An interactive installation exploring how self-affirmation and spoken compliments unfold as social performances, giving presence a form through mirrored interaction, spoken exchange, and printed memory.
Kind Words Club examines affirmation and compliments as a small act of recognition. A compliment does not define who someone is; it gives form to the moment in which their presence is seen, addressed, and acknowledged.
The installation unfolds through two encounters: a mirrored screen interface where participants say affirmations aloud, followed by a person-to-person exchange where a spoken compliment is recorded.
The printed receipt gives this exchange a physical form. What would normally disappear after being spoken becomes a small artifact of recognition, response, and memory. The system does not measure sincerity or emotional truth. It frames the conditions around being seen, heard, and remembered.
The Interaction
01
Mirrored Interaction
The participant sees themself on screen and repeats affirmations aloud. The interface listens only when a smile is detected, using the smile as a visible social gesture rather than emotional proof.
02
Spoken Exchange
The interaction shifts from self-address to interpersonal acknowledgment. Another person holds down a button while giving a compliment, making the act of speaking intentional, visible, and physically marked.
03
Printed Trace
The recorded compliment is printed as a small receipt for the receiver to keep. A temporary spoken exchange becomes a physical trace of presence: something said, received, and carried forward.
04
Shared Display
Every compliment also joins a shared screen, collecting as drifting bubbles. What passed between two people becomes something the whole room can see.
0:00 / 0:00
How It Works
Kind Words Club is a browser-based interactive mirror that combines facial expression recognition with live speech transcription. The camera analysis runs locally on the device, while spoken words are transcribed through the browser's speech-recognition service.
Reading the expression
MediaPipe Face Landmarker tracks facial movement in real time. Rather than detecting a smile from a single measurement, the system compares smiling with signs of facial tension, helping it respond more reliably to a relaxed expression.
Connecting face and voice
Speech is accepted only while the smile is held. Words spoken without the expression are discarded, and the interaction resets when the smile is lost. A brief grace period allows for natural changes in expression without interrupting the experience immediately.
This coupling makes the affirmation both verbal and physical: the words must be spoken with the expression, rather than simply repeated afterward.
Responding through voice
Once the phrase and expression are recognised together, the mirror responds with one of two pre-generated voice recordings. These responses were created with ElevenLabs and stored within the project rather than generated live.
Try It Yourself
The mirrored interaction runs here as it runs in the room: it opens listening when it sees you smile, and the words land as you speak them.
Kind Word Karaoke, liveNeeds a camera and a microphone. Speech recognition is Chrome and Edge only.
