A ~$50 Raspberry Pi turns any HDMI screen into a remote-controlled photo frame. A ~$5–20/month Linux VM runs the website that controls it.
Give one to your parents. Point it at a collection of family photos. When you add new pictures, every frame updates itself. When something goes wrong at their house, you diagnose it from yours.
▶ Live demo: myslideshow.craig0409.com/slideshow/?delay=4000&fade=800 — exactly what a frame displays (press F11 for the full effect). The player is steered entirely by URL parameters, so "configuring" a frame just means choosing its URL:
| Try it | What it shows |
|---|---|
/slideshow/ |
every image, shuffled, calm default pace (12 s) |
?group=Family Moments |
a single named collection |
?group=Scenery,Family Moments |
two collections blended |
?delay=4000&fade=800 |
brisk demo pacing (4 s slides, 0.8 s crossfade) |
?group=Sunsets & Water&delay=20000&fade=3000 |
gallery mode: long holds, slow dissolves |
?img=thumb&shuffle=0 |
bandwidth-light thumbnails, fixed order |
┌──────────────┐ heartbeat every 45s ┌───────────────────────────┐
│ Raspberry Pi │ ──────────────────────► │ your Linux VM │
│ (Chromium │ ◄────────────────────── │ ASP.NET Core + SQLite │
│ kiosk) │ URL + commands │ behind Caddy (HTTPS) │
└──────┬───────┘ └────────────┬──────────────┘
│ HDMI │
any TV or web UI: image library,
monitor device admin, user view
- The Pi boots straight into a fullscreen browser showing a slideshow.
- The server hosts the images (uploaded and auto-resized to 1080p, or referenced by URL), grouped into named collections.
- Each frame can show any collection — switched remotely from a web page, including by non-technical "owners" you assign per device.
- Devices report health (memory, temperature, Wi-Fi signal, power quality) with every heartbeat, so you can support a frame you can't SSH into.
- Wi-Fi onboarding is a phone-friendly captive portal: the frame broadcasts its own hotspot when it can't find a known network. No keyboard needed, ever.
This repo is both the source code and a step-by-step guide aimed at a developer who is comfortable on Windows but has never operated a Linux server. Every Linux command you need is written out; nothing assumes prior experience.
Read and do these in order. Each step ends with something working.
| Step | Guide | You end up with |
|---|---|---|
| 1 | Buy the hardware | ~$50 of parts on the way |
| 2 | Rent and learn your Linux VM | a server you can SSH into, and enough Linux to be dangerous |
| 3 | Deploy the server app | https://slideshow.your-domain.com live, with HTTPS |
| 4 | Build the first frame | a Pi that boots into your slideshow |
| 5 | Clone more frames | as many frames as you have Pis |
| 6 | Operate the fleet | day-2 skills: health, updates, troubleshooting |
| 7 | Reuse the VM for your next idea | a repeatable recipe for hosting any .NET app |
| Thing | One-time | Monthly |
|---|---|---|
| Raspberry Pi 3 Model A+, PSU, SD card, HDMI cable | ~$50/frame | — |
| Hetzner Cloud VM (CX22/CPX11 class) | — | ~$5–8 |
| Domain name (any registrar; DNS at Cloudflare, free plan) | — | ~$1 amortized |
The VM is wildly oversized for this app alone — the whole point of step 7 is that the same box hosts your next several ideas at no extra cost.
| Path | What |
|---|---|
server/ |
ASP.NET Core app: image library + device hub + slideshow player |
device/ |
Everything installed on a Pi (scripts, agent, systemd unit, installers) |
deploy/deploy-vm.ps1 |
One-command deploy from your Windows machine to the VM |
docs/ |
The guide |
This design was extracted from a working four-frame deployment. The non-obvious decisions, all learned the hard way, are documented inline where they live, and summarized in docs/06-operate.md:
- Wayland (
cage) with CPU rendering, not X11 — on Raspberry Pi OS Trixie, the Pi 3-era GPU scans out black frames for tiled GPU buffers while everything looks healthy internally. --no-memcheckbefore the URL — Pi OS's Chromium wrapper otherwise blocks forever on an invisible "low RAM" dialog on 512 MB boards.- Chromium singleton lock cleared each launch — a cloned SD card otherwise refuses to start the browser ("profile in use on another computer"), and so does one that lost power at the wrong moment.
- Server-side 1080p variants — a 512 MB board decodes 1080p images happily and full-resolution photos not at all; periodic browser reloads shed the memory creep that image-heavy pages accumulate.
- Write-minimized SD card — logs to RAM, browser cache to RAM, batched filesystem commits, no background apt: people will pull the plug.
MIT — see LICENSE.
If you appreciate this project, send me a tip! Click a QR (or scan it from your phone):
| PayPal | Cash App | Venmo |
|---|---|---|
 |
 |
 |
| paypal.me/Craig0409 | cash.app/$Craig0409 | venmo.com/Craig0409vt |