CHYIHAUR: From a 3D City to Signage Simulation, Letting Clients See Their Sign Before It Is Made
Before a sign is built, most clients can only look at a flat design and imagine how it will look once it is mounted on the building and lit up. Founded in 1984 with 40+ years of hands-on experience, CHYIHAUR specialises in outdoor signage, wayfinding systems and interior advertising. The question behind this website redesign was simple: can clients see the finished result before anything is fabricated?
POLISH™ built the site around a 3D city made with Three.js as its entrance, paired with a 3D signage showroom of four demo buildings and a real-photo simulator, and delivered the whole site in Traditional Chinese, English, Japanese and Malay. Case articles are written for both search engines and AI search, while the back office uses an AI-assisted workflow in which every content change and social asset is confirmed by a person before it goes live.
- 3D city homepage: texture memory cut from about 681MB to about 30MB
- Upload an SVG logo or a storefront photo and preview the sign instantly
- Four-language site; AI drafts changes, a person confirms before anything is written
1. From a Signage Catalogue to a 3D City You Can Walk Through
Signs live in the city. Lining up photos of finished work like a catalogue makes it hard to feel how a sign relates to its building and street. So POLISH™ turned the homepage into a 3D city: among the towers, a Ferris wheel and city blocks sit six hotspots leading to Consultation, Featured Projects, About, Services, the Signage Showroom and ESG, letting visitors get to know CHYIHAUR by strolling around.
The richer the city, the heavier the load on the browser. We scaled the eight 4K city textures down to 2K, cutting GPU texture memory from about 681MB to about 170MB, then switched to KTX2 compressed textures to bring it down to about 30MB, and self-hosted the Draco model decoder. Materials are precompiled behind the loading screen and the interface only fades in once the city is ready; devices that cannot run WebGL get a static cover with the main links instead.
2. From Imagining the Sign on the Building to Seeing It On Screen First
The Signage Showroom offers 3D simulations on four demo buildings. Visitors pick a template or upload their own SVG logo and the sign is applied to the facade immediately; they can then choose front-lit, back-lit, front-and-back-lit, painted lettering or a light box, and adjust colour, border and brightness as well as position, rotation, size and depth. Pressing Showcase hides every control and slowly orbits the scene, ready to present to a client.
However detailed a 3D model is, what clients really want to know is how the sign will look on their own storefront. Photo simulation lets visitors upload a photo of their shop and drag the sign into place, so they can see the result before deciding. A seven-step guide appears on the first visit, and the colour picker, sliders and upload dialog all work with a keyboard and screen reader, not just a mouse.
3. From a Chinese-Only Site to a Four-Language Case Library Search Engines and AI Can Read
The site supports Traditional Chinese, English, Japanese and Malay under /en, /ja and /ms, with hreflang tags linking each language version. The easiest place for multilingual sites to break is where interface text doubles as a value the code checks, such as the material options in the showroom. We separated display text from identifiers and translate only what users see, so the 3D scene keeps working whichever language is selected.
Every case article has an accordion FAQ below the story. The answers stay in the page even when collapsed and are published as FAQPage, Article and BreadcrumbList structured data, alongside a sitemap, robots file and an llms.txt for AI crawlers. Because 3D canvases are invisible to crawlers, the homepage and showroom also carry crawlable text descriptions, and unpublished cases correctly return a 404.
4. From Editing Articles by Hand to AI Proposals Confirmed by People
Maintaining content in four languages by editing each article by hand invites missed or mistaken changes. POLISH™ added an AI-assisted workflow to the back office: an admin highlights text in an article and leaves a change request, AI drafts a proposal, the review step marks every changed character, and nothing reaches the live site until a person confirms it. Each case can also produce a social pack of 4:5 Instagram carousel slides and caption copy, exportable as 1080×1350 images.
Tracking starts with the visitor’s choice. The site has its own cookie consent banner wired to Google Consent Mode v2; until visitors choose, statistics and marketing cookies stay off, and they can accept all, customise or allow only necessary cookies. We also measured colour contrast across the back office and brought 125 text, badge and button colours that fell below WCAG AA up to standard, so the people using it every day can read it clearly.
Website: https://chyihaur.net
