k-clock

A voice-controlled LED clock running on a Raspberry Pi. Say "hey clock" to wake it, then speak a command to change patterns, show the moon phase, or toggle the clock display. LED animations are driven by a Pixelblaze controller over the local network.

Hardware

• Raspberry Pi (any model with audio input)
• Pixelblaze LED controller named k-clock on the local network
• USB microphone or audio input device

How it works

1. The Pi continuously listens for the wake word "hey clock" (via Picovoice Porcupine)
2. On detection, the active pattern switches to C wake word to signal the clock is listening
3. You have 5 seconds to speak a command, which is parsed by Picovoice Rhino
4. The command controls the Pixelblaze over WebSocket
5. Every 15 minutes, brightness is adjusted automatically on a cosine curve (dim at night, brighter during the day)

Voice commands

Phrase	Action
"set pattern to moon phase"	Fetch today's moon phase, display it for 5s, then return
"set pattern to clock on"	Switch sequencer to clock-on playlist (C patterns)
"set pattern to clock off"	Switch sequencer to clock-off playlist (C- patterns)

Pixelblaze pattern naming convention

Patterns come in pairs:

• C <name> — clock-on variant (clock overlay visible)
• C- <name> — clock-off variant (background only)

On startup, the sequencer playlist is built from the C variants. The "clock off" command swaps to the paired C- variants without leaving the sequencer.

Setup

Dependencies

Requires Python 3.11+ and a Picovoice access key.

pip install -r requirements.txt

Picovoice models

Two model files are required (not committed — generate from the Picovoice Console):

• hey-clock_en_raspberry-pi_v4_0_0.ppn — Porcupine wake word model
• k-clock-pattern_en_raspberry-pi_v4_0_0.rhn — Rhino intent context model

Running directly

python3 k-clock.py \
  --access_key <PICOVOICE_KEY> \
  --keyword_paths hey-clock_en_raspberry-pi_v4_0_0.ppn \
  --context_path k-clock-pattern_en_raspberry-pi_v4_0_0.rhn

Useful flags:

Flag	Description
--show_audio_devices	List available audio input devices
--show_inference_devices	List available Porcupine inference devices
--audio_device_index <n>	Select audio device (default: -1)
--sensitivities <float>	Wake word sensitivity 0.0–1.0 (default: 0.5)

Running as a systemd service

sudo systemctl start k-clock
sudo systemctl stop k-clock
sudo systemctl status k-clock
journalctl -u k-clock -f

The service unit is defined in k-clock_service.service. It calls startup.sh, which activates the virtualenv and passes the access key and model paths.

Files

k-clock.py               # Entire application (single file)
startup.sh               # Launch script for the Pi (activates venv, passes args)
k-clock_service.service  # systemd unit file
requirements.txt         # Python dependencies

Logs

Rotating log file at k-clock.log in the working directory (40 KB, 5 backups).