Convert MP4 to JPG

Taking a screenshot of a paused player is the obvious way to get a picture out of a video, and it is the reason most people end up here disappointed. A screenshot records your screen: the video scaled to whatever size the window was, a control bar that faded in when you moved the mouse, and the colour handling of the display rather than of the file.

Reading the frame out of the file avoids all three. A video recorded at 3840 by 2160 gives you an image at 3840 by 2160 even if the player was a quarter of that on screen, with no interface in it and nothing scaled. On a phone, where the player is small and the recording is not, the difference is dramatic.

The frame is chosen by a timestamp in seconds from the start, not by scrubbing. That sounds less convenient and is usually faster: if you know the moment you want is around the fifty-second mark, typing 50 and converting takes less time than dragging a playhead, and typing 50.5 gets you the next half second.

Converting again with a different number is cheap, though nothing is kept between runs: the file is opened again and decoded forward from the keyframe before your timestamp rather than from the start, which is a fraction of a second wherever in the video you land. Comparing three candidate frames is three numbers and three downloads, and none of it touches a server.

Video codecs do not store every frame in full. They store occasional complete pictures and, in between, the differences from the frame before — which is what makes an hour of video fit in a few gigabytes at all. A frame reconstructed from differences during fast motion carries less real detail than one from a shot where nothing moved.

So a still from a hand-held pan looks softer than a still from a locked-off interview, and no setting on this page changes that. It is the video’s own encoding, decided when the file was made. Pausing the player on the same moment shows you exactly the same softness — the conversion is not adding it, it is showing you what was there.

JPG is right when the picture is going somewhere: attached to an email, dropped into a document, posted. It is a fraction of the size and, on a photographic frame, indistinguishable at any sensible quality setting. That covers most of what people take frames for.

PNG is right when the frame is going into an editor and will be saved again, and when the content is not photographic — a title card, a slide, a screen recording of an interface. Those are flat colour and hard edges, which is the structure JPG handles worst; the haloing around text shows up immediately. The PNG will be several times larger for the identical frame, which is the price of not deciding anything away.

Video generators hand back short MP4s, and a single frame from one is wanted constantly: a thumbnail, a preview for someone who will not watch five seconds of footage, or a still to feed back into an image prompt as a reference. Pausing and screenshotting captures the player furniture too, at whatever size the window happened to be.

A frame grab takes the encoder’s own pixels at the clip’s own resolution. The caveat from further up applies with particular force to generated video: motion frames are softer than still ones, so a frame from a fast camera move will disappoint. Step to a moment where little is moving and the still is as sharp as the clip can give.

The decoding runs on your own device, through the same hardware path the browser uses to play video. The file is never sent anywhere, so there is no upload to wait through, no queue and no daily count — but the free tier takes files up to 100 MB, which a few minutes of phone footage passes, and the frame has to be decoded inside the tab’s memory once it is in.

It also means this works with the network disconnected once the page has loaded, which is the simplest way to satisfy yourself that the claim is true rather than taking our word for it — switch off the wifi, convert a frame, and watch it succeed.