Part 4

First-Time Setup

Both editions greet you with a setup wizard the first time you start them. This Part walks through it: pointing the app at your camera, setting your location, configuring notifications, and (optionally) joining the community feed. By the end you'll have a connected camera and your first detection rolling in.

Preloaded Raspberry Pi — start here

If you received a Raspberry Pi with BirdWatchAI already installed (a ready-to-run bundle rather than a bare Pi you imaged yourself), you can skip Part 3 — Installation entirely. The software is already on the SD card and set to start automatically every time the Pi powers on. All that's left is to give it power and a network, open the dashboard, and run the wizard below.

Not sure which path is yours? If you flashed the SD card yourself with Raspberry Pi Imager, follow Part 3 → Raspberry Pi server (from a blank SD card) instead. If the Pi arrived ready to plug in, you're in the right place — the four steps below are all you need.

1. Get your camera ready

BirdWatchAI watches an RTSP camera over your network, so have the camera powered on and joined to the same Wi-Fi / LAN as the Pi before you start. For TP-Link Tapo cameras, first open the Tapo app and set a Camera Account under Advanced Settings → Camera Account — that username and password are what the wizard's RTSP URL uses, and the camera won't accept an RTSP connection until you've created one. Full per-brand details are in Part 2 — Camera Hardware.

2. Power on the Pi

Connect the Pi to your network — plug in the Ethernet cable, or trust the Wi-Fi it was preconfigured with (check the insert or sticker that came with it). Wired is more reliable for an always-on camera.
Plug in the official USB-C power supply. Underpowered supplies cause random lockups under camera + AI load, so use the one provided.
Wait about 1–2 minutes for it to boot. BirdWatchAI starts on its own — there is nothing to launch by hand.

3. Open the dashboard

From any browser on the same network, go to:

http://birdwatch.local:8080

(Use whatever hostname is printed on your unit if it isn't birdwatch.)

Can't reach it, or don't know the Pi's address? The Pi is headless — there's no screen to read an address off. Two fallbacks: look for the hostname in your router's admin page to get its IP and browse to http://<pi-ip>:8080, or download BirdWatchFinder.exe, a double-click Windows tool that scans your network, finds every BirdWatchAI server on it, and hands you a clickable link to the dashboard. (It also scans for feeder cameras and prints their IPs — handy for the RTSP URL in the next step.)

4. Run first-time setup

The first time you open it, the dashboard shows a "👋 First-time setup" card pointing you at /setup. Click Start setup and follow The setup wizard below — from this point the experience is identical to every other server install.

It stays running for you The BirdWatchAI container is set to restart automatically, so it comes back up on its own after a reboot or power cut — you never have to start it manually. Your detection history and settings live on the Pi under ~/birdwatch/data/; back that folder up if the data matters to you.

The setup wizard

The very first time you run BirdWatchAI, you'll see a splash screen (desktop) or "👋 First-time setup" card (server) directing you to the wizard. Both wizards cover the same ground in slightly different UIs.

On the desktop, the wizard appears automatically after the splash. You can press Skip on most steps and configure things later in Settings, and you can re-run the wizard any time from the Help menu.

On the server, the dashboard at http://<your-host>:8080 shows a "First-time setup" card on the home page. Click Start setup to go to /setup.

The five steps

Welcome — quick orientation; lists what you'll need (camera, Wi-Fi, optional email + ZIP code).
Camera — RTSP URL, or for the server edition you can pick a wired Pi camera instead.
Location — ZIP code (used for outdoor temperature and regional rarity).
Notifications — optional email + push.
License — optional; a free trial starts automatically if you skip.

Camera step

This is the only step that really matters on first run — without a working camera the app has nothing to watch.

On the desktop wizard

Pick your camera type from the dropdown (Tapo, Wyze, Amcrest, Reolink, Hikvision, or Other). Tabbed instructions adapt to your choice. Then fill in:

Camera IP Address — use the How to Find IP button if unsure, or see Part 2 → Find the camera's IP address.
RTSP Port — default 554.
RTSP Username / RTSP Password — the camera account credentials (for Tapo, this is separate from your TP-Link login — see Part 2 → TP-Link Tapo).

Click Build URL to assemble the RTSP address, then 🔗 Test Connection to verify. A successful test shows the camera resolution. There's also a 📖 Full TAPO Guide button.

On the server wizard

Paste the camera's RTSP URL straight in. For TP-Link Tapo cameras, the form is:

rtsp://<username>:<password>@<camera-ip>/stream2

For example: rtsp://birduser:secret123@192.168.1.105/stream2.

