openai/birdingpal

openai/birdingpal

Description: BirdingPal is an ChatGPT-enabled partner for identifying birds.

Language: Kotlin

License: MIT

Stars: 0

Forks: 0

Open issues: 0

Created: 2026-06-25T21:48:47Z

Pushed: 2026-07-09T16:58:07Z

Default branch: main

Fork: no

Archived: no

README:

Birding Pal


Birding Pal is a small field companion for birding. We paired a plush bird with a custom Bluetooth controller and a ChatGPT-based voice interface so you can press the bird, talk about what you see or hear, record a bird call, and save confirmed bird sightings to a personal bird book.

Our goal was to make the interaction feel less like pulling out a phone and more like carrying a tiny friendly AI-enabled buddy outdoors.

What We Built

The system has two main pieces:

• The Hardware: a plush bird with pressure inputs, Bluetooth, status lights,

microphone, speaker, and a microcontroller.

• The Software: a mobile app that powers the conversation, bird-call

recording, optional BirdNET-compatible analysis, and bird book.

The software can run by itself, but the hardware gives the project its intended shape: a tactile companion you can clip to a bag, hold in your hand, and use without staying focused on a screen.

Hardware

The prototype hardware includes:

• Adafruit QT Py ESP32-S3, no PSRAM
• DFRobot BT401 Bluetooth module
• Force-sensitive resistor inputs
• Addressable NeoPixel LED
• LiPo battery and charger/breakout
• Microphone, amplifier, and speaker
• Jumper wires and a plush/enclosure

The firmware reads the pressure inputs, drives the status LED, configures the BT401 module, and hosts a local Wi-Fi dashboard for debugging and threshold tuning.

How It Works

Pressing the plush sends Bluetooth events to the companion app. One action starts or stops a voice session; another records ambient audio for bird-call analysis. The assistant can identify likely species, explain what it heard, and add confirmed sightings to the bird book.

The voice experience uses the OpenAI Realtime API. The optional BirdNET-compatible endpoint receives recorded WAV audio with location and week-of-year metadata, then returns species candidates for the assistant to interpret.

Note: GPT-Live-1 is not yet available on the OpenAI API platform; this repository uses gpt-realtime for the Realtime API integration.

License

This project is licensed under the MIT License. See [LICENSE](LICENSE).