How long should a shot last? Ask an editor and you'll get "it depends." Ask a film school professor and you'll hear about pacing and rhythm. Ask YouTube and you'll find essays about how the average shot length in Hollywood has dropped from 12 seconds in the 1960s to 4 seconds today.
None of that helps when you're sitting in front of a storyboard at 2am wondering if shot 14 should be 3 seconds or 6.
What does help is data. And data on individual shot durations, broken down by what the shot actually does in the story, is weirdly hard to find.
The dataset
The dataset we're drawing from was compiled by Shanyin, an award-winning AI filmmaker, as part of the open-source Shanyin Director Master project. Shanyin analyzed 551 shots across 9 short films (3-7 minutes each) and categorized each by narrative function and information density — producing the kind of empirical timing data that's hard to find in filmmaking education.
We should be upfront about the limits. 551 shots from 9 short films is a real dataset, not a guess, but it's also not a massive sample. The films are all in the 3-7 minute range, which means feature-length pacing will differ. Long-form drama holds shots longer. Action films cut faster. These numbers are baselines, not rules.
That said, for short-form content, storyboards, and AI-assisted filmmaking where you need a starting point for duration estimation, this is the most granular public data we've found.
The headline numbers
- Average shot duration: 4.2 seconds
- Median: 3 seconds
- 76% of shots fall between 1 and 5 seconds
- Only about 8% of shots exceed 8 seconds
The median being lower than the average tells you the distribution is right-skewed. Most shots are short. A few atmospheric or thematic shots pull the average up. This matches what you'd expect from well-edited short films: the default is brisk, with deliberate slow moments for emphasis.
Timing by narrative function
This is where the data gets useful. Instead of asking "how long should a shot be?" you ask "what is this shot doing?" and the duration follows.
| Function | Duration | Why |
|---|---|---|
| Push / Upgrade | 2.5-3.5s | Forward momentum. These shots drive the story forward. Quick cuts maintain energy. |
| Explain / Establish | 4-5s | The audience needs time to read a new location, a new character, or a new situation. Rushing setup creates confusion. |
| Display / Present | 4-5s | Showing something that matters. A prop, a detail, a reaction. Long enough to register, short enough to not overemphasize. |
| Emotion / Atmosphere | 4-5s | Let it breathe. Emotional moments need air. Cutting too fast through a grief scene undermines the feeling. |
| Turning Point | 5-7s | The weight of change. When the story shifts direction, the shot needs to hold long enough for the audience to feel the shift happen. |
| Sublimation / Theme | 6-10s | The slow revelation. Final images. Thematic statements. These are the shots the audience remembers, and they need time to land. |
The pattern is clear: shots that move the plot forward are short. Shots that carry emotional or thematic weight are long. The middle ground — establishing, displaying, atmosphere — clusters around 4-5 seconds.
This means function dictates duration more than content does. A close-up of a face might be 3 seconds (push) or 7 seconds (turning point) depending on what it's doing in the story. The framing is the same. The narrative role is different.
Timing by information density
Another way to think about shot duration: how much information does the viewer need to absorb?
| Density | Duration | Example |
|---|---|---|
| Single piece of info | 2-3s | A hand picks up a key. One action, one piece of information. |
| Two pieces | 4-5s | A character enters a room and sees something unexpected. Two beats: the entry and the discovery. |
| Multi-layered | 6-8s | A tracking shot through a crowded market. The viewer needs to register the environment, the mood, specific details, and the character's movement through it. |
| Long take (comprehensive) | 10-15s | An unbroken shot that covers an entire conversation exchange or a complex blocking sequence. |
The rough rule: the eye absorbs about one piece of visual information every 2-3 seconds. A shot with one thing to say needs 2-3 seconds. A shot with four things to say needs 8-10. If you're holding a shot for 6 seconds but it only contains one piece of information, the audience will feel it dragging.
Practical rules
Here's how we've been applying this data when designing storyboards.
The 3-second test. If a shot has one piece of information, 3 seconds is probably enough. If you're holding it longer, ask yourself what else the viewer is supposed to be getting from the extra time. If the answer is "atmosphere," that's valid. If the answer is "I don't know," the shot is too long.
The "hold it longer" trap. When a shot feels like it needs more time, the problem is often not duration but design. The shot might need a different angle, a closer framing, or a cut to a reaction shot. More seconds don't fix a weak composition. They just make the audience wait longer for something that isn't coming.
First shots are longer. The first shot of a scene is almost always the longest. The audience needs to orient — where are we, who's here, what's the mood. An establishing shot of a new location at 2 seconds will feel rushed. Give it 4-5.
Last shots vary. The final shot of a scene can be very short (hard cut to the next scene, creates momentum) or very long (lingering on a face or an empty room, creates emotion). This is a deliberate choice about what you want the audience to carry into the next scene.
Genre adjustments:
- Horror holds shots longer than average. Dread builds in stillness. A 6-second shot of an empty hallway is boring in a comedy and terrifying in horror.
- Comedy cuts faster. Comic timing depends on the cut landing at exactly the right moment. An extra second kills a joke.
- Action alternates: 1-2 second combat shots for intensity, then 4-5 second breathing room shots so the audience can re-orient. All fast cuts and the fight becomes illegible. All slow and it loses energy.
A sample scene, annotated
Here's what these rules look like applied to a simple 5-shot scene — two characters meeting at a train station.
| Shot | Description | Function | Duration |
|---|---|---|---|
| 1 | Wide: the train station at dusk | Establish | 5s |
| 2 | Medium: Character A waiting on the platform | Display | 4s |
| 3 | Train arrives, Character B steps off | Push | 3s |
| 4 | Close-up: they see each other | Turning Point | 6s |
| 5 | Medium: the embrace | Emotion | 5s |
| Total | 23s |
Shot 4 is the longest because it's the moment the scene has been building toward. Shot 3 is the shortest because it's pure forward motion. Shot 1 is second-longest because the audience needs to read the location. The durations aren't random. They follow from what each shot is doing.
How we use this
Genkee's storyboard agent uses these duration baselines when estimating shot timing. When you design a storyboard, the agent tags each shot by narrative function and suggests a duration range. You can override it — these are suggestions, not constraints — but having a data-informed starting point means you spend less time guessing and more time making deliberate pacing choices.
You can see the full list of references, directors, and techniques that inform our approach on our References & Open Source page.
The timing data in this article comes from the Shanyin Director Master, an open-source directing methodology by AI filmmaker Shanyin. All illustrations generated using Genkee AI.