We built the beat engine expecting that the most common feedback category would be structural gaps: missing act breaks, underdeveloped midpoints, absent inciting incidents. That is what the structural analysis literature would predict, and it is what most discussions of first-feature screenplay problems focus on.
What we found when we reviewed the patterns across eighty scripts from our early-access cohort of fifteen writers was different. The structural elements were mostly present. The act breaks existed, the inciting incidents were there, the character arcs had identifiable turning points. What was wrong, in the majority of scripts that the beat engine flagged as needing significant revision, was pace: specifically, the management of narrative velocity across act transitions.
This was a productive surprise. It changed what we were building and it changed how we think about what structural analysis actually needs to do.
What Pacing Actually Is
Pacing is not the same as the speed at which events happen. A fast-paced film is not a film where things happen quickly; it is a film where the narrative never gives the audience's attention a reason to wander. A slow-paced film is not a film with long scenes; it is a film where the dramatic tension has dissipated and the audience is waiting for it to return.
What controls pacing, at the structural level, is the relationship between the information the audience has and the information they need to understand what is at stake. When the audience knows exactly what is at stake and the scene is delivering on that stake, pacing is fast even if nothing is happening externally. When the audience is uncertain what they are supposed to care about, or when a scene is not doing anything to advance the dramatic question they came in with, pacing feels slow even if the scene has a lot of action.
Act transitions are the highest-risk moments for pacing problems because they are the moments when the dramatic question shifts. At the Act 1 to Act 2 transition, the story's central question is being established for the first time. At the Act 2 to Act 3 transition, the question is being moved toward resolution. Both transitions require the audience to update their understanding of what the story is asking, and any sequence that delays that update, or performs it clumsily, produces a pacing valley that is very difficult to recover from.
The Pattern in the Scripts
The pacing problem we saw most consistently in the early-access cohort scripts was what we started calling "the transition buffer": a cluster of scenes immediately after an act transition that seemed to re-explain what the transition had just established, rather than advancing the new dramatic question.
Here is the pattern in practice. A strong inciting incident occurs at page ten. The protagonist makes a decision in response to it and enters Act 2. Then pages eleven through twenty are spent showing the protagonist coming to terms with the decision they already made. The scenes are not useless; they contain character texture and relationship development. But they are doing emotional processing of an event the audience has already understood and integrated. The dramatic question of Act 2 is not advancing during those pages; it is waiting for the processing to finish.
The same pattern appeared at the Act 2 to Act 3 transition: a strong crisis or low point, followed by several scenes of the protagonist in emotional response to that crisis, before the story moved into its final sequence. The emotional response scenes were not wrong in themselves. They were in the wrong place. The audience's appetite at the bottom of Act 2 is for action and resolution, not reflection. Reflection belongs in the spaces within the action, not as a holding pattern before the action resumes.
Why This Happens
Writers produce transition buffers for understandable reasons. An act transition is a significant event, and it feels natural to give the characters time to respond to it. Writers who work from their characters' psychology outward will naturally want to show the internal process of adjustment. The intuition is correct; the placement is wrong.
There is also a structural honesty problem at play. Some transition buffers exist because the writer has not fully worked out what the next dramatic question is. The processing scenes fill space while the writer figures out where the story needs to go. In a finished script, those scenes stay in because removing them feels like cutting material that adds depth. But they are costing pace in exchange for depth that the audience does not experience as depth; they experience it as drag.
The third cause, specific to Indian commercial material, is the convention of emotional underscoring: allowing the emotional weight of a significant moment to be felt before moving forward. This convention has a legitimate place in Indian narrative traditions. What we saw in the early-access scripts was writers applying it at act transitions without distinguishing between emotional moments that earn extended underscoring and transitions that need to keep moving. Not every transition is a moment that deserves to breathe.
What We Changed in the Beat Engine
After identifying this pattern, we revised the beat engine's transition analysis to track narrative velocity specifically at act transition zones. The system now flags sequences where dramatic question advancement has stalled for more than a threshold number of consecutive beats in the first section of a new act. The annotation explains the pacing function of the transition zone and asks whether the current scenes are advancing the new dramatic question or processing the previous act's events.
This is a harder analysis to do reliably than structural gap detection. Structural gap detection asks whether an element is present; pacing analysis asks whether the narrative is moving, which requires a more nuanced judgment about what "moving" means at different points in the story. We are still calibrating the sensitivity. Our early testing suggests the transition buffer flag has too many false positives for certain kinds of character-driven material where the processing scenes are genuinely advancing the protagonist's internal arc in ways the system does not yet recognize well. We are working on this.
We are also adding a separate pacing view that shows narrative velocity across the whole script as a graph, giving writers a visual representation of where momentum is building, where it is holding steady, and where it drops off. The underlying data comes from the beat engine's sequence-by-sequence analysis. The visualization is for writers who respond better to a spatial representation of pacing than to sequential annotations.
What the Writers Taught Us
The most important thing our early-access cohort taught us was that our initial theory of what screenwriters needed help with was incomplete. Structural knowledge is more widely distributed among working writers than we had assumed. They know about acts and beats and turning points. What they do not always know is how to feel the difference between a script that is moving and a script that is formally correct but experientially stalled.
Pacing is an experiential problem, not a formal one. It only shows up when someone reads the script and notices that their attention drifted at a specific point. First-time writers often do not have enough readers at a sufficiently early stage to catch this before the script is further along in development than is cheap to fix.
We are not claiming the beat engine catches everything that a skilled reader would catch. It does not. What we are building toward is a system that catches the patterns that show up consistently enough to be identifiable without a human reader, so that the human reader's time is spent on the things that require genuine interpretive judgment. Transition buffers are consistent enough and common enough to be one of those patterns. Finding them earlier in the development process, before a completed draft, is the goal the early-access data pushed us toward.