Picture this. You’re on stage at a conference, running a live demo, and the clock beats you. You’ve got a complete brief and a stack of written prompts, but your time’s up. Most of us would close the laptop, grab a coffee, and move on.
I just came across a LinkedIn post from an AI video expert who hit that exact wall, and I love what happened next. During an AI video masterclass at the Content Marketing Summit in Seoul, the author ran a live demo on how to start an AI drama from scratch. Time ran short, so the demo ended at the brief and prompts. But over the weekend, the author kept going. That side project turned into a 17-minute AI film. Even the creator says, “I never thought this was possible.”
A 3-minute plan that got out of hand
The original plan was modest: a 3-minute film. It’s a sensible length for a proof of concept, and nobody would’ve blamed the creator for stopping there. But curiosity got in the way, and one question changed the whole project: “What if we spend more time telling the full story?”
The source material is Journey to the West, the 16th-century novel. The brief covers just 3 chapters, which is about 3% of the book. Do some quick napkin math, and at that pace the full novel would run for more than 9 hours. That’s my calculation, not the author’s, but it shows how much room there is when you stop rushing a story.
There’s also a detail here that I think matters more than any tool in the stack. The original poster has read this novel at least 5 times, at different stages of life. So this wasn’t someone asking AI to invent a story from nothing. It was someone who already knew these characters and this world, using AI to bring them to the screen.
The budget problem and a surprising answer
A longer film creates an obvious problem. More minutes usually means more generation, more money, and more time. A 17-minute film could easily become a very expensive weekend, so the workflow had to stay lean. It came together in three stages:
- Build the foundation: The author rebuilt the character sheets and refined the storyboard using a set of Claude skills and GPT Image 2.5.
- Generate the footage: Video generation ran on Dreamina and took 2 hours at most, using Seedance 2.5 at 480p, the cheapest option available.
- Shape the film: The edit happened in Claude Code, which also took the film from 480p to 1080p without an additional AI video model.
If you’re new to this, a character sheet is a visual reference for how a character should look, while a storyboard maps out the film shot by shot. Think of them as the blueprint. Long AI videos need that kind of planning because the audience will notice when a story or a character starts to drift.
Now look at that second step again. Choosing 480p for a film sounds like a mistake. It’s a low resolution, and most people would chase the best quality their budget could buy. But the creator picked it on purpose, both to save money and to treat the project as a proof of concept. That choice sets up the most interesting part of the story.
The plot twist: the edit did the heavy lifting
You’d expect the video model to be the star here. It usually gets all the attention. But generation took 2 hours at most, and the author says plainly, “The heavy lifting was the edit in Claude Code.”
Most of the time on this project went into working with Claude Code to direct the storyline, the cuts, and the edits. So a tool many of us think of as a coding assistant ended up in the editor’s chair. And it did this while bringing the final film up to 1080p, with no extra AI video model involved. I was honestly surprised by that part, because a resolution jump like this is usually where people reach for another paid tool.
Generation gives you raw material. The story gets made in the edit. That’s how traditional filmmaking has always worked, and this project suggests AI filmmaking may be heading the same way.
The result, according to this talented creator, is the longest AI drama they’ve produced in such a short time.
What you can take from this
You don’t need to adapt a 16th-century classic to use these lessons. Here’s what I’d borrow from this workflow for your next video project:
- Start with a story you know cold: Pick material you can judge with confidence, whether that’s a favorite book, a customer story, or your company’s origin story.
- Write the brief first: The whole film grew from a complete brief and a set of written prompts, so get those right before you generate anything.
- Prototype cheap: Test your idea with the lowest-cost option, then decide where quality deserves more spending.
- Budget time for the edit: Keep generation short and give most of your hours to the storyline, the cuts, and the pacing.
- Test past your original plan: A 3-minute brief became a 17-minute film because the author wanted to see what was possible, so give yourself room for that kind of experiment.
A fair caveat: 480p footage and a fast turnaround are a proof of concept, and you’ll want to judge the final quality for yourself. Still, it’s wild that a demo cut short by the clock could turn into a film this long in one weekend.
There’s more to this story
This post shares Part 1 of the film along with the process behind it, and the person who shared it plans to release the rest in the next two posts. So the next time a demo runs out of time, maybe don’t close the laptop just yet.
Head over to the full LinkedIn post to watch Part 1 and see the process for yourself. It’s well worth your time if you’ve been curious about long-form AI video.