) daily videos on our X feed with the outputs from our machines, and will continue to update this website as we develop new machines for film.

machine for film is a collaboration between Richard Flint and George Kenwright, with support from Jordan De Souza.