It’s an exciting project: automatically generated commands lead to a photo of the place where you are. Or rather: to an image of how an AI “imagines” the place.
You have to come up with that first: A virtual camera that “shoots” photos based on location data and the associated context information. But it is precisely this concept that the interaction designer Bjørn Karmann came up with, implemented and named Paragraphica.
who the Paragraphica page in the browser sees the 3D model of Karmann’s special camera. If you have also enabled location access, a descriptive text about your location appears on the display in English. This includes information such as time of day, address (coordinates), weather, temperature and surrounding buildings, shops or landmarks.
Descriptions become “photos”
With this description as a command, text-to-image generators like the AI software Stable Diffusion are able to output images. Using Stable Diffusion directly is chargeable. However, one can switch to services that offer the use of the software free of charge after registration to a limited extent. These include about “Playground AI“, “InstantArt” or “encyclopedia“. Karmann also offers to generate the images directly on its website for a small fee.
But that’s not all: In addition to the virtual version, the designer has also developed a physical prototype of his camera: its 3D-printed housing contains, among other things, a mini-computer (Raspberry Pi) and a touchscreen.
Mole nose instead of lens
Of course, the camera does not need a lens. Where a camera usually has its optics, in the Paragraphica only the stylized, skin-fringed olfactory organ of the star-nosed mole is symbolically resplendent, with which the animal can smell, touch and even perceive electrical stimuli.
Since light is all but useless to the mole, who lives almost exclusively underground, Karmann has chosen the creature’s extraordinary sensory organ as a metaphor for his physical but lensless AI camera. Because Paragraphica does not draw with light, but rather generates its images using electronically transmitted information.
What does that have to do with reality?
Does the Paragraphica with its image results even come close to the realm of reality? How big are the differences to a real photo and what exactly are they? These questions and the comparison are what make the project so appealing.
On his project page, Karmann gives a first impression of what the image “ideas” of the artificial intelligence are like and what it really looks like at the respective location with some comparison images. There is also a short one Videoshowing the physical prototype of the Paragraphica in action.