A tiny computer in my backyard is eavesdropping on birds. 24/7. They have no idea.

Real-time acoustic bird classification running on a Raspberry Pi 3B+, listening to the birds of Durham, North Carolina.

This is a modernized fork of BirdNET-Pi, built around a modern web stack for performance, maintainability, and a much better day-to-day operating experience.

Want to see what's singing right now? → durm.pibirds.org

┌─────────────────────────────────────────────────────────────────┐
│                        BirdNET-Pibird                           │
├─────────────────────────────────────────────────────────────────┤
│  Frontend (SvelteKit + Tailwind CSS)                            │
│  • Modern, responsive UI with dark mode                         │
│  • Real-time detection feed                                     │
│  • Mobile-first design                                          │
├─────────────────────────────────────────────────────────────────┤
│  Backend (FastAPI + Python)                                     │
│  • RESTful API for all operations                               │
│  • Reuses existing BirdNET Python utilities                     │
│  • SQLite database for detections                               │
├─────────────────────────────────────────────────────────────────┤
│  Analysis Pipeline (unchanged from upstream)                    │
│  • BirdNET TensorFlow Lite model                                │
│  • Audio recording & spectrogram generation                     │
│  • BirdWeather integration                                      │
└─────────────────────────────────────────────────────────────────┘

The modern web app covers the main day-to-day and admin workflows:

• Dashboard with current detections, live audio access, and public status
• Review feed for recent detections and moderation actions
• Library for historical recordings, spectrogram inspection, frequency-shifted clips, and Temporal Zoom playback
• Species and Insights views for trends, charts, date-filtered species lists, and exports
• Weekly Report at /reports/weekly
• Settings, Advanced Settings, and System administration
• Live Logs for service log inspection
• Admin-only File Manager at /files, linked from Library

Intentional exclusions from earlier admin tooling:

• No Web Terminal
• No Adminer

The File Manager is intentionally scoped to BirdNET-owned directories rather than acting as a general server browser.

Recording players include Temporal Zoom presets (1.0x, 0.85x, 0.7x, 0.6x, 0.5x) that slow playback while preserving pitch. The feature is framed as a listening aid for humans, not as a simulation of another animal's hearing. On mobile browsers, Pibird uses a lighter native playback path and can prepare cached tempo-rendered clips in the background to avoid render-on-tap delays over Cloudflare Tunnel or other remote access paths.

On a fresh Raspberry Pi with 64-bit RaspiOS:

curl -s https://raw.githubusercontent.com/cpieper/BirdNET-Pibird/main/newinstaller.sh | bash

You can also record an update preference during install:

RELEASE_CHANNEL=stable bash newinstaller.sh

Supported channels are stable, prerelease, and edge.

• stable is the default and installs the latest stable release tag.
• prerelease installs the latest prerelease or release tag, but requires explicit opt-in.
• edge follows the selected BRANCH, and also requires explicit opt-in.

This installs everything: BirdNET analysis pipeline, the new web interface, and all services.

# Replace BRANCH_NAME with the actual branch name
curl -s https://raw.githubusercontent.com/cpieper/BirdNET-Pibird/main/newinstaller.sh | RELEASE_CHANNEL=edge BRANCH=BRANCH_NAME bash

curl -s https://raw.githubusercontent.com/cpieper/BirdNET-Pibird/main/newinstaller.sh | RELEASE_CHANNEL=edge BRANCH=fastapi-svelte-migration-mk1 bash

REPO_URL=https://github.com/yourfork/BirdNET-Pibird.git RELEASE_CHANNEL=edge BRANCH=your-branch bash newinstaller.sh

If you already have BirdNET-Pibird checked out on a system and want to refresh the installed services and web app:

cd ~/BirdNET-Pi
git pull
./scripts/install_web.sh

The installer script will:

• Install Node.js and build the new frontend
• Install FastAPI backend dependencies
• Reconfigure Caddy for the new architecture
• Start the new web service

Your detection data, configuration, and recordings are preserved.

• Hardware: Raspberry Pi 5/4B/400/3B+/Zero 2W
• OS: 64-bit Raspberry Pi OS (Bookworm recommended)
• Microphone: USB microphone or RTSP stream

For local development on macOS, use Python 3.11 for the backend environment. TensorFlow does not currently provide a compatible install path for the newer Homebrew python3.14 toolchain, so backend setup and tests are expected to run from a 3.11 virtual environment.

python3.11 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt
cd backend
python -m pip install -r requirements.txt
uvicorn app.main:app --reload --port 8080

Backend tests currently need both the BirdNET runtime dependencies from the repository root and the FastAPI backend dependencies from backend/:

python3.11 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt
python -m pip install -r backend/requirements.txt
python -m pip uninstall -y pyarrow
pytest tests -q

On Apple Silicon macOS, uninstalling pyarrow in the local development venv avoids a TensorFlow import crash during test collection. This is only a local test-environment workaround and is not required on the target Pi install path.

cd frontend
npm install
npm run dev

The frontend dev server proxies API requests to localhost:8080.

• versions.md stores machine-readable release metadata used by the API:
  • service_version (SemVer)
  • git_hash (commit hash)
  • git_branch
  • api_version
  • build_date_utc
• version.md remains the human-readable changelog/history.

Update metadata before a release/build:

./scripts/update_version_metadata.sh --service-version 0.14.0

Use the guided release script to keep ordering correct (metadata -> commit -> tag -> push):

./scripts/release_flow.sh --tag v0.14.0-rc6

BirdNET-Pibird/
├── backend/                 # FastAPI application
│   ├── app/
│   │   ├── main.py         # Application entry point
│   │   ├── config.py       # Configuration management
│   │   ├── dependencies.py # Auth, database connections
│   │   ├── routers/        # API endpoints
│   │   └── models/         # Pydantic schemas
│   └── requirements.txt
├── frontend/                # SvelteKit application
│   ├── src/
│   │   ├── routes/         # Page components
│   │   ├── lib/            # Shared code, components, stores
│   │   └── app.css         # Tailwind entry point
│   ├── tailwind.config.js
│   └── package.json
├── scripts/                 # Analysis & utility scripts (unchanged)
│   ├── birdnet_analysis.py # Main analysis pipeline
│   ├── utils/              # Python utilities (reused by backend)
│   └── *.sh                # Shell scripts
├── model/                   # BirdNET models & labels
├── versions.md              # Machine-readable release metadata
└── version.md               # Human-readable changelog

This project builds upon the incredible work of:

• Wiki — Full documentation & troubleshooting
• Web UI Guide — Settings surfaces, reports, file management, live audio, and tunnel guidance
• External Integrations Notes — Bird image provider behavior, local asset caching, and Wikimedia safeguards
• Discussions — Community Q&A
• BirdWeather — Share your birds with the world

Currently listening in Durham, NC... 🎤🐦