When you paste a video link into ramrom.io and tap the button, nothing happens for a second or two. The browser is waiting on resolve.php, which is asking a provider API what that link actually contains. The question I had to answer was simple: what should the visitor see during that wait?

My first version showed a spinner. A spinning circle in the middle of the page, and the rest of the page sitting empty underneath it. It worked, technically. But it told the visitor nothing. A spinner says "something is happening" and nothing else. It does not say what is happening, how long it might take, or what will appear when it finishes.

Worse, the spinner made the page feel broken on a slow connection. On a Pakistani mobile network, resolve.php can take three or four seconds when the provider is slow. Four seconds of a spinning circle feels like the site has frozen. I watched people tap the button again, and then again, because they thought the first tap had not registered. Every extra tap burned another provider API call, and those calls cost quota.

So I replaced the spinner with a skeleton loader. A skeleton is a grey, shimmering outline of the thing that is about to appear: a block where the video thumbnail will go, a short bar where the title will sit, and a row of bars where the quality options will be listed. It is the shape of the answer, shown before the answer arrives.

The idea is not new, and the implementation is deliberately boring. In the page I have a hidden container with the skeleton markup in it: plain divs with a class that gives them a light grey background and a slow pulsing animation in CSS. No JavaScript library, no extra dependency. When the visitor taps the button, I hide the empty results area and show the skeleton container instead, then fire the fetch to resolve.php. When the response comes back, I swap the skeleton out and render the real results into the same space.

Two details matter. First, the skeleton has to match the shape of the real result. If the skeleton shows three quality rows and the real answer has two, the page jumps when it swaps. So I built the skeleton to match the most common result: one thumbnail block, one title bar, and three option rows. A small jump is still possible, but the layout barely moves, and that matters more than it sounds. Layout shift is the thing that makes a page feel cheap, even when everything technically works.

Second, the skeleton has to have a limit. If resolve.php fails or times out, the skeleton must not sit there forever. I set a hard timeout on the fetch, and on failure I replace the skeleton with a plain error message that says what happened: the link did not resolve, or the provider is busy, or the video is private. An honest error after a skeleton feels like the site tried. An endless skeleton feels like the site died.

There is one more reason I prefer the skeleton. It stops the double-tap problem. While the skeleton is visible, the button is disabled, so a second tap does nothing. The visitor can see the page is working on their request, and they can see roughly what shape the answer will take. They wait. On the slow connections I test on, that patience is worth more than any optimisation I could make to the API call itself.

A spinner is a promise that something is happening. A skeleton is a preview of what is coming. For a tool site, where the visitor's whole job is paste, wait, download, the preview is the better promise. It costs about twenty lines of HTML and CSS, it needs no library, and it made the wait feel shorter without making anything actually faster. Sometimes that is enough.

If you are building a small tool on shared hosting, try it. Hide the spinner, draw the shape of your answer in grey boxes, and let your visitors see what they are waiting for.