.jpg)
I’ve been watching train videos this week every night before going to sleep. My wife and I put the kids to bed, finish the dishes, clean the house, do whatever else needs to be done before climbing into bed and shunting off onto Lorirocks777 Youtube channel, one of the many I’ve found that feature - exclusively - driver cab POV videos of trains crossing landscapes. These landscapes range from the snowy whitecaps of the Swiss Alps to the three kilometer long tunnels beneath Sarajevo to the peacefully humid-green villages of summertime Okinawa. There are thousands of these videos all offering the brain-swaddling fluid motion of a 4K rail-guided camera gliding through space. The sound is on low, a gentle hum.
I’ve found myself driving trains to sleep thanks to Artist Andre Oshea’s current AiR project, which examines how the same video (a POV train video, in this case) can be reinterpreted through two different stable diffusion setups—Automatic 1111 and ComfyUI. What’s fascinating is that even though these programs are built to serve as objective platforms for stable diffusion resources, they produce entirely different visual outputs when given the exact same inputs and settings. This project digs into how the tools we use, even those designed to be neutral, shape the content we create in subtle ways.
Curious to see the differences? There’s more below.
We have a new section on the home page featuring all of the LoRAs trained by AiR Artists! You can stay up to date with the newest resources being used in our artists' cutting edge projects, or dive deep to find older LoRAs from past projects. Check it out on the home page!
Civitai is pleased to present the latest collection of wearable art by artist Anne Horel! This week we are releasing a limited edition of 4 designs featuring Anne’s AI generated creature masks, Digital Felines. Made using an early DALLE2 beta model only released to a few artists, these masks form the basis of a larger body of work that she has expanded upon by training her own Flux LoRAs. Each tshirt design will be limited to 25 units, each priced at $30. Here they are:
Anne has officially been accepted into the Artefact AI Film Festival with her new film called Dips Fake! It is a fantastical french-language coming of age story of a potato dealing with the existential corkscrew of life. Go to the link below and vote for it now!
Finally, if you want to learn more about her art practice check out Anne's live stream from Thursday. We had a fantastic conversation about artists rights and identity in the context of AI, which will be expanded upon in Friday's Artist Round Table!
Rachel has been hard at work producing a series of Artifact-related LoRAs over the past few weeks, and the results have been nothing short of impressive. The images she’s created are striking in their fidelity, capturing not only the lighting but also the formal elements of archaeological documentation. Details like objects meticulously arranged on graph paper lend these images an uncanny authenticity, as if they were plucked directly from a historical archive.
Stone Tablet Burrows Cave LoRA
Her work ties beautifully into broader conversations around “Techno Primitivism” and the layered meanings behind false artifacts. In Issue 02 of Techno Primitivism, the Artist Journal Rachel recently released, we see a deep dive into how sight, as outlined by Foucault, has been privileged as the primary sense by which we establish proof and reality. Foucault argues that since the Classical age, perception has been systematically stripped of ambiguity, leaving sight as the dominant faculty for certifying knowledge, while taste, smell, and touch were deliberately marginalized for their subjective nature. This creates a precarious relationship between what we see and what we believe, especially in an era where images can be easily falsified.
Rachel’s work with these false artifacts—like archaeological documentation mimicking reality—dovetails with this critique of how authenticity is constructed through visual representation. The term “artifex,” meaning both artist and trickster, underscores the tension between what is genuine and what merely appears to be. The false artifact occupies a space where its believability hinges on how well it conforms to preconceived visual standards—just as Rachel’s LoRAs manage to invoke the visual language of archaeological documentation with an almost deceptive precision.
Once she’s collected enough images they will make their way into the final cut of her film about Burrows Cave.
Riccardo has developed several models of his own, incorporating wild and imaginative concepts as part of his work in the “Zone of Forty Howlers” (40H). One notable example is Iris Topography, a series of images that accompany a story he’s written about the topography of a person’s iris as a data storage mechanism, capturing and preserving an individual’s past experiences. This concept, which melds the personal and the speculative, exemplifies the surreal yet deeply philosophical nature of Riccardo’s work.
