I go to my favourite industrial design reference site https://lemanoosh.com/, time to time. It’s always filled with new renders of new hardware concepts and devices. I was sure that I will find latest iPhone 17 renders over there.
This time I had an idea to test some new stuff. So, I opened up Gemini for image generation on the side tab, to test its nano banana - the image generation model.
In this article you will see what I tested by combining multiple images and prompting to produce worthy compositions.
Nano Banana : A game changer for visual creativity
Nano Banana is the most recent model of image generation from Google Gemini. It allows to edit images using natural languages with consistency, details and photorealism. Pose changes, background swaps, objects additions among many other features. Making it ideal for industrial design compositions. It allows images to edit multiple times with complex editing.
Let’s check what the hype is all about. In the next stage I wanted to test out the tool, without no over expectations.
Collecting References
I started collecting images of different products not related to each other. A shoe, a earphone, a bike and an iPhone.
Editing with Nano Banana
Prompt 1 :
Athletic woman in black sportswear and red-tinted sunglasses, standing confidently with a sleek, modern look against a plain background.
I attached the image of shoes and earphones, there was no reference image of woman attached.
I was already in awe and inspired with the first image.
Prompt 2 :
Add bike
I attached cropped image of the bike.
Prompt 3 :
Add iPhone
I attached image of the orange iPhone.
Prompt 4 :
Now at this point I wanted to stop, not making the composition over crowded. But I wanted to change the pose. What the maximum it can do?
Make her breakdance
This was absurd and crazy. I was beyond impressed. It’s bananas. It truly lives up it is name.
The final image was composed in stepwise process. Without dumping images in one go. Making incremental changes.
To my further investigation I found a tool (not tested yet) which can help in making variations systematically and keep style alignment. https://www.kittl.com/features/ai/flows
This is also in line with my previous thoughts on node based ai editors. Please check that article here:
With Nano written in the name of the model, I can safely assume there is a bigger model cooking in Google’s labs.
Conclusion
We can conclude that:
Can make great believable images.
Can maintain character with previous details in addition to new objects.
Can work with incomplete/cropped references.
Can change poses of the characters. Even dramatic poses.
In my search I didn’t find much information pertaining industrial design and nano banana. I am pretty confident, this will change change my composition workflow.
In following weeks I will experiment with material & style, integrate with environment and scene composition.
Announcement:
I'm hosting free branding workshops for healthtech startups and companies wanting to improve their market presence. This two-part sessions will help participants reassess their current branding approaches. Book your session here:
https://calendly.com/ziro-nikhil/30min
For One Page Landing needs check my work here:
https://ziro.agency/
For health related design work here:
https://ziro.health/
