Pamk
An open-source desktop robot project based on the ESP32-S3 microcontroller. It supports custom wake-word local voice activation and integrates Gemini for natural language understanding. The device provides productivity features including task management, schedule reminders, and a Pomodoro timer, and is equipped with a 3D-printed interchangeable front panel.
Timeline (last 90 days)
Pamk is a desktop robot built on the ESP32-S3 that can understand voice commands, manage tasks and schedules, and proactively send reminders.
Pamk is activated by the custom wake word "Hey Pamk" with no phone app involvement required throughout the process.
Pamk has a built-in Pomodoro mode, with a progress ring displayed around the face on the screen to reflect the remaining time.
Pamk's face panel is 3D printed and secured magnetically, allowing it to be swapped out at any time.
Pamk's main control chip is Espressif's ESP32-S3, and the display uses a 280x240 resolution screen.
Pamk's audio input is handled by an INMP441 I2S digital microphone, and output drives the speaker through a MAX98357 I2S amplifier.
Pamk integrates Google Gemini for natural language understanding, parsing spoken commands into structured tasks and schedules.
The voice in Pamk's responses is assembled from audio clips recorded by the author personally, rather than being synthesized by TTS.
Pamk's author trained a dedicated wake word model using thousands of self-recorded audio clips, rather than using an off-the-shelf framework.
Pamk adopts a hierarchical architecture combining local wake detection and cloud-based understanding: the self-trained wake model runs offline, while Gemini-based understanding requires an internet connection.
1 more timeline events
All Facts (11)
Pamk is a desktop robot built on the ESP32-S3 that can understand voice commands, manage tasks and schedules, and proactively send reminders.
50%UnverifiedPamk is activated by the custom wake word "Hey Pamk" with no phone app involvement required throughout the process.
50%UnverifiedPamk has a built-in Pomodoro mode, with a progress ring displayed around the face on the screen to reflect the remaining time.
50%UnverifiedPamk's face panel is 3D printed and secured magnetically, allowing it to be swapped out at any time.
50%UnverifiedPamk's main control chip is Espressif's ESP32-S3, and the display uses a 280x240 resolution screen.
50%UnverifiedPamk's audio input is handled by an INMP441 I2S digital microphone, and output drives the speaker through a MAX98357 I2S amplifier.
50%UnverifiedPamk integrates Google Gemini for natural language understanding, parsing spoken commands into structured tasks and schedules.
50%UnverifiedThe voice in Pamk's responses is assembled from audio clips recorded by the author personally, rather than being synthesized by TTS.
50%UnverifiedPamk's author trained a dedicated wake word model using thousands of self-recorded audio clips, rather than using an off-the-shelf framework.
50%UnverifiedPamk adopts a hierarchical architecture combining local wake detection and cloud-based understanding: the self-trained wake model runs offline, while Gemini-based understanding requires an internet connection.
50%UnverifiedPamk's V2 version is being designed with a custom PCB, and the author plans to launch a crowdfunding campaign and release build materials after accumulating followers on MakerWorld.
50%