A camera-like object that remembers a moment by turning it into music.
I designed the device experience around a physical shutter button, connecting image capture, AI interpretation and music playback in a handheld prototype.
The problem
Most cameras preserve an image, but not the atmosphere of a moment. mmento-1 starts from a different idea: what if taking a picture could create a musical memory instead of only a visual archive?
Design & implementation
The prototype runs on a Raspberry Pi style setup with camera, screen, button, battery monitoring, and local playback. When the button is pressed, it captures an image and freezes that moment on screen.
That image is then interpreted locally through a vision-language model to produce two structured outputs: lyrics and style. Those outputs are sent to a music generation service, which later returns a generated track through a callback. The device downloads the result, plays it, and overlays the generated text on the captured image.
What matters here is the meaning of the loop: image capture, interpretation, generation, callback handling, and playback all happen inside one object whose purpose is to turn a lived instant into a replayable musical souvenir.
What it does
- Camera-like object designed around musical memory
- Multimodal pipeline from image to lyrics to music
- Local vision-language inference with remote music generation
- Embedded playback, display, and hardware-button interaction
- Single-device interaction loop with callback-driven audio return
Why I built it
mmento-1 matters because it treats memory as something that can be translated rather than merely stored. It is less about a model in isolation than about building an object that can transform a moment into a piece of generated media.
