TfNSW Sydney Metro
Infrastructure Hoarding: Use of AI in Design
THE BRIEF
Hoardings were required for the Bankstown Station plaza ahead of the Sydney Metro opening — not standard construction signage, but campaign-quality design carrying refreshed messaging, an updated Sydenham-to-Chatswood line map and people photography. The goal was a welcoming, optimistic environment consistent with the broader Sydney Metro brand as the plaza opened ahead of services.
THE CHALLENGE
The key obstacle was a complete absence of suitable photography. Available assets either showed the Metro underground or featured older viaduct shots from the Tallawong–Chatswood section — neither reflecting the new, fully above-ground Bankstown corridor. Commissioning bespoke imagery had previously cost around $16,000. I explored AI image generation as a way to bridge this gap without sacrificing brand authenticity.
Existing Imagery
Chat GPT ai generation Dec 2025
Copilot ai generation Dec 2025
PROCESS
After testing Copilot, ChatGPT and Adobe Firefly — each of which struggled with train-model accuracy or multi-reference image input — Higgsfield (running Google’s Nano Banana Pro model) proved the strongest tool.
Using two reference images and a targeted prompt, it generated a convincing above-ground exterior scene. Real train details and branding were then composited back in Photoshop, and the result AI-upscaled within Higgsfield to clean up the blending of features.
OUTCOME
The final image placed the teal Sydney Metro train in an authentic urban rail corridor matching the campaign aesthetic. It was incorporated directly into the hoarding layout alongside messaging panels and the Metro line map — delivering a campaign-grade result at a fraction of the cost of a traditional commission.
Adobe Firefly ai generation Dec 2025
Nano Banana Dec 2025(chosen ai generation)
Photoshopped Layer
