AI camera designs itself, (kinda).
By Jason Robinson & Daniel Huang
Written in collaboration with Jenna Fizel, and Danny Deruntz. Part two of a series on prototyping a Generative AI Camera. Part one is about the initial software prototype by Jenna Fizel.
So, what would a physical AI camera look like, anyway? Would it have a lens? How would you interact with it? How might AI play a role in the design process? Could it design itself? These questions ran through our minds when we first talked with Danny and Jenna about building a tangible version of the “Vibe Camera” described in part one of this series.
Daniel Huang and I are both Industrial Designers who pick up the challenge to imagine a physical version of a Generative AI-enabled imaging device. First we need some…
Ground Rules and Design Constraints
- We will involve AI tools in the process as much as possible.
- We would “de-emphasize” the traditional lens and the focus on image quality.
- Camera designs have to print a tangible photo of whatever “Vibe” it sees in the environment.
- Concepts have realistic volumes that can accommodate feasible technology, with the goal of making a working prototype as a next step.
With these guidelines in place, we take the next logical step that any designer would do these days…
Daniel Huang (DH): asks Midjourney to imagine a generative AI camera that captures vibes and de-emphasizes the lens and image quality:
imagine a 3D Keyshot C4D rendering of a futuristic black camera that has no lens shot on a yellow background — v 5 — ar 4:1
Jason Robinson (JR): Ok, great. But not great. It just looks like a normal camera. AI has the “don’t think of a pink elephant” thing going on. Meaning if you mention “lens,” it will definitely imagine a lens. Also, this process was neither intellectually or creatively interesting.
DH: I agree. This is not quite right. What if instead, AI gave us some prompts and then we can try to interpret it? Kind of flipping the Midjourney workflow on it’s head.
JR: Nice. AI as a thought partner, not as a freelancer. We’re looking for something that challenges the mental model of a traditional camera. Maybe ChatGPT can help us think metaphorically about how this camera might be expressed.
DH: [to ChatGPT] “You’re designing a camera that shows off how AI would see the world. Give me some poetic metaphors of what this camera might look like.”
ChatGPT: [replies]
- It’s like a kaleidoscope, reflecting and refracting the world into new and beautiful patterns.
- It resembles a luminous mirror, where AI sees its own reflection, recognizing the interconnectedness between observer and observed.
- It’ like a mystical artifact, transcending the ordinary, capturing the extraordinary moments that define our existence.
- It’s like a looking glass, reflecting the infinite possibilities of perception, revealing the hidden depths and dimensions of existence.
- It’s like a crystal ball that reveals the hidden patterns of the world, and predicts its future with uncanny accuracy.
- It’s like a conductor’s baton, orchestrating the symphony of visual experiences, harmonizing the elements of perception.
- It manifests as a compass, guiding AI’s vision through the labyrinth of possibilities, navigating the vast expanse of understanding.
- It resembles an enchanted book, filled with pages of visual poetry, where each image tells a story and unlocks the mysteries of the universe.
DH: Ok, now this is starting to get interesting as these prompts push us to imagine new constructs of a camera.
JR: Love these. Let’s take 5 of them and see what we can come up with.
Industrial-Designer-Text-To-Image
Below, you can see sketches and mock-ups by Daniel and me that resulted from the “design prompts” provided by ChatGPT.
SX70AI
A mystical artifact, transcending the ordinary, capturing the extraordinary moments that define our existence. In this design, we pay homage to the beloved Polaroid Land Camera but with an added layer of delight. Here, each frame of film comes with randomized prompts built into the print. Sort of like a new type of AI metadata that is only revealed on a small display when the expose is taken. The approach builds on the original innovation of embedding the chemicals and battery into the Polaroid film itself and the emergent nature of the [not quite instant] film developing process.
Looking Glass
Reflecting the infinite possibilities of perception, revealing the hidden depths and dimensions of existence. This concept uses the metaphor of a “Looking Glass” to examine tiny subjects to capture unseen realities through the magic of augmented AI. A small board mounted camera in the handle of the device detects the subject being examined through the magnifying lens. This image is then processed through the vibe app and sent wirelessly to the printer in the base. The lid also serves as a print tray to collect the images and reinforce the archival theme of this design.
Kaleidoscope
Rotates to unveil prompts, reflecting and refracting the world into new and unexpected patterns. In this interpretation, we put a new twist on text based AI prompting by blending the familiar interactions of a camera with a kaleidoscope where rotating the focus ring of the lens toggles between different facets of AI perception. We call this interaction “Physical Prompting”.
The Luminous Mirror
Where AI sees its own reflection, recognizing the interconnectedness between observer and observed. This concept places the camera’s lens inside a two way reflective half dome. By suspending this lens inside a mirrored box, it gives the illusion of a perfect sphere floating in space. The images captured by this device is further manipulated through symmetrically altering AI prompts.
The Crystal ball
A stationary camera that sporadically generates new possible futures, revealing the hidden patterns of the world, and predicts its future with uncanny accuracy. This camera prints 360 degree spherical images that can be refracted and viewed through the crystal ball lens to create a unique imagining experience.
Our take-aways from this exercise.
We explored several AI tools in the course of our experiment, including Midjourney, Dall-E, Adobe Firefly, ChatGPT, and Sloyd. We found that AI served as a helpful design thought partner to inspire divergent thinking, rather than a turnkey solution provider.
ChatGPT turned out to be the most inspiring tool for us. It pushed us to think differently about the brief, and the back and forth nature felt fluid and conversational. It suggested the physical prompts on the Kaleidoscope concept, and we even used it to write some python scripts for Blender that would create 3D CAD of a camera, more on that later.
We abandoned Midjourney early on in the process, but Adobe Firefly did the grunt work of creating all the Polaroid image creations you see in the renders.
AI served as a helpful design thought partner to inspire divergent thinking, rather than a turnkey solution provider.
As product designers, our ultimate goal is to create something that exists in physical space and possesses real-world functionality and utility. This is beyond the current functional state-of-the-art of AI. However, there’s no doubt that such a workflow will become mainstream soon. Even today, ChatGPT was able to write simple Python scripts that approach something inspired and usable in 3D CAD, albeit with a lot of patience and trial and error. Below is a simple camera as imagined by ChatGPT in Blender. It’s very basic, and yet inspiring in its primitive form. With a bit of added detail and the right color, material, and finish (CMF) treatment, this could evolve into something very cool. It could also potentially be sent to a 3D printer for rapid prototyping.
This workflow feels like a great topic for a future chapter where we might just do that. Stay tuned as we hope to share more soon…
