This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
Most AI products are designed to keep us on our screens.
HearWild is designed to do the opposite.
HearWild is an open-source AI nature companion that lets you record a bird sound, identifies the bird using local AI, tells you a little about what you heard, and encourages you to go back outside.
The idea is simple:
Listen. Discover. Learn. Go outside.
You don't need to take a photo or know what bird you're looking at. Just point your microphone toward the sound, record for a few seconds, and let the AI listen.
When a bird is confidently identified, it is automatically added to your Flock and you earn XP:
- 🐦 New bird → +100 XP
- 🔁 Bird you've already discovered → +25 XP
- ⭐ Discover more birds → level up
The important part is what happens next.
HearWild doesn't try to keep you inside the app.
Once you've discovered the bird, the goal is to put the screen down and keep exploring.
It's built for anyone who's curious about nature, especially people who don't know much about birds but want a reason to step outside and pay attention to what's around them.
Demo
The MVP currently runs locally:
Record a bird sound ↓ BirdNET V3 identifies it ↓ See the bird + confidence ↓ Bird is added to your Flock ↓ Earn XP ↓ 🌿 Keep Exploring Code
The complete source code is available on GitHub:
How I Built It
The core of HearWild is BirdNET V3, running locally through its ONNX model.
I wanted the AI itself to be part of the open-source stack rather than simply calling a closed AI API.
🧠 The AI
HearWild uses:
BirdNET V3 → ONNX → local inference
The audio recorded by the browser is sent to a local FastAPI backend, converted to the required format using FFmpeg, and passed to BirdNET V3.
The model returns the predicted species and confidence.
I then use confidence-aware behaviour:
| Confidence | HearWild's response |
|---|---|
| 70%+ | Confident identification → add to Flock + earn XP |
| 20–69% | Possible identification → show result, but don't add |
| <20% | Unknown → don't pretend to know |
That last part was important to me.
If the model isn't confident, HearWild doesn't pretend it knows the answer.
🏗️ Architecture
┌──────────────────────────┐ │ Vue 3 + Vite │ │ │ │ 🎙️ Record bird sound │ └────────────┬─────────────┘ │ ▼ ┌──────────────────────────┐ │ FastAPI │ │ │ │ Audio processing │ └────────────┬─────────────┘ │ ▼ ┌──────────────────────────┐ │ FFmpeg │ │ Audio conversion │ └────────────┬─────────────┘ │ ▼ ┌──────────────────────────┐ │ BirdNET V3 ONNX │ │ │ │ Local bird identification│ └────────────┬─────────────┘ │ ▼ ┌──────────────────────────┐ │ SQLite │ │ │ │ 🐦 My Flock │ └──────────────────────────┘ The frontend is built with Vue 3 + Vite and the backend uses Python + FastAPI.
SQLite keeps track of the birds I've discovered, how many times I've heard them, and my XP progression.
For bird imagery and short descriptions, HearWild uses Wikimedia when available.
Why Does Open Innovation Matter?
This project could have been built by calling a commercial bird-identification API.
But that would have missed one of the most interesting parts of the challenge for me.
I wanted the AI itself to be something I could run and understand locally.
Using an open model made it possible to:
- Run bird identification locally
- Avoid sending bird recordings to a proprietary AI service
- Experiment with the model directly
- Control the inference pipeline
- Build without depending on an AI API subscription
- Keep the project reproducible for other developers
More importantly, it changed the way I thought about the product.
The AI doesn't need to become the destination.
It can simply be the bridge between technology and the real world.
That's what I like about open innovation here: the model becomes a building block that anyone can experiment with and build something completely different around.
🌿 What I Learned
The biggest lesson from building HearWild was that an AI project doesn't necessarily need to give users more to look at.
It can help them notice what's already around them.
Birdsong is happening all the time.
Most of us just don't stop to listen.
So instead of building another AI assistant that says:
"Here are 10 things you can do on your screen..."
I wanted to build one that says:
"You heard a bird. Now go find it."
That's HearWild.
Listen to the world. Discover what lives in it.
