There’s a very interesting new entrant into the radiance field platform world, with Postshot from Jawset. Eliminating the need for complex command lines, Postshot offers an incredibly straightforward and user-friendly experience to begin exploring and creating with radiance fields.
Their radiance field methods, both for Splats and NeRFs are are all custom implementations directly using CUDA in the backend. Meaning no existing radiance field related 3rd party projects are used.
File uploads are as simple as drag and drop onto the window and then selecting your training settings.
Me at the Magnopus office in summer 2022.
You can watch the training take place live, which is one of the original things that got me hooked on Instant NGP in early 2022. Additionally, you can see the PSNR as it trains, which is something I’ve always been curious about in my captures.
Some parts of Postshot could use a bit of refinement, such as hoverable information on icons would be helpful, but this *feels* like a more complete product than some of the companies that we’ve seen thus far. It has functionalities that might feel more akin to the Adobe Suite or a word processor than an cutting edge technology platform and that is a good thing. You can save, edit, and load files the same exact way that you would in Microsoft Word.
All of this bypasses the command line. I’ve been running tests pretty much all day yesterday and today and it’s a straightforward application to use. Users can choose between creating a NeRF or a Splat model, with simple dropdown menus for setting parameters. Catering to various needs, Postshot provides five distinct NeRF modes:
S Mode: Ideal for basic, toy-like testing.
M Mode: A significant upgrade, allowing for the capture of real scenes with minimal memory requirements.
L Mode: The recommended default for producing high-quality images.
XL and XXL Modes: Designed for achieving fine detail in the center of scenes or managing large-scale scenes.
With all the features that are packed into Postshot, it’s hard to believe that I haven’t even mentioned you can export directly into After Effects.
For those who have felt intimidated by the technology, I would recommend giving this one a go. You will feel comfortable using the application and understand how to navigate it. All radiance fields trained by Postshot are done locally, meaning that the data stays on your device.
Postshot is free to use and during the Beta phase, you can use it commercially.
This is a beta that literally released yesterday, but their floor is already high. I will be paying very close attention to the Germany based team and I strongly suggest that if you are on Windows to download it.
New to Gaussian Splatting? Start here
