Snap Brings 3DGS Trainer into Lens Studio 5.3

Michael Rubloff

Michael Rubloff

Nov 17, 2024

Email
Copy Link
Twitter
Linkedin
Reddit
Whatsapp
Snap Lens Studio
Snap Lens Studio

Over the past year and a half, Snapchat has steadily embraced Radiance Fields, starting with MobileR2L within the Snap app, introducing 3DGS support in Lens Studio, a pipeline from Scaniverse into Spectacles V5, and now today we're also getting trainer and pipeline within Lens Studio.

It's worth noting that Snap also owns PlayCanvas, the maker of SuperSplat, further embedding the AR camera company into the Radiance Field ecosystem more than it might initially seem. Unlike Polycam, Scaniverse, and until recently, Luma AI, Snap’s implementation focuses on individual objects rather than larger scenes, offering a workflow that feels somewhat akin to Dolly, albeit serving different purposes.

Previously, creating these captures required training on a separate platform before importing them into Lens Studio. However, with the introduction of Lens Studio 5.3, this barrier to entry has been significantly lowered.

Using Snap’s new trainer requires downloading the Lens Studio desktop app for Mac or Windows. Once installed, you simply sign in with a Snap account to access the Gen AI Studio. While Gaussian Splatting might not typically fall under this designation, the corresponding extension is now housed under the 3D Model feature.

From here you can upload a video of an object you would like reconstructed and type in the focus of the scene. For the below example, I typed in baseball cap. Additionally, you don't need crazy strong coverage. The videos I used were about 20 seconds each.

The platform appears to be cloud-based, meaning your captures are trained off-device. The reconstruction process is swift—usually taking about 10 minutes or less. Once reconstructed, the object can be seamlessly imported into Lens Studio with a single click.

As someone new to Lens Studio, I found the platform surprisingly intuitive. This project took under an hour to complete. There are so many ways to leverage this platform to bring. The original Snap documentation for the workflow can be found here, but please note that the category name has been updated from Gaussian Splatting to 3D Capture in Lens Studio.

Snap's Lens Studio can be downloaded here for free.

Featured

Recents

Featured

News

Deckard Render in Unity adds Free Tier

The Unity plugin now has a free tier for higher rendering effects.

Michael Rubloff

Nov 15, 2024

News

Deckard Render in Unity adds Free Tier

The Unity plugin now has a free tier for higher rendering effects.

Michael Rubloff

Nov 15, 2024

News

Deckard Render in Unity adds Free Tier

The Unity plugin now has a free tier for higher rendering effects.

Michael Rubloff

News

Radiance Fields in St. Peter's Basilica

One of the most iconic landmarks gets hyper-real.

Michael Rubloff

Nov 14, 2024

News

Radiance Fields in St. Peter's Basilica

One of the most iconic landmarks gets hyper-real.

Michael Rubloff

Nov 14, 2024

News

Radiance Fields in St. Peter's Basilica

One of the most iconic landmarks gets hyper-real.

Michael Rubloff

NeRFs appear in NY Times Profile of Artist

OpenAI Artist in Residence, Alexander Reben used NVIDIA's Instant NGP for his recent artwork.

Michael Rubloff

Nov 13, 2024

NeRFs appear in NY Times Profile of Artist

OpenAI Artist in Residence, Alexander Reben used NVIDIA's Instant NGP for his recent artwork.

Michael Rubloff

Nov 13, 2024

NeRFs appear in NY Times Profile of Artist

OpenAI Artist in Residence, Alexander Reben used NVIDIA's Instant NGP for his recent artwork.

Michael Rubloff

Guest Article

Exploring Cross-Modality in Radiance Field Technologies and its Intersection with Physical AI

Radiance field technology has grown more prominent in computer vision in the past few years, enabling exact 3D reconstructions using neural representations of volumetric input.

Preetish Kakkar

Nov 13, 2024

Guest Article

Exploring Cross-Modality in Radiance Field Technologies and its Intersection with Physical AI

Radiance field technology has grown more prominent in computer vision in the past few years, enabling exact 3D reconstructions using neural representations of volumetric input.

Preetish Kakkar

Nov 13, 2024

Guest Article

Exploring Cross-Modality in Radiance Field Technologies and its Intersection with Physical AI

Radiance field technology has grown more prominent in computer vision in the past few years, enabling exact 3D reconstructions using neural representations of volumetric input.

Preetish Kakkar