Pokémon Automation

Closed-loop automation scripts for Pokémon FireRed/LeafGreen using a Raspberry Pi Pico H, a Raspberry Pi Debug Probe, a USB capture card, Open Broadcasting Software, and Python/OpenCV for visual detection.

Projects

shiny_hunter.py

Automates soft-reset shiny hunting. Detects the yellow star indicator via HSV color detection and stops the macro automatically when a shiny is found. Optionally saves an OBS replay buffer clip on detection.

nugget_bridge.py

Automates the Nugget Bridge trainer loop in FireRed/LeafGreen. Walks the route, fights the Rattata trainer, and loops back. Uses purple pixel detection to know when the battle ends.

Hardware Required

• Raspberry Pi Pico H (flashed with the controller firmware)
• Raspberry Pi Debug Probe
• 2 USB Cables
• USB capture card (Elgato HD60 X or Magewell USB Capture HDMI Gen 2)
• Switch running FireRed/LeafGreen via capture card

Software Dependencies

• Open Broadcaster Software
• Python
• Arduino IDE

pip install opencv-python numpy pyserial obsws-python

Setup

1. Flash your Raspberry Pi Pico H with the controller firmware
2. Run tools/find_devices.py to find your capture card index
3. Open the script you want to use and set COM_PORT to your Pico H's port
   • Windows: "COM5" (check Device Manager)
   • Mac/Linux: "/dev/ttyACM0"
4. Set CAPTURE_INDEX to the number from step 2, or leave it as None to auto-detect
5. Run the script and type STARTMACRO to begin

Tools

Script	Purpose
tools/find_devices.py	Lists all capture card indices and resolutions
tools/check_res.py	Checks what resolution each capture index reports
tools/find_coords.py	Click on a live frame to get pixel coordinates
tools/star_test.py	Live preview of shiny star detection region and mask

Commands (while running)

Command	Effect
STARTMACRO	Begin the macro
STOPMACRO	Stop the macro (program stays open)
EXIT	Close the program
TESTDETECT	Run detection on the current screen (nugget_bridge.py)
TESTMACRO	Run detection on the current screen (shiny_hunter.py)
STARTWALK	Skip straight to the walk sequence (nugget_bridge.py)
A B X Y UP DOWN LEFT RIGHT	Send a single button press manually

Notes

• Log files are saved automatically in the same folder as the script
• shiny_hunter.py requires OBS to be open with WebSocket and Replay Buffer enabled for clip saving — it works without OBS but won't save clips
• Detection regions are tuned for a 1920x1080 capture feed — if your resolution differs, use tools/find_coords.py to recalibrate