From photo to AR, in one pipeline.
Every product flows through the same path: ingest, reconstruct, optimize, render, deliver. Each stage is engineered to remove friction for both the merchant and the shopper.
Vizbl is a vertically integrated stack, AI that builds 3D models from photos, a proprietary compression layer that ships an 80 MB model as 2.5–4 MB, a WebGL renderer tuned for true-to-scale fidelity (±5 mm), and a WebAR layer that anchors objects to floors, walls, ceilings, even to other objects. The same engine runs on the product page, the showroom iPad, and the construction-site iPad.
Every product flows through the same path: ingest, reconstruct, optimize, render, deliver. Each stage is engineered to remove friction for both the merchant and the shopper.
Upload product photos, CAD or existing GLB / USDZ. Vizbl normalizes inputs into a single internal format.
The Vizbl model rebuilds geometry, separates surfaces, infers materials and applies physically-based rendering textures.
Decimation, Draco compression, KTX2 textures and LODs make models small enough for mobile, sharp enough for hero shots.
Models stream from the edge, render with WebGL, and launch AR via WebXR, Quick Look (iOS) and Scene Viewer (Android).
Some Vizbl categories don't need per-SKU 3D at all, for rugs, flooring, wall art and wheels, a flat product photo is enough and the engine generates AR on the fly. For everything else (furniture, electronics, accessories, jewelry, watches) you need actual 3D models. Here are the paths we offer.
Our recommended path today. Send us product photos and dimensions; our 3D team produces optimized, mobile-ready GLB/USDZ, color and material variants included. See services →
Status: production · Most clients use this
We're training a neural network that reconstructs an AR-ready 3D mesh from a few standard product photos, no scanner, no studio. Currently in active development; we'll roll it out to existing clients first.
Status: in development · Pilot access on request
We have an in-house photogrammetry rig you can use for organic and complex shapes. It's a real option, but quality depends a lot on the object and the lighting, we'll be honest about whether it fits your category before you commit.
Status: available · Quality varies by category
Rugs, flooring, parquet, tile, vinyl, wallpaper, wall art and posters auto-generate AR from your existing flat product photography. Zero per-SKU 3D modeling required.
Status: production · Built-in to Vizbl
Wheels and rims are tried on the customer's actual car using AI segmentation, only flat catalog imagery needed. Family 03 →
Status: production · OEM-ready
If you already have GLB, USDZ, FBX or OBJ assets, drop them into the Object Manager. Vizbl auto-converts, validates dimensions and applies our compression pipeline (80 MB → 2.5–4 MB) before publishing.
Status: production · Free
For an in-depth breakdown of every option to get 3D models for AR, including third-party libraries, manufacturer assets and freelancers, see the Learn article.
Vizbl's renderer is tuned to look great on a $200 phone and on a 5K display. Real-time IBL, shadowed materials and tone-mapped output deliver console-grade visuals through plain WebGL.
Scenes inherit a soft studio HDRI by default, or your brand's own environment.
Three quality levels stream automatically depending on bandwidth and device class.
Loaded after first paint. Vizbl never blocks your product page.
Vizbl uses each platform's native AR runtime, Apple Quick Look on iOS, Scene Viewer on Android, WebXR on the open web, so users never download anything. The same model and the same call site light up across all three. Objects anchor to floors, walls, ceilings and to other Vizbl objects, with placement accurate to ±5 mm.
<!-- Drop into your product page --> <script src="https://cdn.vizbl.com/v1/viewer.js"></script> <vizbl-viewer model-id="sofa_halden_pebble" ar poster="/img/sofa.jpg" scale="auto"> <button slot="ar">View in your room</button> </vizbl-viewer>
For accessory categories, jewelry, watches, eyewear, hats, scarves, body-tracking AR is overkill. The customer just needs to see the product on themselves. Vizbl Try-On runs an AI pipeline that takes a single photo and renders the product onto it with real lighting, scale and perspective.
Detects face, hands, neck, wrist, ears and head orientation from a single static image, no live camera, no body tracking, no calibration.
Vizbl's diffusion-based renderer matches the product's shadows, highlights and skin-tone tints to the original photo, so it reads as a real shot, not a sticker.
Vizbl Try-On takes a 3D model of the accessory and the customer's photo, and renders the result with photo-real fit, drape and lighting. We help with 3D production if you don't have models yet.
Switch metals, gemstones, frame colors or lens tints live, the same uploaded photo restyles in under a second.
Photo is processed in-session and not retained by default. Storage is opt-in (saved looks, boards). GDPR & CCPA compliant.
Jewelry retailers using AR try-on see +32.7% add-to-cart and −17.4% returns (Snap × Deloitte). Vizbl Try-On is built to match.
Read the product page: Family 06 · Vizbl Try-On.
Vizbl is async-loaded, edge-cached and lazy-initialized. It never blocks your Time To First Paint, and it doesn't fight your existing storefront for resources.
Loaded after first paint. Zero impact on Lighthouse scores.
Models served from a global CDN with HTTP/3 and Brotli.
Vizbl ships an 80 MB raw model as 2.5–4 MB without visible quality loss, no overheating, no freezing, even on standard phones.
Renders on iOS 12+, Android 8+, all modern desktop browsers.
All assets are served over TLS, with strict CORS and signed model URLs.
Camera frames stay on the device. Vizbl never uploads what the camera sees.
Operational controls modeled on SOC 2 Type II for enterprise customers.