Photoshop drops a shiny new feature, and somewhere out there, a creator decides to build their own version from scratch. That’s exactly the kind of story that makes you stop scrolling.
I came across a post from this talented creator who saw Adobe’s new “Rotate Object” feature in Photoshop and thought: what if I could vibe code something similar using Claude Code? Not to compete with Adobe, but to push AI to its absolute limits and see what’s possible in just a few hours, on a budget of only $10 in total API costs.
The Setup
The creator already had serious experience under their belt. They’re a frequent user of Nano Banana 2 and had previously built a tool called “Anything Anywhere” that lets you drag and drop objects into a scene and blend them seamlessly. But this time, they wanted to go further: adding a full 3D layer on top of that foundation.
Where Things Got Interesting
Here’s where the story takes a turn. The creator first asked Gemini to handle the job, but it couldn’t deliver. So they pivoted to Claude Opus 4.6 inside Claude Code. To keep costs low, they instructed Claude Code to rely on open-source tools and models. Smart constraint, smart move.
The result? Two brand-new features emerged from that session:
- Background removal for any dragged-in element
- 2D-to-3D conversion using a depth estimation model
The workflow is surprisingly smooth: drag and drop an element, remove the background, convert it to 3D, place it on your background image, hit generate, and you get the updated image.
The Honest Take
I appreciate how transparent the creator is about the limitations. The “3D” here is essentially depth estimation, not true 3D modeling. It’s a prototype built on 2024 technology. But that’s exactly the point: this is what you can build in a few hours for ten bucks.
The creator believes that connecting this approach with top image-to-3D models like Hunyuan 3D could significantly improve the results. The potential is real, even if the current version is a starting point.
Practical Takeaways
- Constraint breeds creativity: A $10 budget and open-source-only rule forced the creator to build something lean and functional
- Claude Code as a prototyping partner: When Gemini couldn’t deliver, Claude Opus 4.6 stepped up and produced working features in a single session
- Know your tools: The creator points out that Nano Banana itself remains the best alternative for production work. The key is finding a workflow that fits your specific use cases
- Depth estimation as a stepping stone: Even basic 2D-to-3D via depth maps can add a convincing spatial layer to image compositing
What I find most valuable here isn’t the tool itself. It’s the mindset. Instead of waiting for Adobe to ship every feature, this innovator just built a working prototype to explore what’s possible. That kind of hands-on experimentation is how you actually learn what AI can and can’t do right now.
Check out the full LinkedIn post for the original walkthrough and visual demos.