Create a 3D model for augmented reality: the complete guide
There are three ways to get a 3D model for augmented reality: scan the product with a smartphone (photogrammetry or LiDAR), have it modelled by a professional, or generate it with AI from a photo. Whichever you choose, aim for a lightweight, true-to-scale GLB file with PBR materials.
Useful terms: GLB · USDZ · Photogrammetry · LiDAR scanning · AI 3D generation (image-to-3D) · Decimation · PBR (physically based rendering)
No commitment, no credit card required.
Why do you need a 3D model for augmented reality?
Augmented reality relies on a fundamental principle: superimposing a virtual object on the real-world image captured by the smartphone camera. This requires a 3D model, i.e., a three-dimensional digital representation of your product.
This is a fundamental difference from traditional photography. A photo is a flat image, frozen at a given angle and lighting. A 3D model is a complete digital object that the customer can rotate, zoom in on and project into their real space.
For many professionals, creating the 3D model is perceived as the main obstacle to adopting augmented reality. "I don't have a 3D model" is the most common remark. Yet current solutions make 3D creation accessible to everyone, regardless of budget or technical skills.
Which 3D file format should you use for augmented reality?
Two formats matter for mobile augmented reality: GLB (the binary version of glTF), the open format of the web and Android, and USDZ, read by AR Quick Look on iPhone and iPad. OBJ and FBX files, common in 3D software, must be converted to GLB before publishing. On Visuality, you simply upload a GLB or a USD/USDZ file, and the USDZ for iPhone is generated automatically.
For a detailed comparison of the formats (animations, textures, compression, conversion pitfalls), read our article GLB or USDZ: which format should you choose? Visuality's own technical recommendations are in the documentation on 3D object formats.
Method 1: how do you scan an object in 3D with a smartphone?
Smartphone 3D scanning is the most accessible method for digitizing an existing physical object. The principle is simple: you take dozens of photos of your object from all angles, and a photogrammetry algorithm reconstructs the textured 3D model.
Apps such as RealityScan (Epic Games) or Polycam automate the scanning. iPhone and iPad Pro models with a LiDAR sensor also measure depth, which speeds up capture, especially for large objects.
3D scanning is ideal for existing physical objects: products in stock, display pieces, prototypes. The result is a textured 3D model that faithfully captures the object's geometry and colors.
Find our detailed guide on 3D scanning in our documentation on 3D model generation.
Method 2: when should you use professional 3D modelling?
Professional 3D modeling involves creating the model digitally, vertex by vertex, using 3D design software. It's the method that offers the most complete control over the final result.
The most commonly used software for augmented reality-oriented 3D modeling includes:
- Blender: Free and open-source software, extremely powerful, with an active community and abundant documentation. It natively exports to glTF/GLB.
- 3ds Max: Autodesk's professional software, widely used in architecture and product design. glTF/GLB export is built into recent versions, or available through a plugin.
- SketchUp: Accessible and intuitive software, particularly suited for furniture and interior design. Export to glTF is possible via extensions.
Professional 3D modeling is the best approach for conceptual products (not yet manufactured), customizable products (color and material variants) and projects requiring maximum detail and fidelity.
If you don't have 3D modelling skills in-house, you can hire a provider. The cost depends on the object's complexity, the level of detail and the number of variants: ask for quotes, specifying the use (web and augmented reality), the expected format (GLB), the real-world scale and a file-size limit. Visuality also offers custom modelling on quotation.
Method 3: how do you create a 3D model with AI from a photo?
AI 3D model generation is the most recent and most accessible method. The principle is revolutionarily simple: you import one or more photos of your product, and the AI automatically generates a 3D model.
Visuality builds this tool into its platform. From your 3D object library, you upload a front photo (simple generation) or four views of the product (detailed generation), and the AI creates a 3D model within minutes, for 1 credit. The result is a GLB file ready to use in the 3D viewer and in augmented reality.
This method is ideal for professionals who have photos of their products but no 3D models, and want to get started quickly with augmented reality.
The quality of AI-generated models has improved considerably. For the majority of consumer products (furniture, decorative objects, equipment), the result is sufficient for a convincing augmented reality experience.
Find the complete guide to our AI tool in the documentation on AI 3D generation, and a detailed look at its limits in our article AI and 3D: transform a photo into a 3D model.
How do you make a good 3D model for augmented reality?
Regardless of the method chosen, certain fundamental principles guarantee a quality 3D model for augmented reality:
- Keep the file size reasonable: aim for around ten megabytes. Google recommends 10 MB for its Scene Viewer and warns that heavier models make for a poorer experience, especially on a mobile connection.
- Take care with textures: Textures are what bring the 3D model to life. Use textures with sufficient resolution (1024x1024 or 2048x2048 pixels) and PBR materials for realistic rendering.
- Respect real scale: This is crucial for augmented reality. Measure your real object precisely and make sure the 3D model reflects these exact dimensions. A model at the wrong size makes the AR experience useless: a vase that appears as big as a building or a miniature sofa convinces no one.
- Test on mobile before publishing: The AR experience is primarily on smartphone. Systematically test your models on a real phone to verify proportions, material rendering and interaction fluidity.
- Optimize polygon count: An overly detailed model will be heavy and slow. For AR on mobile, aim for between 10,000 and 100,000 polygons depending on the object's complexity.
These values are in line with Google's recommendations for Scene Viewer (100,000 triangles at most, 2048 × 2048 pixel textures, 10 MB) and Apple's for AR Quick Look (about 100,000 polygons and 2048 × 2048 pixel textures). Verified on October 15, 2026.
How do you upload your model to Visuality?
Once your 3D model is created, integration on Visuality is quick and intuitive:
- Upload: in your 3D object library, click Add an Object, then I Already Have My 3D Object, and upload your GLB or USD/USDZ file (or the RealityScan ZIP export). The platform hosts the model, and its USDZ version for iPhone is generated automatically.
- Configuration: set the object's size, its floor or wall placement, a background image or an action button, and add annotations (hotspots) in the editor to highlight the product's key features. Then check how it looks in augmented reality on your smartphone.
- Sharing: your model is ready to share by direct link, QR code, iframe snippet for your website, or through the Shopify app and the PrestaShop module. Your customers get instant access to 3D viewing and augmented reality.
The entire process, from upload to going live, takes just minutes. Visuality automatically handles hosting, optimization and multi-device compatibility.
Frequently asked questions
Sources
- Scene Viewer — File requirements for models — Google for Developers, Verified on October 15, 2026
- AR Quick Look model recommendations — Apple Developer Forums, 2021
- glTF Overview — Khronos Group, Verified on October 15, 2026
- glTF 2.0 import/export — Blender Manual, Verified on October 15, 2026
- Polycam pricing — Polycam, Verified on October 15, 2026