AI Cinema Workflow

January 8, 2026By Alan Kent · AI agent architect; building Ordinary Animator

Previously I talked about the overall workflow used by many animation studios, including Pixar, and how it still makes sense for AI generated content. I came across another AI creator talking about their workflow, and its pretty similar, so I thought I would drop a mention for reference as well.

First, my old post if you want to refer back to it.

The new video I came across is from The AI For Real Life Channel on YouTube. He mainly uses online services rather than ComfyUI, but he has an interesting perspective from someone doing paid work using these technologies.

Here is his workflow.

Diagram of an AI cinema production workflow with music brought in early and a narration pass near the end

What is different?

Music is brought in early - if creating an ad for example where timing and mood is important, selecting the music early is important as you may want to align the emotional beats of the music with the content.
Edit to Music is where in the video editing software you trim the video clips to align with the music.
The Narration Pass is towards the end I assume because he more often creates videos where there is a narrator rather than lip sync. You can move dialog around more freely in the timeline than music where the visual and music need to align to hit beats.

The rest is pretty similar.

Work out the idea first, confirm it.
Build up your visual assets - the characters (I would add locations and props if doing a series where you return to the same locations multiple times).
Break the full content down into scenes.
Then break the scenes into shots, where a 9-frame storyboard is a useful approach to generate a set of shot-starter images the look consistent.
Only generate the videos when you have everything else locked in (the start shots).

So nothing radical, just interesting to see another pipeline for an AI professional that is pretty much the same.

Here is the full video if you want to watch.