Find Every Scene Cut in a Video
Upload a video and the detector walks through it, finds where one shot ends and the next begins, and gives you a numbered list of scenes with a start time, an end time, and a thumbnail for each. It runs in the browser by comparing frames, so the video never leaves your device and there is no app to install.
This is the mechanical job, splitting a video at its cuts, not an AI guessing what the video is about. If you need a cut list to start editing, a way to jump to each shot, or a quick count of how many scenes a clip has, this is the tool for that.
What a Scene Detector Actually Does
A scene cut is the moment the picture changes from one shot to another. Detecting them turns a single video file into a list of segments you can act on, and that list is the thing most editing and captioning workflows actually need first.
Worth being precise about the words, because search results mix them up. A shot is one continuous run of camera between two cuts. A scene is usually several shots that belong together. Frame-difference detection like this finds shots, the hard visual boundaries, which is what “scene cut detection” and “shot boundary detection” both point at in practice. It does not read the story to decide where a narrative scene begins, that is a judgment call no automatic tool makes reliably.
How It Finds the Cuts
The method is deliberately simple, which is why it can run on your own machine with no server. The tool samples the video at a steady step, shrinks each sampled frame down to a tiny grayscale thumbnail, and measures how different each frame is from the one before it. A hard cut makes almost every pixel change at once, so the difference spikes. When that spike crosses the sensitivity threshold, the tool marks a cut there and opens a new scene.
Because it samples rather than checking every single frame, a cut lands on the nearest sample, so a timecode can be off by a fraction of a second on a long video. That is the trade for keeping it fast and in the browser. For an edit list or a rough segmentation it is close enough, and you can see each boundary in its thumbnail to confirm it.
The sensitivity control is the knob that matters. High catches subtle changes and returns more scenes, low only breaks on big changes and returns fewer. Start on Medium, then move it if the count looks too high or too low for the footage.
Hard Cuts, Fades, and What Gets Missed
Hard cuts are what this does well. A straight edit from one shot to another shows up clearly because the frame difference jumps.
Gradual transitions are the honest weak spot. A slow fade to black, a cross dissolve, or a long morph changes the picture a little at a time, so no single frame-to-frame difference is large enough to trip the threshold. Raising the sensitivity helps, but a soft dissolve can still read as one continuous scene. Fast motion or a hard camera flash can push the other way and split one shot into two. The thumbnails are there so you can spot and ignore a false split at a glance.
None of this is a defect unique to browser tools. Frame-difference detection has these limits everywhere. What you get here is a fast, private first pass that is right on the great majority of ordinary cuts.
Scene Edit Detection, Scene Cut Detection, Same Job
The naming is a mess and it is worth clearing up, because people search all three and get different results.
DaVinci Resolve calls it Scene Cut Detection, and it lives on the Media page rather than in the timeline.
Filmora calls it Smart Scene Cut. CapCut calls it Split Scenes on desktop, and folds it into AutoCut on mobile and web. PySceneDetect just calls it detection and gives you detect-content and detect-threshold modes.
Different names, one job: find the frames where the shot changes. What differs is not the detection so much as what the tool hands you afterwards. Premiere and DaVinci cut your timeline. CapCut and Filmora split clips ready to edit. PySceneDetect prints a CSV. This gives you a timecode list with thumbnails you can read and copy, which is the right output when you want to know where the cuts are rather than immediately edit them.
Browser Tool vs DaVinci and PySceneDetect
The tools people usually reach for are DaVinci Resolve and PySceneDetect, and both do far more than this. The question is what you are willing to set up for one clip.
| Tool | Where it runs | Setup needed | What you get back | Free |
|---|---|---|---|---|
| This detector | Browser | None | Timecode list with thumbnails | Yes |
| Premiere Pro | Desktop app | Creative Cloud subscription | Cuts, subclips, or markers on the timeline | No |
| DaVinci Resolve | Desktop app | Large install, import first | Cuts on the Media page | Free tier available |
| CapCut | Desktop, mobile, web | Account and app | Split clips ready to edit | Yes on desktop |
| Filmora | Desktop app | Install, paid for export | Scenes extracted as highlights | Trial only |
| PySceneDetect | Command line | Python and pip | CSV, stats, or split files | Yes, open source |
| HandySaw DS | Windows app | Install | Footage split into scene files | Trial |
Verified September 2026 from each vendor’s documentation.
DaVinci Resolve has a scene cut detection feature, but it means installing a large editing suite and importing the clip before you can run it. PySceneDetect is an open-source command-line tool with content-aware and threshold modes, which is powerful if you are comfortable in a terminal and installing Python packages. Neither is something you open for one quick check on one clip.
This tool is the no-install, no-account version for exactly that quick check. You drop a file in, read the cut list, copy the timecodes, and move on. For a heavy batch job or frame-perfect boundaries inside a finished edit, the desktop tools still win, and that is a fair place to send you.
Who Uses Scene Detection
Editors use a cut list to break raw footage into clips before assembly, or to find the exact seconds where B-roll changes so they can drop in music or captions on the beat.
Captioning and subtitling work benefits because a subtitle file reads better when lines break on shot changes instead of running across a cut. A scene list gives you those break points.
People logging or archiving footage use it to index a long recording by shot, so a two-hour capture becomes a navigable list instead of a scrub bar. And anyone comparing two versions of an edit can count scenes and check the cut points quickly.
Scene Detection on an iPhone or Android Phone
This runs in the phone browser, so there is no app to install and no App Store detour. Open the page in Safari or Chrome, pick a video from your camera roll, and you get the same cut list with thumbnails.
Two honest caveats. Phones are slower at this than a laptop because the detection compares frames one at a time and a phone has less to spare, so a long clip takes noticeably longer. And a 4K video shot on a recent iPhone is a big file to chew through, so trim it first if you only care about one section.
If you are editing on the phone anyway, CapCut’s Split Scenes is the more practical route, because it cuts the clips where you are already working instead of handing you numbers to type in. Use this when you want to read the cut list rather than act on it immediately.
Beyond the Cut List
Once you know where the scenes are, the next step is usually to do something with them. To understand what is inside each scene rather than just where it cuts, the AI video analyzer reads the content, objects, and events with timestamps. To pull a still image at a specific moment or on an interval, the video frame extractor saves frames as PNG or JPG. And if your question is whether a clip was AI-generated or manipulated rather than where it was cut, that is the AI video detector, a different job entirely.