“The Zone of the Forty Howlers - Anthology of Non-Cartographic Representations of Space” (40H) is an ambitious, long-term research project currently in its early stages. It is conceived as a collection of “fictional geographical moments,” forming an encyclopedia of textual and visual narratives that blend reality and imagination. These narratives serve as pearls on a string, interweaving ephemeral relationships between humans and territories across various cultures and historical moments—both real and imagined.
At its core, the project seeks to redefine geography, positioning it not just as a physical science but as a geosophical practice, a space to explore the philosophical, creative, and visionary potential of both inner and outer landscapes. This multidimensional approach draws on interdisciplinary fields, ranging from botany and archaeology to anthropology and religion, fusing them into compelling and speculative spatial relationships.
A key element of the project is its iconography. Riccardo’s use of AI-generated images helps blur the line between reality and fiction, adding depth and texture to the stories. These images include fictional archives, imaginary archaeological finds, fantasy ethnographies, and even photographic records of nonexistent places and cultures. This visual storytelling enhances the project’s overall narrative scope, transforming it into a multifaceted investigation of space and the human experience within it.
The core of Andre’s project is to take the same video and run it through two different programs—ComfyUI and Automatic 1111—while keeping all resources, settings, and parameters identical. By comparing the outputs, he aims to highlight how seemingly neutral tools can produce divergent results. To extend this exploration, Andre will upload the resulting videos to two distinct social media platforms, posting daily on both Instagram and TikTok, further emphasizing how platforms influence perception and engagement.
His work challenges the notion of objectivity in both the tools we use and the platforms we rely on. In an era where, as Jerry Saltz recently claimed, “Art Criticism is dead,” and where everyone has an opinion but no definitive answers, it becomes crucial to reexamine our reality by scrutinizing the mechanisms that shape it. In this case, Andre is probing two of the most popular programs in generative AI to reveal how the tools we trust impact the very content they produce.
Here are a couple of initial tests, running a POV car video through both programs with the same settings, seed number, prompt, and resources:
In discussing these tests I uncovered Andre’s hidden fascination with trains. He’s an avid viewer of train videos in his spare time. In the same way that one might put on the news or a college football game as background noise, Andre opts for train videos. The repetitive yet calming nature of trains in motion became a subtle metaphor for the project’s exploration of technological differences. Watching trains has long been associated with a hypnotic quality, one increasingly also associated with social media via doom scrolling, etc.
This led me to recall the well known 50 second film “Train Pulling Into the Station” (1896) by the Lumière brothers, a cinematic milestone that, much like Andre’s project, questions how we perceive technological advances. At the time of its release, viewers were reportedly shocked by the realism of the arriving train on the screen. Today, while we’re no longer startled by such simple images, Andre’s work reminds us that technological tools can still shape our perceptions in profound ways. The subtle distinctions between the outputs of ComfyUI and Automatic 1111, much like the arrival of that first screenbound train, challenge us to look closer, to question what is real and what is mediated by our tools..
In both cases—the early cinema experiment and Andre’s use of gen AI—what’s most striking is the examination of objectivity: how the same ‘input’ (a train, a video) can generate a different reaction based on the medium or tool delivering it. By choosing to upload to two separate platforms, Instagram and TikTok, Andre continues to push the inquiry into how our tools shape not only the output but also how it’s received by audiences on different platforms, each with its own algorithms and viewership dynamics.
Chris has produced a bundle of LoRAs this week, at least 6 now. He will be implementing these in his total redux of 2001: A Space Odyssey, interpreting it through style transfers of the major visual language shifts over the past 100+ years and beyond. A few samples below:
He’s also been refining his Comfy workflow by meeting with a few other AiR artists as well as one of our Creator Club members. Saftle, special shout out and thank you for doing a deep dive on your vid2vid workflow with Chris!
We look forward to sharing tests from these sessions soon!
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)