Automated Video Creator

Description

I built an automated long form video pipeline that turns scripts, images, and narration into publish ready videos. The goal was to make long form production repeatable without relying on manual editing for every output. I wanted a workflow that could take a concept and produce a consistent video structure, while still leaving room for review and improvements. The system is designed for builders who want predictable results rather than one off demos.

The pipeline starts with script generation using OpenAI GPT prompts that produce a scene outline. I then use TTS models to create narration with timing metadata. Images are selected or generated to match the script, and assets are staged in a predictable layout. FFmpeg handles final rendering, while a C# orchestration layer manages job state, caching, and retries. Redis backs the job queue and SQL stores run metadata so I can resume or audit a render later. Docker keeps the environment consistent across machines.

I normalized audio levels and basic timing metadata so the final render feels consistent across episodes. I also cache intermediate assets so reruns are faster when only the script changes. These small controls keep quality stable without adding heavy manual work.

I focused on orchestration and determinism. Each step writes intermediate outputs, which makes debugging easier and prevents redoing work when only one stage fails. Naming conventions and folder structures are consistent so the pipeline can find what it needs without brittle logic. I also added sanity checks for missing assets and timing alignment so narration and visuals stay in sync. This kept failures visible and reduced wasted compute on bad renders. I also keep a small config file for pacing rules and default transitions. It keeps renders consistent and makes review faster.

The result is a repeatable workflow that produces long form videos quickly and predictably. It reduces manual editing time while keeping a clear structure that can be reviewed before publishing. The pipeline also provides a strong base for enhancements like template variations, different voice options, or richer visual styles. This project showed me how careful orchestration and simple checks can make complex media production reliable for a solo builder.