Research
September 27, 2024

NEX V1.0

Heading 1

Heading 2

Heading 3

Heading 4

Heading 5
Heading 6

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.


        const greeting = "Hello, world!";
        function sayHello() {
            console.log(greeting);
        }
        
Block quote Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.

sdsds

Ordered list

  1. Item 1
  2. Item 2
  3. Item 3

Unordered list

  • Item A
  • Item B
  • Item C

Text link

Bold text

sdsdsd

Emphasis

Superscript

Subscript

It was only last year that AI became capable of creating high-fidelity digital media. The technology has improved significantly since then with the release of Midjourney v6, DALLE-3, and SDXL.

However, there are still challenges with the models of today. In particular, all of them are primarily text-driven! This hinders artistic control and it's often hard to get the outputs that you want.  We believe that controllable, steerable AI will usher in a new wave of human creative expression, and NEX will be the place to build and house it.

Today we've made progress towards that goal with the release of:

  1. PRIMUS-1, a powerful multimodal model for image synthesis and
  2. The NEX platform, a new creative interface that makes it easy to create, remix, and organize content

PRIMUS

We’ve developed a brand-new foundation model for image synthesis with built-in multimodal controls. We’ve trained it from scratch— which means that our model has an entirely new aesthetic style and set of capabilities.Here are some samples from our model:

Primus has excellent visual quality at native 1024x1024 resolution, on par with state of the art models like SDXL. We've trained this on a cluster of A100s using techniques to achieve linear scalability and high model-flops utilization.

But that's not all.

We’ve also incorporated native color control into Primus, which you can use by uploading an image with the color palette you want. Our model extracts its palette and fuses it in the generation process to create an output with the colors in your input image.

Primus also has native depth control: you can upload an image with a structure you want to preserve, and our model will preserve it while change it's semantics.

In the example above, we've turned some flowers into cherry blossoms on a lake, while keeping the flower stem & body intact— and in the color example, we've transferred the color palette from one of our favorite films into our output image! Moreover, these conditions can be fused, so you can use depth + color together!

We're can't wait to see what you will create with these new capabilities!

The Platform

AI has to be accessible for it to be useful— so we’ve built a brand-new platform for you to easily create, share, and organize your content. We looked at some of the most popular tools out there and asked ourselves, "How can we make this easier?"

We are unveiling three primary features today:

  1. You can create content with Primus in a simple session interface
  2. You can share sessions between friends and colleagues to show off your work
  3. And you can add your favorite images to descriptive, private collections

Here's a collection that I put together recently:

Yummy 😋

With these features we hope to make it as easy as possible for both experts to leverage the power of multi-modal fusion and for newbies to create and remix the content they want.

More to Come

We've released a lot this month but have much more in store. Like the computer and the camera before it, AI will be a democratizing force for creativity, and we believe that new communities of expression will form around these new tools. We'd love your feedback on where we should go from here.

Fundamentally we’re big believers in human storytelling and want a world in which everyone is empowered to tell great stories with technology. If this resonates with you, please send a note to me at darius@nex.art ! Today we’re a small team based in London and Palo Alto, and we're looking for experts to join our AI and Application teams.

Sincerely,

Darius

12/01/23