The still is what recruits the play

Every moving piece is first a stationary one

Before anybody watches your video, they see a single frame of it, in at least four places:

• the cover in a grid or profile,
• the poster in a player before it starts,
• the link preview when somebody shares the URL,
• the first frame as it scrolls into view, paused or muted.

In each case the still is doing the recruiting. A piece with a good opening thirty seconds and a black first frame has a good opening thirty seconds that nobody reaches.

The default is usually the worst available frame

Left alone, most systems take frame zero, or a frame chosen by an automatic heuristic. Frame zero is very often the darkest frame in the piece, because editors fade in. Automatic selection tends to pick a sharp, well-exposed frame, which is not the same as a meaningful one: a crisp shot of an empty workbench beats a slightly soft shot of somebody’s face by every criterion the machine has and by none that matter.

So set it explicitly, everywhere it can be set:

• Upload a custom cover wherever the platform offers one.
• Set poster on an HTML video element, so the player shows something before it loads.
• Set og:image for link previews, at about 1200 × 630, which is the size most cards crop to.

Link previews are cached, which is why yours is still wrong

The first time a URL is shared, the platform fetches the page, reads its og: tags and stores the result. Change the image afterwards and shares keep showing the old one, sometimes for a very long time, because nothing re-fetches on your schedule.

The practical consequence: get the preview right before the link goes anywhere. Most platforms offer a debug or card-validator tool that forces a re-fetch and shows what they actually read, and using it before you publish costs a minute. Using it afterwards costs an apology.

<meta property="og:title"       content="The £90 tool that halved our reject rate">
<meta property="og:description" content="A workshop, a jig, and four months of measurements.">
<meta property="og:image"       content="https://example.com/cover-1200x630.jpg">

What makes a good recruiting still

It is not the prettiest frame. It is the frame that makes the piece’s promise legible at the size it will appear.

• One subject. Two competing subjects at cover size is a blur.
• A face with a real expression, if the piece has one, because faces survive shrinking better than nearly anything else: people read expressions at resolutions far below what reading text requires.
• Legible at a fingernail. Shrink it to 150 pixels wide and look. Everything that disappears was decoration.
• Contrast against both light and dark interface, since you do not know which the viewer is using.
• Not a frame from the fade-in. Pull from a point where the piece is running.

Design a frame rather than finding one

The better habit is to plan for a frame rather than hunt for one. While shooting, take a deliberate still: the composition arranged, the subject placed, shot as a photograph. It costs thirty seconds on location and produces a cover that no frame grab will match, because video frames are compromises made for motion: a shutter speed chosen for movement, a moment chosen for what came after it.

For animated and graphic work the equivalent is to build the cover as its own artboard rather than exporting frame one.

The line that matters

A cover is a promise. The same discipline applies here as to a headline: it may withhold, and it may frame, and it may not depict something the piece does not contain.

The specific failure is a cover showing a moment that never happens: a reaction, a result, a dramatic instant assembled or generated for the cover alone. It works once. It is also, for synthetic imagery, increasingly a labelling question, and the duties differ by country and keep moving, so that is a matter for a lawyer where you are rather than a rule this page can give you.

The limitation

Optimising the still is optimising the front door, and it has a ceiling. A well-made cover on a piece that does not pay it produces a click, a few seconds, and an audience slightly less likely to click the next one. Platforms measure this directly: a high click-through paired with an immediate exit is a pattern their ranking systems are built to detect and demote.

So the cover is worth real effort and cannot rescue anything. It determines who arrives. The first second and the structure determine whether arriving was worth it.