written by
Ruben Verhack

Reflecting on the first AI for Digital Arts/Entertainment/Game Dev meetup

Meetups 4 min read , April 11, 2024

The event

Last night, Datameister, Howest Digital Arts & Entertainment and Flanders Game Hub organized the first event of the AI in digital arts game development and entertainment BE meetup community. It was hosted in the beautiful event space underneath our offices at the Boldhouse in Ghent, BE. It was great to see so much enthusiasm and attendance to the event. Full House!

Full house at the Boldhouse!

The goal is to connect people ranging from AI algorithm techies to AI-powered creatives. Both ends are really looking at each other for help since there is so much noise at the moment. It feels like the entire playing board has been completely shaken up. The techies wonder what to make, while the creatives wonder what is possible. We believe that those who take the lead in navigating this new future will be the ones who ultimately own it.

The people

I was honestly pleasantly surprised by the creative talent that was present. In general, there is always more talent in Belgium than you would expect. People came from far and beyond and were showcasing their experiments to each other on their laptops. That really brought a smile to my face.

It was a nice mix of people, and I have to give a shoutout to Howest DAE Research for being the ideal bridge between the creatives and techies. Researchers and teachers were all present to share their findings and look for collaborations.

Apart from the creatives, it was great to see other industries present. People who work on photogrammetry and 3D assets in industrial settings, work with 3D people developing their new AI products in stealth mode, or other people working on all kinds of simulations and training environments in AR/VR.

The talks

Glenn Van Waesberghe (DAE Research) gave a great overview of how text-to-image works, not shying away from technical details like the embedding spaces and diffusion, something I really appreciate because I don't want to be fluffy about AI in this community. Give me the cold, hard, technical details. He concluded his talk with a comprehensive analysis of all the available tools, evaluating them on various aspects. Glenn shared his slides here.

Glenn presenting text-to-image: deep dive

Vince Buyssens (Starhaven) showed us how he has been successfully applying AI for big international campaigns and discussed how it was received. He specializes in supplying the change management necessary for companies and all the nuances of how to apply new techniques to broaden your toolbox but still respect your audience and keep humanity relevant within the AI revolution. He was kind enough to share his presentation slides. Check out his videos on which he collaborated: “Welcome to the Latent Space”, “Under Armour - Anthony Joshua - Forever is made now”.

Me, Ruben, gave a talk on AI-driven breakthroughs in image-based rendering which I already wrote extensively about in this blog post.

My reflections

First of all, I found it extremely interesting how tools, which are still very limited in quality or creation length, can be put together to create something great. It’s a great new type of creativity that emerges, which often reminds me of the 80s and 90s aesthetics that also tried to be as creative with shitty graphics as possible but gave rise to 8-bit art etc. It itself becomes an art.

Another trend I am seeing is that the modalities on which creatives will work in the future will be volumetric; it won’t just be images and video. Some studios already treat video productions a lot more like 3D environment productions than plain 2D video productions. For this, NeRFs and Gaussian Splats will become great intermediate formats that bridge this camera-captured data with 3D worlds. We will work in media in a different modality akin to radiance fields in a more volumetric environment, and this will require radically new tools. This is exciting for a company like ours that specializes in spatial AI.

What I learned was how creatives now have access to open source, and how open source is so important to creatives. The idea that this technology should not be only dominated by Adobe and the likes is actually a stance on open source that I never thought of. For me, as a techie, I always assumed open source was something us techies only cared about. Especially ComfyUI caught my attention.

Last observation was that creatives seem to care a lot more about the impact on their industry than I have ever given thought of. And with a lot, I mean A LOT. The criticism and even boycots company face when experimenting with AI is mind-blowing. They need to deal with inertia a lot more than us techies need to deal with. It opened my eyes to what kind of techno-optimistic bubble I live in.

Next steps

This event placed a significant emphasis on text-to-image generation, image in- and out-painting, and generative video, along with our discussion on radiance fields, including Gaussian Splatting and NeRFs. I am really looking forward to discovering what other topics will emerge. I am particularly excited about those that involve 3D pose, animations, 3D world design, and visual effects.

We already set the date for the next event, being June 5th. More details will follow, follow us on our meetup community page.