Studio Update Three
April 16, 2024 3:38 AM
Chris Adler
April 16, 2024 3:38 AM

At the close of the sixth week, Arvid and Noah are nearing a midway point while Domenec is closing in on completing his work (and a bit ahead of schedule, to boot). As we get closer to wrapping up our first artist residency, questions of how to present and archive the works produced are front of mind.

Domenec Miralles

Domenec’s Infinite Mesh site is nearly done. You can find it currently hosted on his artist website here: (!) There are still a few last minute tweaks that we want to introduce, such as giving each story its own URL slug so they are easily shareable. But, we are now looking towards the final step of moving the work to the AiR site for permanent archiving in our Gallery section. Over the next week we will be working with our stellar AiR website designer, Sufiyan, to figure out how best to integrate Domenec’s work into the AiR site.


We want the Gallery section of the site to function as a vitrine containing artists' finished works, archived in perpetuity for future viewing. As the first work going into this new archive, the interactive nature of Infinite Mesh presents a fun challenge: how do we organize this archive visually so that the AiR site is able to hold all types of content and experiences, while still making sense as a navigable and consistent archive of past artists’ work. This presents somewhat of a UX challenge. It’s actually ideal that this oddly shaped project should be the first to enter the archive, since it forces the boundaries wide open from the outset. For example, if we had designed it from the beginning to just hold images, perhaps we would run into issues hosting a project like this in the future. Of course, the primary goal will be to maintain the fidelity of the work as it exists currently on his site, keeping it open and interactive for years to come. I like to think about the way a work like this might age. It's both at the cutting edge of what's possible today, but also harkens back, embedded stylistically with a 90s retrofuturistic nostalgia. The LoRA he trained for the work also just creates wonderfully gorgeous imagery.

Since the core function of the work is to give people the ability to create a story unique to them, we are considering the idea of treating the individual stories produced as an open edition. This borrows an idea from photography that really makes sense within the context of generative art. Archival prints are often offered as an open edition to encourage the accessibility of a work, as closed editions tend to be much more expensive. But this sort of turns the photographic notion of an open edition on its head since this isn’t an edition of a single work, but an infinite edition of unique works produced by an artist-created tool. This agential function truly embodies the open-ended nature of generative art, and is indeed an idea at the heart of the work. To further support both the Artist and the Creators (we're not just Viewers!) who engage with the work to create these digital pieces, we’ve discussed adding the ability for creators to add a unique Certificate of Authenticity to log the creative origin and future provenance of the works as they are shared out into the world (again, infinitely on into the future).

Noah Miller

Noah has completed his entire Animatic which now comes in at a staggering 22 minutes. He has also completed a full rewrite of his script v2 and is now tasked with cutting the film down to within his 12-14 minute target run time. This will be no easy task, since the film traverses so much emotional ground in such a short period of time. You can check out the full animatic on vimeo now, posted below. This is basically a shot for shot sketch of what his final film will be once he actually goes into production, processes, edits, and scores it. It already feels so complete!

Noah has settled on a gorgeous anamorphic 2.35:1 aspect ratio, a nod to the true cinematic soul of this film. It’s also a way to maximize the real estate of his current 1.5 workflow, since his SDXL workflow was getting a bit glitchy and unreliable. Not to worry, once the film is “in the can”, there will be plenty of AI upresing after the fact to get to his final output screener.

Noah is also in talks with a certain LA film school about offering the production of this AI film as a summer class. This could potentially solve for location, gear, and labor all in one fell swoop. Of course, with the limited financial support we can offer, we are still having to get a bit scrappy with the production of his project. Since academic summer will soon be upon us, a summer course that offers credits to students for participating in the production in May/June could be a perfect fit.

An update on Noah’s full “LoRA stack”: each character will get a separate LoRA for every costume/setting they appear in. Tara will get one LoRA for her indoor scenes, and another for her Outdoor scenes. Rick will get his own for indoor clothing, and though he does venture out into the arctic.. well you’ll just have to watch the animatic to find out why he only needs the one :).

To make these LoRAs Noah will stage each actor in front of a green screen in full wardrobe for each setup. Then, he will capture still images of the actors from all angles, rotating a three point lighting setup at the same time to capture different lighting impressions. These still images will then be used to train his character LoRAs.

We’ve also been discussing the possibility of releasing a full collection of every AI resource Noah uses to create this film on Civitai. To our knowledge, this would be a first in filmmaking, and would effectively kick off a whole new category of cinema: Open Source Filmmaking. The collection could include all of his custom character LoRAs, scene/worldmaking LoRAs, character voices, the full script, etc. This would allow fans to expand the storyline by using these resources to make their own films set in the ZERO cinematic world. There are 52 known structures, and this is just the story of Structure Zero. I wonder what happens after Tara enters Structure Zero, or how Zero is related to Structure 1, 3, 8, 45? Open Source Filmmaking could allow for the Rhizomatic expansion of IP like this cinematic universe in a truly community-driven way.

Arvid Tappert

Behold, multiple videos in this update! Arvid has recently been collaborating with another animator to expand the ODDBIRDS universe. They are leaning into the royal kingdom setting with a new work featuring a queen being courted by successive suitors. This is essentially his ODDBIRDS take on tinder in the middle ages. One suitor tries dancing for the queen, another gives her a flower, and the next brings her a fish. It’s an endearing setup that captures the essence of Arvid’s approach to making work. His animations always seem to be embedded with a wacky honesty that examines the odd superficialities of everyday life. It’s Wes Andersen meets Wallace and Gromit. While this specific new video isn't available yet, check out a few below.

To date he has completed 4 “situations” short films. This new work is likely to be a bit longer, closer to 40 seconds or a minute. Even with the new workflows he’s created, which according to the artist speed things up by ~50x over traditional animating, it still takes 4-5 days to execute a short in his universe of characters.


His process is a very good example of weaving AI into a traditional workflow at every step. It is emphatically NOT a simple text to video prompt. First, he sketches the birds out roughly in Krita, which uses their built in AI diffusion workflow to translate them into real ODDBIRDS using a custom trained LoRA. He does this 5-10 times, drawing them all from different views, front, back, side, top etc, then takes these 2D images and brings them into blender, which merges them to create an AI 3D mesh. He animates the birds in Blender, then drops them into the background scenes he’s created in Krita, using ample inpainting to achieve the specific scenario he has in mind. He’s also been playing around with Mixamo, a free adobe plugin that allows you to do motion capture to animate anything. Also, here’s the full version of his AI Erase Workflow Demo mentioned in the last Update.


More to come from Arvid soon!

Studio Update Three is a wrap. Look out for the final launch of Domenec’s tool and for new Artist Shinji Murakami to join us starting next week!

Post link
©Civitai 2024