Video Shot Breakdown & Episode Planner

Measure how often a reference video cuts, then split your script into a series of short episodes that follows the same rhythm. The file never leaves your device, and no AI is called, so the same input always gives the same result.

MP4 · MOV · WebM — Your file is never uploaded; it is analysed on this device only. No sign-up, no limits.

Build an episode plan

Enter a topic and a script to get the episode count and, for each episode, a hook, beats and a next-episode slot. If you analysed a video above, you can bring its cut rhythm straight in.

How to use the shot breakdown and episode planner

Press “Choose a video file” and pick a reference video. Analysis starts at once and lists every shot with its start, end, length and a thumbnail.
If too many or too few cuts are found, move the “Cut sensitivity” slider. A lower number treats smaller changes as cuts.
Press “Use this rhythm in the episode planner” to send the average shot length and the first-shot length to the planner below.
Paste your series topic and script (or storyline) and press “Build the plan”. You can take the result away as Markdown, CSV or JSON.

What the numbers mean

Average shot length (ASL) is the running time divided by the number of shots, and it is the standard measure film researchers use to compare editing pace. A study of 160 Hollywood films released between 1935 and 2010 found ASLs of about 10 seconds in the 1930s and 1940s falling to below 4 seconds after 2000 (Cutting et al., “Quicker, faster, darker”, i-Perception, 2011). Vertical shorts are often cut far tighter than that, so measuring a video from the channel you want to learn from tells you more than any rule of thumb.

Median is the middle value when the shot lengths are put in order. When one long shot drags the average up, the median is closer to how fast the video feels.
Cuts per minute is how many times the picture changes in a minute. An average shot of 2 seconds means about 30.
First shot (hook length) is the time from the start until the picture first changes: how long the opening image is held to catch the viewer.
Pacing sorts the average shot length into four bands: over 10 s is slow, 5–10 s steady and explainer-like, 2–5 s dynamic and social-style, 2 s or less rapid-fire. The bands match the reference-video analyser in OpenMontage, an open-source video production system.
Silences over 1 s are stretches where the sound stays below −40 dBFS for more than a second. Use them to find gaps in the narration.

How cuts are found, and the limits

The video is sampled every 0.1 seconds. Each sample is shrunk to 64×36 pixels and compared with the previous one on a 0–255 scale. When the difference is above the threshold (27 by default) and at least three times the average of the samples around it, it counts as a cut, and the tool then narrows the gap to place the cut to within 0.01 seconds. The threshold of 27 and the factor of three follow the defaults of the open-source PySceneDetect. Dissolves and fades change too little between samples and can be missed; strobe lighting or a full-screen caption can be counted as a cut. Videos longer than four minutes are limited to 2,400 samples, so the sampling step grows accordingly.

How the episode plan is split

The plan is calculated by fixed rules, not written by an AI. The length of the script divided by the reading speed gives the total time; dividing that by the target length gives the number of episodes; the script is then cut at the sentence ends that make the episodes most even. It never cuts in the middle of a sentence.

Structure: the first sentence becomes the hook, the last one the payoff, and everything between is divided into context and three beats. Each episode has an empty “next-episode tease” slot, and the last one is reminded to answer the question episode 1 raised.
Timing: the hook takes the hook length you set (3 seconds by default), the payoff the last 20%, context 13%, and the three beats share the rest.
Shot count: episode length divided by average shot length. A 60-second episode at 2.5 seconds per shot needs 24 shots, which tells you how much footage to prepare.
Reading speed: the starting value is 150 words per minute (OpenMontage's short-form guide gives 125–150 words for 60 seconds). To calibrate it, read part of your script aloud for 30 seconds and enter twice the number of words you got through.

Choosing a target length

YouTube classifies square or vertical videos of up to three minutes, uploaded after October 15, 2024, as Shorts (YouTube Help, checked October 9, 2026). Length limits on other platforms change often, so check each app before you post. If the plan warns that an episode runs over the target, trim sentences or add an episode.

If you would rather have an AI refine the split, copy the prompts under the plan. The full set, from finding stories and checking sources to the script and a pre-publish checklist, is at promptsorted.com. To lift one section out of a reference video use Trim video; to turn footage vertical use Resize.

FAQ

How are cuts detected?

The video is sampled every 0.1 seconds and each picture is compared with the one before. If the difference is above the threshold (27 by default) and at least three times the average around it, it counts as a cut. This is the same principle as the open-source PySceneDetect.

Is my video uploaded?

No. The analysis runs entirely in your browser and the file never leaves your device.

It misses cuts, or finds too many.

Gradual transitions such as dissolves and fades can be missed. Lower the sensitivity slider to find more cuts, raise it to find fewer.

Is the episode plan written by AI?

No. It follows fixed rules: split at sentence ends so the episodes are even, then fill the hook, context, three beats, payoff and next-episode slots. Ready-to-copy prompts are included if you want an AI to refine it.

Does it work on long videos?

Yes, but above four minutes the sampling step grows beyond 0.1 seconds, so very short shots may be missed. It works best on Shorts and Reels length videos.