3D Modeling - Everything that You Need to Know about WebAR

Last updated on May 18th, 2024

Marvin XR can 2x your online furniture sales with 3D Ecommerce

The foundation of any AR/VR/WebAR/XR experience lies on 3D models. But building 3D models are complex, costly and time consuming process, especially in professional arena.

So how to build a 3D asset library, especially for ecommerce, where you need numerous 3D models in brief timeline to setup virtual try-on, without bleeding from high budget issue? How to ensure quality check 3D models for WebAR usage?

How 3D is captured in 2023?

The 3D capture space is evolving so rapidly in 2023 that it is hard to stay on top on everything. Especially AI is driving this space so fast (Read the article published on this before).

However, we can broadly categorize the methods, based on which 3D models are generated today:

3D Scanning

3D scanner technology appeared in the 1980s. Objects are scanned by laser and images are catured by camera. 3D scanned objects are reversed engineered using CAD drawings. You need scanning devices, follwed by software to process the outcome.

Image by Surface Scan

Photogrammetry

It's is the technique of taking multiple overlapping photographs and deriving measurements from them to create 3D models of objects or scenes. In reality, sometimes hundreds of images from different angles are needed to be processed via software to get a 3D model. An exhaustive technique to build 3D but with comparatively high quality.

How Photogrammetry is done

LiDAR

iPhone 12 Pro brought this into surface. Built directly into the device, the Light Detection and Ranging (LiDAR) scanner emits a laser to more accurately measure distances of surrounding objects and then create a volumteric 3D scene. Good to capture 3D fast but often needs too much manual editing.

LiDAR Scanner iPhone 12 demo - Instant 3d scans!

NeRF (Neural Radiance Field)

A neural radiance field (NeRF) is a fully-connected neural network that can generate novel views of complex 3D scenes, based on a partial set of 2D images. In 2023, NeRF based 3D modelling is getting more popular but mostly for amateur purposes.

Luma NeRF 3D Scanner App

Point Cloud

This is essentially the later processing part of 3d scanning, lidar, photogrammetry. Based on the volumetric capture, the process creates 3D represntation by spatial points (dots) and turning itself into a 3D model.

Point Cloud scanning

Text to 3D

Generative AI next generation software are getting more smart. Unlike Midjourney or DeepAI, these new AI systems let users to create 3D models based on text inputs. These are quite cool. But it's hard to mimic a real world object. It's hard to describe a designer sofa with cushion texture to be described only by words.

Using AI to create text based 3D model

Image to 3D

Another generative AI masterpiece evolution - creating 3D model based on image input. This can change the game on how accurately digital twins can be built in the future. But most systems are often not stabilized yet to build realistic 3D models to be used for professional purpose.

sudo.ai generating 3D model from 2D image

Read more about AI based 3D modeling in our blog.

Manual 3D Modelling

This is the classic way of building 3D model with professional quality as of now. 3D artists can cover real object geometry, lighting, texture, skin, sizing, rendering and many aspects using software like Blender or Maya. But like any manual process, the output is slower and often needs a higher budget (based on spent hours allocated for 3D building).

3D Model by Marvin XR - based on sofa images

Quality check of 3D models for WebAR

Virtual try-on is special for product experience because users can try it from browsers, without downloading any apps. Loading 3D model in a browser calls for some special parameters to consider:

Low Poly count

Decreasing the polygon count for 3D model drastically reduce the size of it, thus rendering it in the browser becomes lighter and easier. Keeping a perfect balance between edge, vertex and face is the secret of a high quality 3D.

3D Modeling basics

Low poly 3D can load faster in user's browsers

Use GLTF/GLB format

There are several 3D model extensions out there. GLTF 2.0 or GLB is the most lightweight version of all 3D extensions. Packaging your 3D model into GLB can end up into a better user experience from WebAR.

UV unwrapping & Texturing

UV unwrapping and texturing of the 3D model present the real object like look and feel. Too much detailed texturing can increase the 3D model size. A balanced wrapping and texturing can bring the professional look while keeping its size under check.

Blender is excellent at 3D modeling

3D Model Size

Traditional 3D models used to be of several hundreds MBs. But that does not fly with WebAR. Try to compress 3D models to fit within a small size. A 3D model between 2MB-10MB can do the trick. Ideally 5MB size is the sweet spot for a 3D model to be used in virtual try-on or WebAR.

3D Modeling is critical for WebAR experience

What are the challenges for mass 3D production?

Most automated and AI based 3D modeling need handholding by 3D artists if you need to use them for professional usage. This is a real challenge to produce mass scale 3D asset library. On the other hand, if you focus more on perfection, you can end up hiring bunch of 3D artists increasing the budget and time manyfold for the similar delivery (of course unless budget is not a limiting factor which seldom happens).

How to speed up 3D building Process?

If you need to build more 3D models within short period of time, you need to adopt some smart strategy. Depending only on AI may not be the ultimate answer to your challenge. You need to build a hybrid process where AI can help to build the 3D skeleton within minutes and then deploy 3D artists to curate the texturing, skin, color, lighting, scaling, look and feel. Thus in the end, you'll be able to produce high quality 3D models in less time, budget without comprosmising with the professional quality standards.

How Marvin builds 3D assets?

How Marvin XR builds professional low cost 3D models faster

Marvin XR has built a hybrid AI pipeline to build 3D models at scale. The generative AI helps to create quick 3D prototype only from a handful images of an object. Then professional 3D artists work on them to create the digital replica of it in 3D - making it a true digital twin. The fast 3D production helps ecommerce brands to build up 3D asset library and implement virtual try-on even within a very tight schedule. The budget friendly approach can accelerate online sales conversion quickly. You can see results from just within 5 weeks once you start using Marvin WebAR for your ecommerce store.

Contact Now