What a catalogue owes you

Every platform that lets people share captures has ended up with the same three problems, and every one of them is documented by the people who use it. We went and read those threads before writing a line of the Hub, because the cheapest research available is somebody else's decade-old bug report.

A zero that means silence

The largest rig exchange in this category shows a star rating on every listing. On the first page of its newest entries, nineteen of twenty read 0.00 / 5.00.

They are not bad. Nobody has rated them. But 0.00 does not read as *nobody has said* — it reads as a verdict, and it poisons the sort that depends on it.

The mechanism is worth knowing, because it is an ordinary design mistake rather than negligence. A staff member explained it on their own forum: a rating only counts if you give it inside a particular pane of the desktop application, and any rating you give after saving the file is stored locally and discarded. So the only gesture that counts happens somewhere almost nobody goes.

Two rules came out of that. An unrated thing never renders as a rating here — below three ratings the page says how many there are and shows no average. And the rating is asked for where the judgement happens, next to the thing you just listened to, not three screens away.

A file you cannot hear

A capture is unlistenable until it is loaded into something. That is the whole problem, and the request for a fix is eight years old in one forum we read:

although there are audio demos (sometimes) available, it's not really a reliable indicator of how it will sound with your own guitar, and how it feels to play. I would even prefer to pay double for a profile that I can try before

One catalogue has since solved this properly, with an in-browser player that runs the model against a choice of source recordings and lets you bypass it to hear the dry signal. That is the bar, and we are not at it yet — what the Hub does today is let you attach a short clip and filter for the entries that have one. It is the smaller version of the right answer, and it is honest about being that.

Facets somebody typed

The last one is structural. Filters only work if the thing they filter on is a property of the file rather than a sentence somebody typed.

Where gear names are free text, EVH 5150 and EVH 5150 III Stealth become two different things, and filtering for one silently hides the other. Where the instrument is a dropdown that defaults to a value, the default is what most rows end up saying. A sample library that gets this right derives every sortable facet — tempo, key, length — from the file itself, and never asks a human for a field it later sorts on.

So gear here is an entity, not a string: you type "Marshall JCM800" and it resolves to a maker and a model, and every capture of that amp is findable together. When we later discover two entries are the same thing, the duplicate is pointed at the survivor rather than deleted, so every link that already exists keeps working.

One we added

There is a facet almost nobody has, which one hardware catalogue does have, and it turns out to be the most useful question on the page: what did the rest of your rig look like? Direct, four-cable, power amp into a cab, effects loop, straight into an amp input.

It is not a question about whether something sounds good. It is a question about whether it will work in your setup, which is the one you actually have. It costs the person uploading a single dropdown, and we would rather ask it than not.