forge-film: video generator — what it does and what you need to run it

forge-film is an open source video generator that builds the filmmaking process as a dependency graph between scenes and runs independent scenes in parallel. The developer describes it as an engine that models script dependencies as a DAG and schedules execution by the critical path method, rather than simply generating scenes one after another.

What it does

As described by the author, forge-film combines several models in a single workflow and speeds up generation through parallel execution of independent scenes. The speed gain, as the developer claims, scales with the number of independent scenes. The project description mentions text-to-video generation, multimodal scripts and workflow automation. Among the repository tags are ai-cinematography, ai-filmmaking, text-to-video, llm-workflow and gradio, which points to a Gradio-based interface.

What you need to run it

The code is written in Python; running it requires Python 3.11 or newer, as well as ffmpeg. A GPU is optional, but for running CogVideoX locally the authors specify CUDA 12 or newer. The code license is MIT. The repository is available at github.com/F-R-L/forge-film.

Who it suits

The tool is aimed at those who assemble video from several scenes and want to cut total generation time through parallelism. It will suit users who are comfortable working with a Python environment and who are ready to figure out the model stack on their own. It will not suit those looking for a ready-made application with an installer or a fully cloud service with no local setup.

The project is young: the repository was created on March 24, 2026, and the last code change was on March 26, 2026. Keep in mind that at an early stage of development some capabilities may be described by the author but not yet confirmed by independent tests.