A beat sheet built for a 90-minute feature does not shrink cleanly into a five-minute AI short. It has to be re-sized before you write a single scene, not after you have already generated one.

I wrote a beat sheet for a Lost Garden episode the way I was taught: fifteen beats, the full Save the Cat shape, each one earning its place on the page. It read well. It had a clean opening image, a proper midpoint, an all-is-lost moment with real weight. Then I started breaking it into shots and generating, and I watched a four-minute episode try to carry the structural load of a feature film. Half the beats never made the final cut. I had already paid to generate shots for some of them.

The fix wasn’t a better beat sheet. It was a shorter one, sized to the film before a single scene got written. For a short AI-generated film, that usually means three to five core beats, not fifteen, because every beat you keep is a beat you are about to pay to generate.

What is a beat sheet, and why doesn’t the standard one fit a short film?

A beat sheet is a beat-by-beat outline of a story’s structural turns, written before scenes or dialogue exist. Blake Snyder’s Save the Cat! system, still one of the most widely taught beat sheets in screenwriting, lays out 15 beats mapped to percentages of a feature-length runtime, from the opening image at the very start to the final image at the close.

That percentage mapping is the part that breaks first. Fifteen beats sized to a 110-page feature assume you have the room to let a beat breathe for several minutes before the next one arrives. A short film doesn’t have that room, and an AI-generated short has even less of it, because every beat you write down is a beat you will eventually have to turn into shots, and every shot is a real generation cost in time and money.

Fifteen beats were never a law of storytelling. They were a budget, sized to a runtime you almost certainly do not have.

How many beats does a short AI-generated film actually need?

For a short film in the three-to-seven-minute range, working screenwriters typically compress the Save the Cat structure down to a handful of essential turns rather than all fifteen beats. Boords’ guide to beat sheets notes that a five-to-seven-minute short may only carry three or four beats in its middle section, because character setup has to happen in one or two quick beats with no room for backstory, and the story has to turn fast enough that an audience feels the shape of it before it’s over.

In practice, for a short AI-generated film, four beats usually do the real work:

• The catalyst.The event that breaks the character’s normal. Everything downstream exists because of this one beat.
• The midpoint.The stakes stop being theoretical and become unavoidable.
• The low point.The plan fails, and the character has nothing left to fall back on.
• The finale.The change the film was building toward, earned or not.

Everything else Save the Cat lists, the theme stated, the B-story, the fun-and-games section, either gets folded into one of those four beats or cut outright. That is not a lesser version of the structure. It is the same shape, compressed to what a short can hold.

How do you size a beat sheet to your generation budget before you write it?

This is the step most people skip, and it’s the one that would have saved me the wasted generations on that Lost Garden episode. Every beat you keep is a beat you will eventually price in shots, and pricing it after you’ve already generated is the expensive way to find out you didn’t need it.

The workflow that holds up:

• Write the full beat list first, every beat the story seems to want, with no cost attached yet. This is where you think like a writer, not a producer.
• Price each beat in shots.Ask, honestly, how many shots this specific beat needs to read on screen. A quiet realization might cost one shot. A physical low point might cost three or four.
• Sum the shot count against your actual budget, time, money, or both, before writing a single scene. If the beat list adds up to sixty shots and you can afford twenty, you know that now, on paper, instead of finding out mid-generation.
• Cut the beats that don’t earn their shots.Not the ones that are badly written, the ones that cost more than they contribute to the four beats actually carrying the story.

This is deliberately the opposite of “generate first, edit later.” Cutting a beat sheet on paper costs nothing. Cutting a beat after you’ve already generated and reviewed shots for it costs whatever those generations cost, and that number adds up fast once you remember that a single shot in AI production commonly needs several variations before one is usable.

This plan has to live next to the script itself, which is the specific gap ScreenWeaver exists to close: a workspace where the beat sheet, the shot pricing, and the script sit next to each other instead of scattered across a notes app and a spreadsheet you stop updating by week two.

What’s the fastest way to tell if your beat sheet is still too long?

Run a one-sentence test on every beat before you write a scene for it: can you explain, in one sentence, why this beat has to exist between the catalyst and the finale? If you can’t, it’s not a beat, it’s a placeholder you inherited from a structure built for a different runtime.

A second, faster test: read the beat list out loud in under thirty seconds. If you can’t get through it that fast, it’s carrying more beats than a short film’s runtime can hold, no matter how good any individual beat is on its own.

I went back to that Lost Garden episode with both tests and cut it to four beats. The episode got shorter. It also got the only thing that mattered, a shape an audience could feel in four minutes instead of a shape built for a film I wasn’t making.

FAQ

Do I need to know all 15 Save the Cat beats before I can compress them?

It helps. Understanding the full structure is what makes an informed compression possible instead of an arbitrary one. You’re deciding which beats are load-bearing, and that’s easier to judge when you know what the other eleven were doing.

Does this apply to shorts made entirely from AI-generated shots, or only hybrid productions?

It applies even more to a fully AI-generated short than to a hybrid production. A traditional crew can improvise a beat’s coverage on set for free. An AI pipeline can’t. Every beat you keep gets priced in generations before it exists on screen, so oversizing the structure costs real money on top of time.

What if a beat genuinely needs more than one or two shots?

Let it. The point isn’t to force every beat into a single shot, it’s to know the true shot cost of each beat before you commit to the full list, so the beats that deserve more coverage are a choice, not an accident.

A beat sheet’s job has never really changed: decide what the story needs before you spend anything finding out the hard way. What changed is what “spending” means. On a traditional set, an oversized structure costs a longer shoot day. In an AI pipeline, it costs generations you already paid for on beats that never made the final cut. Size the sheet to the film you’re actually making, not the one the textbook was written for, and the four beats that survive will be the ones doing the real work.