Extract the perfect frame from any video
Step one true frame at a time, take a whole run at an interval, or let the difference graph find the cuts for you. Everything decodes in this tab and the file never goes anywhere.
Drag & drop a video or click to upload
MP4 · WebM · MOV and anything else your browser can decode — the file never leaves your device
How it works
Drop a video
Anything the browser can decode. The size, the resolution and the real frame rate are read off the file itself — the fps box is gone because it was a guess.
Land on the frame
Arrow keys step whole frames and each step waits for the picture to change. Zoom with the wheel to check focus, and hold Alt to compare against your last capture.
Or take a whole run
Every N seconds, N evenly spaced, or every frame of a range. The scene scanner draws the difference between samples so you can move the threshold instead of trusting it.
Export
PNG, JPG or WebP, at full size or scaled down, one at a time or as a ZIP. The gallery tells you how much memory it is holding, and hands a frame straight to the next tool.
The capture waits for the frame
Seeking is asynchronous, so drawing the video straight after moving grabs whatever was on screen before — usually the previous frame. Every capture here awaits the frame that is actually presented.
The frame rate is measured
requestVideoFrameCallback reports the presentation time of each frame, so the rate comes from the video instead of a box you had to fill in. Common rates snap, but 23.976 stays 23.976.
Batch extraction
Every N seconds, N evenly spaced across a range, or every frame between two points. It shows how many it will take before it starts, and stops the moment you ask.
Scene-cut detection
A luma difference between downscaled samples, plotted as a graph you can click. The threshold is a number you control, not a hidden verdict.
Zoom to the cursor
Wheel-zoom up to 12×, drag to pan, double-click to reset — for checking focus or reading small text before you commit to a still.
Memory you can see
Full-size blobs add up fast, so the strip shows thumbnails, every frame reports its own weight, and the total warns you before the tab gets heavy.
Nothing is uploaded
The video is decoded by the browser from a local object URL. No server, no queue, no telemetry on what you are working with.
Hands off to the rest of the suite
Send a frame straight to CropSnap, CompressSnap, the background remover or MemeBolt without downloading and re-uploading it.
The frame you saw, not the one before it
The frame rate is measured from the video itself and every capture waits for the frame to actually be on screen, so the still you save is the still you were looking at.
Decoded by the browser you already have
The video is played from a local object URL and the frames are drawn to a canvas, so extraction runs at the speed of your own decoder with no encoding queue in front of it. Nothing is uploaded, nothing waits on a server, and closing the tab is all it takes to be rid of the file.
Thumbnails, evidence, reference, proof
Pulling the cleanest still for a thumbnail, catching the exact frame where a bug appears, collecting reference for a design, or documenting what happened at 00:04:11 — all of them need the frame you were looking at, not the one before it. With the rate measured and the seek awaited, the frame number on screen is the frame number in the file, and stepping back and forth lands on the same picture every time.
Your video stays on your device. There are no uploads, no databases and no tracking — the file lives only in local memory and disappears when you close the tab.
Frequently Asked Questions
QIs my video uploaded anywhere?+
No. The browser decodes it from a local object URL and the frames are drawn to a canvas in the same tab. There is no server component, and the file is gone when you close the page.
QHow does it know the frame rate?+
It measures it. The video plays muted for a fraction of a second while requestVideoFrameCallback reports the presentation time of each frame, and the median gap becomes the rate. Where the browser does not provide that callback the tool says so instead of pretending.
QWhy is stepping slower than it used to be?+
Because each step now waits for the new frame to be presented before it says it has moved. That wait is the difference between saving the frame you are looking at and saving the previous one.
QCan I extract many frames at once?+
Yes. The batch tab takes one every N seconds, N evenly spaced across a range, or every frame between two points, up to 500 at a time. It tells you the count before it starts and stops as soon as you press Stop or Escape.
QWhat does scene detection actually measure?+
The mean absolute luma difference between consecutive samples taken at a 64-pixel-wide downscale. Everything is plotted, so you can see how far each sample sits from your threshold and move the threshold rather than trust a hidden score.
QIn what quality are frames exported?+
At the video’s native resolution by default, as lossless PNG or as JPG/WebP with a quality slider. You can also export at 50% or 25% when you need a contact sheet rather than a master.