Shinhan Card The Classic Air
Shinhan Card
Thematic
A hybrid workflow of generative AI and 3D
What We Did
3D scene production, AI imaging, hybrid generative AI and 3D workflow design, compositing
Watch
Context
Shinhan Card has wanted to use generative AI in its films for some time, and greatminds had already been working in that direction on projects such as the Haru Card commercial. The Classic Air extended that line into a different question. Instead of asking how far generative AI alone could go, this project set out to combine generative AI with 3D rendering in one film. Finflow, the advertising agency, built scenes with generative images and video. greatminds built scenes in 3D. Both streams had to arrive as a single commercial.
Challenge
Generative image and video models are strong at mood, light, and texture, but there are scenes they still cannot deliver. Anything that has to hold an exact form across every frame, or follow a defined camera move, sits outside what generation alone can guarantee. Building those scenes entirely in 3D is possible, but it costs time and gives up the range that generation brings. And the two are not interchangeable. Their tone, their sense of resolution, and the way light behaves in them all differ, so placing one next to the other inside the same film exposes the seam.
Solution
We divided the film scene by scene rather than by tool. Where generation was the stronger route, Finflow produced the scene with generative images and video. Where precision mattered, greatminds produced it in 3D. The scenes that neither route could reach on its own were resolved by fusing the two, with 3D carrying the form and the movement while generative imaging carried the surface. Getting that hand-off to hold across a whole commercial meant building a new workflow rather than a one-off fix, and The Classic Air is the film that came out of it. What the project leaves behind is not only the commercial itself but a working method for the scenes that sit between the two approaches.
From the Team
“The useful question is no longer whether AI can make this, but which part of a scene it should make. Once we stopped treating generative AI and 3D as two competing routes and started assigning them to different layers of the same shot, scenes that had been out of reach for either one opened up. That division of labor is what we actually built on this project.”
Credit
- DirectorYoonseop Kim
- Executive ProducerMinseok Kim
- 3DYoonseop Kim
- AI ImagingHyun Kim
