Take a still from a video

A screenshot captures your screen, not the file. If the video is 4K and your window is showing it at a quarter of that, the screenshot is a quarter-resolution picture with the player’s scaling baked in. It will also catch the play button if you are unlucky with the timing.

Reading the frame out of the file gives you the pixels the camera recorded, at the size they were recorded, with nothing drawn on top. For a document, a thumbnail or anything that will be looked at closely, that is a different quality of result.

You give a time in seconds because that is how people know where things are in a video. Frame numbers depend on the frame rate, and 29.97 frames a second is common enough that counting frames is a good way to end up one second out.

The frame you get is the one displaying at that moment. Decimals work, so 12.4 is a little under halfway through the thirteenth second, which is usually precise enough to catch a specific gesture or expression.

JPEG, because a frame of video is a photograph and JPEG is what photographs are for. It is also what every upload form and every document editor accepts without thinking about it.

If you need transparency or a lossless copy, the conversion pages will turn the result into PNG afterwards. Doing that here by default would hand most people a file several times larger for a difference they would not see.

A thumbnail is a still, so everything about it depends on which frame is taken. A face mid-blink, a hand across the lens, a transition frame between two shots — all of them are technically correct grabs and none of them is usable.

Stepping to a moment where little is moving also gives a sharper result, because compressed video spends fewer bits on frames full of motion. A frame from a static shot is meaningfully crisper than one from a fast pan of the same clip, and the difference is not subtle.

The grab is the encoder’s pixels at whatever size the video was recorded, so a 4K clip yields a 3840-pixel still and a 720p clip yields a 1280-pixel one. Nothing is upscaled, and enlarging afterwards will not recover detail that was never recorded.

That is worth knowing when a platform asks for a thumbnail at a specific size. A frame from a 720p video will not satisfy a requirement for 1920 pixels without being stretched, and the honest fix is a higher-resolution source rather than a larger export.

The frame is extracted inside your browser tab, so the video is not uploaded. For a large file that is the practical difference between a grab that takes seconds and an upload that takes minutes — and many upload-based tools refuse video files of any real size outright.

It also means a long recording can be scrubbed and sampled repeatedly at no cost. Taking four frames and choosing between them is a reasonable way to work here, and would not be anywhere that charged an upload for each attempt.