Use stream2 on Tapo Use the sub-stream (stream2) on every Tapo model — it's 720p, reliable, and easy on the camera's Wi-Fi. Avoid the main stream (stream1), especially at 2K: it saturates the camera's Wi-Fi uplink and causes H.264 corruption and instability. See Part 2 → TP-Link Tapo for details.

Click Test camera. A frame should come back in a few seconds. If not, see Part 7 → Camera won't connect — the usual culprits are a typo in the URL, the wrong credentials (camera account vs. cloud account), or the wrong stream path.

If your camera is a wired Pi camera (server only)

Set the Camera type to Pi camera instead of RTSP. Leave the device path at 0 (the libcamera index — the field also accepts /dev/video0). One-time host configuration is required first: see Part 2 → Raspberry Pi camera.

Location & weather

Your location powers three things:

Temperature — fetched from Open-Meteo (no API key required) and stamped on every detection's overlay.
Rarity ratings — each species is rated against regional / seasonal eBird data so a Wood Thrush in suburban Boston gets flagged differently from one in rural Vermont.
Daylight hours — used to pause monitoring overnight (configurable).

On the desktop: enter your ZIP Code and click Look Up. The wizard resolves it to latitude / longitude. There's also an optional Pick on Map button (needs the WebView2 runtime; if you don't have it you can still enter coordinates by hand).

On the server: type your ZIP code into the wizard's location step.

Outside the US? ZIP-code lookup is US-only, but you can enter latitude / longitude directly on both editions and everything else (weather, rarity) still works.

Notifications

Both editions can send a notification with the snapshot when a detection happens. All channels are off by default — enable the ones you want, and use the Test button on each to confirm it works before you rely on it.

Channel	Desktop	Server	Notes
Email (SMTP)	✅	✅	Gmail needs an App Password, not your regular password.
ntfy push	✅	✅	Free, account-less. Pick a hard-to-guess topic name.
Pushover	✅	✅	Mobile push with the photo; small one-time fee per platform.
Windows toast	✅	—	Native Windows 10/11 notifications; instant and local.
Photo frame (FTP or email-to-frame)	✅	—	Push snapshots straight to a digital photo frame.

You can skip this entire step in the wizard and add channels later under Settings → Notifications. The full reference is in Part 5 → Notification channels.

Community sharing

The BirdWatchAI community is a free, worldwide live feed of sightings at www.birdwatchai.com. Sharing is off by default — turn it on if you'd like your feeder to show up on the community map, in the gallery, and on the leaderboards.

What you share is controlled differently on the two editions, so check the one you are running.

On the server

Three checkboxes on Settings → Community: the master switch (off to begin with), Also upload the snapshot image, and Also upload the recorded video clip. The latter two are already ticked, so turning the master switch on publishes your photographs and clips as well as the sighting itself. Untick them first if you would rather share sightings without imagery.

Your location goes with it. If you set a ZIP code or coordinates — which the temperature lookup needs — that is attached to your feeder and places it on the public map. A ZIP lookup shares an area; a street address shares something much closer to your house. There is no setting that shares sightings while withholding a location you have set.

On the desktop

You pick a single Share Level (0–4) that governs what leaves your home:

Level	What is shared
0 — Off	Nothing
1	Species + timestamp (plus confidence, rarity, temperature when available)
2	The above + ZIP code (or GPS coordinates if you've enabled GPS instead of ZIP)
3	The above + the snapshot image
4	The above + the video clip

You stay anonymous on both Your feeder is identified only by a random Device Key (a UUID) — never linked to your name or email. Your public identity is just the Feeder Display Name. On the desktop, location is shared only at Level 2 and above; on the server there is no such tier.

You can change any of this later. The full reference is in Part 5 → Community sharing, and what we do with what you share is in the privacy policy.

Your first detection

When you finish the wizard, you land on the main window / dashboard. The Engine status card (server) or the top status bar (desktop) should show monitoring Running, camera Connected, and within a few minutes — assuming there are birds at the feeder and it's daylight — your first detection.

If it's been a while and nothing has shown up:

Confirm the camera frame in the dashboard / main window actually shows the feeder.
Check that monitoring is running, not paused.
Check that it's within daylight hours (the engine pauses overnight by default, 7 AM – 8 PM).
Try the Test Image button (desktop) with a known bird photo to confirm the AI side of the pipeline is working.

For deeper diagnosis see Part 7 → No / wrong detections.