A generator writes the file. Does the design work?

Turning a PNG into a font provider is a solved problem, and a generator that does it well saves real time. It just answers a different question than the one you usually have.

Two different questions

“Is this a valid font file?” is a question about syntax, and a generator answers it. “Does this design land where I meant it to?” is a question about the client, and only something that models the client can answer it.

The second question is where the time goes: the icon that is two pixels off, the title that is too wide by one, the line that collides with the one below it.

What a preview has to get right to be worth anything

A preview that paints its own idea of the result is worse than none, because you stop checking. This one draws from the same layout function the export uses, one code path and not two, and it refuses to invent values: character widths come from the client's font files, HUD anchors from measurement runs, format numbers from the version's own version.json.

Where nothing was measured, it says so. An unverified version profile is marked as such; a surface whose timing was never recorded says that in the warning rather than offering a number.

When a generator is enough

When the design is one glyph, when the position does not matter, or when you already have a working pipeline and only need the JSON. A tool that fits in your hand beats a tool that fits the problem badly.

What you get here beyond the file

The warnings listed on offsets by hand, the screen view that places every module where the player sees it, the plugin exports for Nexo, Oraxen, CraftEngine and ItemsAdder, and a live bridge that puts the design on a running Paper server without a client restart.