For anyone who started a plugin and never shipped it — where did it actually die?

[circuit-hazw]

Been building my first proper plugin and I keep noticing how little of my time went into the DSP. The audio code was almost the easy part. Most of it vanished into UI, code signing, notarisation, CI, packaging, getting the thing to even load on someone else's machine, sorting out licensing, and all the manual testing across multiple platforms/devices. Got me wondering whether that's universal or just my particular flavour of pain.

So, honest question for anyone who's started something and never got it out the door: where did it die? The idea — not knowing what was actually worth building? The DSP itself? The GUI? The build/sign/packaging side? Testing across hosts? Licensing and payments? Getting anyone to notice once it was done? Or just ran out of time and steam?

Not fishing for anything in particular — genuinely curious which stage is the real graveyard, because I suspect everyone assumes it's whichever bit happened to bite them. War stories welcome.

[Benedict]

This is a massive issue with Coding. And even in SynthEdit: Banging some modules together is easy; doing all the work (even the basics past stock knobs etc) to make something useful is where it gets boring. Then when you post it, you realize that no one cares unless it is called "Pro Proful Uselessizer Pro XVII Uber-Pro Version". Replete with a GUI cloning some piece of gear that most people never heard of (until you told them how rare it was).

One issue that is really common is people working from the inside out. As in, they have a clever idea for a trick. Code that. But have never really thought about how it can be used broadly. They know the single application they have, usually the same darn application as everything else like splattifing drumz, but have not thought about how this could/should be presented for a broader market. This is where older gear often won because while it may excel in one space, it had to work broadly to cover many use cases - therefore buyers - as possible. If a studio was buying a compressor, they only had money for one, so the idea of buying one that only worked for Kermit The Frog doing Sinatra covers was insane when every week they had Rolling Stones, ABBA, Kraftwerk...

Most of mine get posted and that is were they kinda die, because they are not what most want. Largely becuase they are not named something Pro-izer, with a GUI clone of a Rolamaha faceplate, and the real kicker is they need to learn how it works but won't watch the Tut coz my name doesn't end in Dre.

https://bedroomproducersblog.com/2024/0 ... r-windows/

[ToneCarver]

I get ideas for plugins and often start coding them up .. then realize that I am spending an inordinate amount of time doing the coding, refining, UI design (which is my weakest skill) and it just becomes work .. not so fun any more. So the project sits for a while until the motivation returns, and if the motivation does not return or returns but for something else, so be it .. the plugin under development stays on the workbench in a beta/unfinished/not-ready-for-public-consumption state until such time that the motivation returns.

[knob_monster]

it usually dies at the gatekeeper tax because spending a week fighting apple notarization and paying hundreds for windows code-signing certificates to distribute a hobby project is pure psychological warfare

[camsr]

UI design and implementation

An example problem: how to make a circular or n-gon (particularily a convex one) clip region to capture mouse clicks, for a clickable UI element, with other just as complex clickable elements directly adjacent. It is not impossible, but without an automated process to set all the code, or at least the values, demands time because of the complexity. So when you change your mind on the design, good luck

[BertKoor]

They say you can recogise 'agentic' text by the usage of em-dashes ...
Or it's just people using word processors instead of the browsers memo field.

[circuit-hazw]

They say you can recogise 'agentic' text by the usage of em-dashes ...
Or it's just people using word processors instead of the browsers memo field.

I am a real boy.

[mystran]

UI design and implementation

An example problem: how to make a circular or n-gon (particularily a convex one) clip region to capture mouse clicks, for a clickable UI element, with other just as complex clickable elements directly adjacent. It is not impossible, but without an automated process to set all the code, or at least the values, demands time because of the complexity. So when you change your mind on the design, good luck

Both of these are very easy.

Circular test is (x-xc)^2 + (y-yc)^2 < r^2 and for a convex polygon check sign of the signed distance to each line (no need to treat them as line-segments in the convex case) and just bail out as soon as any test says "outside" (and if all succeed, it's a hit).

About 90% of the time though, you'd use either circles or a few axis-aligned rectangles. Axis-aligned rectangles turn the problem of "signed distance" (which isn't difficult, it's just a homogeneous dot-product) into a comparison of a single coordinate.

I guess this is an example of a case where writing GUI code is just another thing that you have to learn. Between event-driven programming and a bit of geometry, it might seem intimidating at first, but when you learn to think in the right way, it's actually fairly straight-forward and the challenge is mostly just figuring out how to reduce the amount of boilerplate.

[camsr]

Both of these are very easy.

Yes and thanks for explaining.
Vector graphics is going to be the norm moving forward. The days of small bitmap plugins are essentially over. It's not something I know a great deal about.

[mystran]

Both of these are very easy.

Yes and thanks for explaining.
Vector graphics is going to be the norm moving forward. The days of small bitmap plugins are essentially over. It's not something I know a great deal about.

Hit testing and graphics don't usually need to have anything to do with each other.

If you're thinking you need "pixel perfect" hit testing, think again. Hit testing (just like collisions in about 99% of games) can be quite approximate. Even for text selection you would usually want to work with glyph metric rather than the actual graphics.

[kerfuffle]

One mistake I've made is to set myself a deadline for finishing the project. Of course things always take longer than you expect, so I decide to drop a few minor features. As the deadline gets closer and closer, I drop ever more stuff (and move the deadline). At some point I'm completely fed up with the project and end up killing the whole thing (or release something I hate).

After this happened a few times, I realized that the reason I start hating the project is that it was no longer what I wanted to make. Dropping features and making compromises turned this from a cool idea into a meh idea, and I don't want to publish meh products. Yet I'm forcing myself to work on it, and so I despise what I'm doing.

The only way to deal with this is to drop the idea of a deadline (even though it might make business sense to release at a certain time of year etc) and don't compromise on the vision for the thing you're building. Give it the time it needs.