I love switching wallpapers a lot, the feeling of intention when I look at my desktop is something I yearn for and like many, I often get my images through Pinterest, but this led to a problem.
See, Pinterest is not really an image host. What you're seeing in your feed is almost always a thumbnail; a compressed resize of the original. When you try to save it, that's what you get. The actual image lives somewhere else, could be on a different site or maybe even gone entirely.
The technical reason: Pinterest crawls the web, indexes images from other sources and serves them fast at scale, they generate their own cached versions, smaller and optimized for feed browsing, not for your desktop. So that beautiful pin you saved? is just Pinterest's copy, not the original.
I figured this out the hard way, many beautiful looking visuals, but when that "set as wallpaper" button is clicked, it looks blurred, a bit off.
App's right here if you'd like to take a look: Poster
One would say to use other platforms, get 4K res, but not many have been able to capture the same level of aesthetic comfort that Pinterest has. My solution was simple, build an app to help. I mean, how hard could that be right?
The Journey
I did a bit of brainstorming, things I would love to have on a wallpaper application, asked a few friends, even downloaded some competitor apps to see what they had going on, and finally arrived at what I thought was a realistic scope.
Pt. 1: Expectations
I envisioned something a bit straightforward, something that could do these:
- Connect to my account, use and access my boards/pins
- AI Upscale the images that make me cringe
- Act as a central library for all my wallpapers
- Implement Unsplash API cause Unsplash is GOATED
but as we all know, "straightforward" almost never happens in our line of work.
So, I got to work.
My package manager of choice? pnpm
My building framework? electron
If you have experience here, you already know that those two don't mix well together, and that was the start of what I thought would be a disaster. More on this later
Pt. 2: Building
I ran the pnpm command, created the project with React within Electron, forgot to configure tailwind but styling wasn't the issue, the real headache was figuring out how Electron actually talks to the OS.
The first problem was understanding how to access user's files since the user would be uploading images they wanna use (at least at the start).
This led to me having to how to work localfile:// alongside file paths on the system, except, whenever it was time to display the files, it never did!
Fixed the problem by having to take out a couple of leading slashes. The full protocol handler ended up looking like this, and the fix was embarrassingly simple:
js// before
//C:/Users/nuel/Pictures/wallpaper.jpg ❌
// after
C:/Users/nuel/Pictures/wallpaper.jpg ✅
pathName = pathName.replace(/^\/+/, "");
And if it wasn't a silent pathing bug, it was something else. The weeks that followed felt like my codebase was yelling at me from every angle.
The pnpm feud: had to switch to npm entirely, this was after I tried fighting a losing battle but ended up realizing that the two are just not compatible.
Native dependencies: packages like sharp have opinions about your environment. Getting them to build correctly inside Electron took more time than I'd like to admit especially with pnpm.
The AI wall: the plan was server-side upscaling via Replicate, billing setup got in the way. Pivoted to local AI upscaling instead. Not what I planned, but keeping it on-device means no API costs passed to the user, so maybe it worked out.
The Unsplash block: turns out Unsplash explicitly restricts their API for wallpaper applications. That whole integration plan was dead on arrival. Painful but still Goated
Auto Updater Setup: Maybe the biggest hassle of them all, config was totally alien for an electron noob like me but I figured it all out in the end.
Every feature was a burden to implement, with each one passed came another, but slowly, the pieces started sticking together.
Pt. 3: The Reality
What Did Poster Actually Become?
Well, not everything I planned made it. Pinterest integration? Still figuring it out. Cloud storage? Probably coming when accounts become a thing. Replicate upscaling? I want it, just need to make sure the app can foot the bill first.
But what's there works, and I'm proud of it.
After the first few versions I realized the performance wasn't really great, especially if one had a multitude of images, the scrolling got choppy. A wallpaper library means you're potentially throwing hundreds of images at the screen at once, so Poster renders thumbnails instead of full images, and anything outside the viewport just stops existing until you scroll back to it. Occlusion culling. Keeps things light and here's where things stand:
Free
- Local image library
- Slideshow mode
Poster Plus
- Time-of-day scheduling, morning, afternoon, evening, night. Different wallpaper for each.
- Local AI upscaling, low-res Pinterest save? Run it through before setting it. The difference is real.
- Boards (coming soon)
To wrap it up
Poster itself is very different from what I initially had in mind. Half the features got cut, one got blocked by an API policy, and a few turned into something else entirely.
But it's out. It works. And I use it every day, which was the whole point.
If you're on Windows and would love to try it, give it a download. It's Free with the Plus tier as a lifetime payment. If something feels off or broken, tell me. I'm still very much building this thing.
Yours truly,
nueldotdev💫
Top comments (0)