Echo of Mind began with a question about radical vulnerability — what happens when we make our most private mental states publicly visible? Dreams occupy a unique position: they are deeply intimate, impossible to fake, and entirely beyond our conscious control.
The project explores whether technology can act as a bridge between interior experience and collective understanding — not through language, but through abstract visual form.
A person presses the blue button and speaks their dream into a microphone. OpenAI Whisper transcribes the voice recording into text, which is passed to a custom DreamWeaver GPT Agent. The agent interprets the dream across three psychoanalytic dimensions — figurative, environmental, and emotional — and outputs a structured Dream Signature as JSON.
This signature becomes a visual prompt sent to FLUX.1-schnell via Hugging Face, generating a unique painterly image. The image is post-processed at 800×480 with grain and vignette, displayed on the Pi screen, and simultaneously pushed to a public GitHub archive.
The Collective Archive — a live website hosted on GitHub Pages — fills up over time as more people use the device. Each grid square represents one person's subconscious state. The archive persists and grows, connecting strangers through their interior lives in a way that language cannot.
The system runs entirely on a Raspberry Pi 5 inside Docker, exposed publicly via a Cloudflare Tunnel. The website polls GitHub every 10 seconds and automatically places each new generated image into the next empty square on the grid.
Voice → Whisper → Transcription → DreamWeaver GPT → Dream Signature JSON → Visual Prompt → FLUX.1-schnell → Painterly Image → 800×480 Post-Processing → Pi Display + Local DB + GitHub Push → Archive Website
System Architecture · Signal Flow
