Welcome to another edition of VP Land.
Two things are all over my feed right now:
So let’s jump into both (and more)!
3D Gaussian Splatting: A New Way to Capture Photorealistic Scenes
Watch out NeRF, there’s a new way to scan and digitize the world: 3D Gaussian Splatting.
As a refresher, NeRFs (neural radiance field) are a way to generate photorealistic 3D scenes from a partial set of 2D images.
What’s 3D Gaussian Splatting?
Single rasterized Gaussian, with a border. Source: Dylan Ebert on Hugging Face
Traditional computer graphics rasterize images by drawing lots and lots of triangles. Now, replace the triangle with a Gaussian.
A Gaussian has data such as
Now combine millions of Gaussians with all their data, generated from a sample of images. That’s 3D Gaussian Splatting.
He also created a great 2-minute explanation and demo:
A lot faster than NeRFs. In this demo from Bilawal Sidhu, he was able to fly around a scene at about 100fps. The same scene as a NeRF would take about 10 seconds per frame.
You can still get photorealistic effects, such as surfaces changing color depending on the angle viewed at. According to Sidhu, “It doesn't use a neural network, but the training loop is similar to deep learning.”
Direct Editing: Select, move, delete, and relight the scene.
One thing - it can be resource-intensive.
Random: Is it a coincidence that all these 3D rendering models have names that sound like they came from the 90s? NeRF and Nerf. Splatting and Nickelodeon. Gaussian Splat sounds like a Double Dare or Guts challenge.
MASV is a faster, easier way to transfer large files to anyone, anywhere in the world—with just an internet connection.
Trusted by media and entertainment organizations across the globe, MASV plays an essential role in thousands of video workflows, giving production professionals absolute confidence in media delivery.
Unlike other file transfer services, MASV has no size limits and always delivers blazing-fast transfer speeds.
DALL-E 3 is coming. And it’ll be inside ChatGPT.
Sam Altman dropped the news that DALL-E 3 is coming in the next few weeks to ChatGPT Plus and enterprise users.
DALL·E 3 understands significantly more nuance and detail than our previous systems, allowing you to easily translate your ideas into exceptionally accurate images.
Altman’s tweet had a comic that he says all came from one prompt.
What really stood out was the prompt had some text that should be in the image…and the output rendered the text! Getting text to display has always been a huge challenge when working in Midjourney.
Besides being (what looks like) a much-improved model, one of the biggest improvements will be how you use it: right inside ChatGPT.
In the same conversation, you can generate images and text outputs. There’s a preview of what this looks like on the DALL-E 3 landing page.
But don’t count Midjourney out just yet. According to Nick St. Pierre, “v6 is apparently going to bring a HUGE leap in natural language understanding.”
🌐 Around the Web
🖼️ LimeWire, the infamous file-sharing tool for pirated materials, rebrands as a creative studio and acquires BlueWillow, a Midjourney rival
🌐 Schedule posts and track your social media with Metricool*
📆 Upcoming Events
XR, Virtual Production & Beyond
St Martins Courtyard, Coney St, York
Engage with experts and industry leaders on the latest trends in film and TV production, games, and software.
October 10 to 11 (other dates and cities available)
Sony’s Professional Display Solutions Roadshow
Sony Corporation, New York City
Meet Sony’s sales and engineering teams to experience hands-on demonstrations of unrivaled display and AV technologies and solutions
Istituto Marangoni Visual Design Courses
Multimedia design for integrated communication
Milano - Sept 2023 to Jul 2024
London - Oct 2023 to Jun 2024
Have a film-related event to share? Just reply back to this email and let us know the details.
Thanks for reading VP Land!