Localising a full campaign without a single new shoot.

Overview

Dongfeng approached us with a library of campaign assets shot in China, with the goal of localising the visuals for the Malaysian market in their stores. Rather than organising a costly local production shoot, we developed a fully AI-driven pipeline to seamlessly transplant the campaign into a Malaysian context keeping the client's original assets intact while making the final output feel authentically local.

Challenges

The core challenge was preserving the look, feel, and quality of Dongfeng's original photography while replacing the environment entirely. Matching lighting, perspective, and surface reflections on the vehicle across different background. On top of that, the solution needed to be efficient and repeatable across multiple assets.

Solution

We trained a custom LoRA model on the client-provided assets to capture the specific visual identity of the Dongfeng vehicle. Using ComfyUI with Flux and experimenting with NanoBanana, we built a generative pipeline that composited the car into Malaysian environments with consistent lighting and detail. This approach eliminated the need for a physical production shoot while delivering localised visuals that met the quality bar of the original campaign assets.