Chatting With The EmProps.ai Team
I had the chance to interview the team behind one of the hottest AI art platforms...
Hello friends,
Today, I am thrilled to introduce a groundbreaking platform dedicated to the world of generative AI art on the blockchain. This innovative marketplace has witnessed the release of exceptional AI series in recent months, and I can't wait to delve deeper into their mesmerizing creations.
What could be better than interviewing the team behind it?
Could you tell us, in short, what is Emergent Properties (EmProps.ai)?
Emergent Properties is a curated platform that enables collections of Long-Form AI Art. We have big plans for growth and expansion, but at the moment the simplest description is that we’ve combined the fields of Long-Form Generative Art (i.e. ArtBlocks and fxhash) and AI Art to create a new medium: Long-Form AI Art.
How did you come up with the idea?
First for a little backstory: we originally launched EmProps in January 2022 as a generative art platform, without any implementation of AI. This was on a blockchain called Secret Network, which has native privacy-preserving contracts. Our idea back then was to be a hybrid of ArtBlocks and fxhash with unique token utility and artist/collector relationships empowered by the privacy characteristics of Secret Network. But we quickly pivoted once it became clear that we needed to be operating where the artist and collector communities were already established.
Fast forward to the fall of 2022, we were slowly working on rebuilding the platform and transferring to Tezos, but we didn’t really have something unique to add the ecosystem. That was, until two of our founders, Emil and Ira, both of whom are artists themselves, started going really deep into AI on their own work.
Ira initially raised the question of how we could automate the creation and minting processes through these burgeoning AI Art platforms, and that led to a series of conversations that became our eureka moment – where we realized that we could take the modular generative pipeline that we had built for our existing platform, and add an AI generator to it.
Could you describe, technically, how it all works?
When a collector mints, they immediately receive a token in an “ungenerated” state. This just means that there is no artwork yet, only a unique seed and the instructions for creating the artwork. The tokens are then placed in a queue to be generated. One-by-one on our servers, we execute what we call our “Generative Pipeline” for each token. The seed of each token is used as the input for a series of processes we call nodes. Each node is a standalone process, be it pure javascript, p5.js, or an AI model like Stable Diffusion or GPT. To make the process as fast and smooth as possible, we employ a group of servers that are the most powerful type available for AI processing.
The most basic version of the Generative Pipeline comprises a javascript node and an AI node. The javascript acts as the generative controller, where all the prompts, image inputs, AI settings are defined and randomized. This allows for a huge range of variability for what actually gets input into the AI model, and this is how the AI outputs become not just randomized and varied, but also generated dynamically at the time of mint.
After the generation process is complete, we save the asset(s) to IPFS and update the token metadata with all the info the pipeline generated. We then commit the changes to the blockchain, and the collector now has a fully generated artwork.
Do you use any platform to handle AI models (stable diffusion, midjourney, etc)?
Yes, AI models are available to the artist as nodes in their custom pipeline. We currently only offer Stable Diffusion, but we are very shortly going to be expanding to other generative AI models including LLMs like GPT and alternatives to Stable Diffusion such as Dall-E.
Midjourney is a closed product at the moment so we do not offer that option, but we would love to partner with them to bring Midjourney support to EmProps. We are really excited about the announcements coming from Stability AI about Stable Diffusion successors and we will be implementing those as they are available.
What does it look like for the artists?
A big part of our internal work is building better tooling for our artists, so the answer to your question is sort of a moving target.
For The Oracles through The Muses, the process was very manual and collaborative. Depending on how technical the artist, they gave us code in p5.js or three.js, or just explained their idea and we wrote the code. Because the input to the code itself was the seed, the input to Stable Diffusion was random/variable and deterministic.
More recently we released a javascript library that allows the artist to fully control the pipeline and test outputs while coding. For non-technical artists nothing will really change, but for technical artists they will soon have a full environment to code in and a set of processes they can use to test outputs. They then pass over what they built and we deploy it. This process allows the artist to be 100% sure the generator is following their instructions and yielding desirable results, and it doesn’t require our team to be directly iterating with them.
In the future, we will release an ambitious set of tools that allows any artist to build projects with or without coding experience in a highly intuitive UI/UX. We hope to get an early MVP of this in the hands of testers very soon.
You have had a great reception so far; why do you think collectors are interested in this new platform?
As mentioned above, Emil and Ira were using AI in their own artwork, and both of them, along with co-founder Victor, have been active in the generative art space both as artists and collectors. That “close-to-the-ground” energy of understanding what’s happening in the space and intuitively feeling what’s next and/or what might be missing, is probably how we achieved strong product-market fit straight away. It’s like this idea was in the water, and someone just needed the vision and skill to bottle it.
It also helps that the artwork is so great. We like to think that the novelty of this new medium is tangible in the artwork, and that collectors can feel how special it is.
How do you curate AI artists to participate?
At this point, our process is fairly informal — primarily a collective decision among the EmProps team. We expect to formalize this process in the near future, including adding outside perspectives and community curation. But right now, the most important criteria is finding diverse artistic visions that embrace the unique challenges and opportunities of this new medium.
What are you focusing on at the moment?
At the moment, we’re focused on three things:
Attracting and supporting the most interesting artists and projects.
Improving the artist’s tools, with an eye toward making them increasingly accessible.
Fundraising to support our vision for growth.
What has been your biggest challenge so far?
We’re a relatively small group working with very limited resources (we’re self-funded, and we all have day jobs). We have an extremely ambitious and exciting roadmap ahead of us, but the biggest challenge is finding the time and resources to build as fast as we want to.
What’s the team at the moment? Where are you located?
We’re a team of five co-founders based in New York (Sandy), Los Angeles (Emil), Dallas (Ira), and Ensenada, Mexico (Luis & Victor); plus a small and growing advisory board that we will announce soon.
There is no question about the AI art potential, and I can’t wait to see what EmProps.ai looks like a few years down the road.
Until next time,
- Kaloh
If you enjoyed this interview, feel free to click the ❤️ button on this post so more people can discover it on Substack 🙏