A tool to map audio input to keypresses for the rhythm game Taiko no Tatsujin
DonKa Detector aims to make the fun of Taiko no Tatsujin accessible anywhere, anytime. It gives people more options to play the game without needing specialized equipment.
- DonKa Mapping: Map your audio inputs to left/right Don's and Ka's by following a metronome. Map left/right Don's and Ka's to native keys.
- Performance Options: Similar to how you can reduce graphics options to increase FPS, you can also modify chunk sizes, buffer sizes, and delay for optimal gaming performance.
This tool is a work in progress. For now, the only way to use this is via the command line.
Make sure Python is installed in your machine. I recommend using your preferred Python package manager. This guide uses Anaconda.
Create and activate the environment using
conda create -n DonkaDetector python=3.12
conda activate DonkaDetector
Then install dependencies using
pip install -r python_requirements.txt
Define audio inputs using
python audio_func/record_donka.py
and follow the instructions.
Then, validate the inputs using
python audio_func/validate_input.py
This should give you some information on expected accuracy for specific note rates (in notes/second) and conduct further fine-tuning.
Now, you can run the program using
python main.py
You can modify paths and keybindings in config.ini.
This application uses normalize=False and K=1 for now. The outputs are in the format (total # of predicted notes, DonKa/Side errors, DonKa errors). The third value corresponds to mistaking Dons for Kas and vice versa, and are more fatal. The second value includes a left/right mistake, which is not critical in real-time gameplay. Roughly, the accuracy is (Donka Errors) / (Total # of predicted notes). For further details and caveats on the metrics, check the risk analysis notebooks.