Real cuts, not guesses
Automatic shot and scene detection
FrameCatcher finds where every shot begins and ends by reading the content itself, so you get real cuts instead of frames grabbed on a fixed timer.
How it works
- 1
Choose your video
No settings to guess at, and no upload. The footage is analysed in this tab.
- 2
Shots are detected
Each real boundary is found from the content, giving you an accurate shot list.
- 3
Use the results
Get a frame from each shot and export them however you need.
- Real content-based shot detection, not fixed time slices
- An accurate shot list for any video
- A representative frame pulled from every shot
- Works on fast-cut and slow footage alike
What people use it for
Knowing how many shots are really in an edit
Counting cuts by hand is slow and inaccurate past about a minute. Detection gives you the number, where each one falls, and how long each is held.
Studying how someone else cut a scene
The rhythm of an edit is the part that is hardest to see while watching it. Laid out as shots with durations, it becomes obvious: where the cutting tightens, where a shot is allowed to breathe.
A starting point for a shot list or a reconstruction
If you have to rebuild, re-time or describe an edit that already exists, a real shot list beats scrubbing for cuts.
Frequently asked
- How is this different from grabbing a frame every N seconds?
- Fixed intervals miss cuts and land mid-shot. FrameCatcher detects the actual boundaries, so each frame belongs to a real, distinct shot.
- Does it work on quick cuts?
- Yes. Fast-cut footage simply produces more detected shots, each with its own frames.
- What counts as a shot?
- A run of frames with no cut in it. FrameCatcher compares consecutive frames and marks a boundary where the picture changes more than a set amount, so a shot ends where the content genuinely changes rather than where a timer goes off.
- Does it catch fades and dissolves?
- Hard cuts are what it is reliable at. A slow dissolve changes the picture gradually, which is exactly the signal a content-based detector is built to ignore, so a long dissolve may be missed or placed slightly late. Worth knowing before you trust a count on a heavily dissolved edit.
- Can I change how sensitive it is?
- Not from the page. It runs at a fixed threshold that was matched against PySceneDetect's own defaults, so results are consistent between videos instead of depending on a setting you had to guess at.
Ready to try it?
Pick a video and see it work. Free, no account, and it never leaves your browser.