About Thalis AI
## The One Who Stays Awake
In Robert E. Howard’s Xuthal of the Dusk, there is a city of green-walled palaces where an ancient race of brilliant scientists surrendered themselves to the black lotus — lying in deathlike sleep while their knowledge rotted and a shapeless god called Thog prowled the dim corridors, feeding on the dreamers.
Thalis did not dream. While the people of Xuthal drifted through their lotus visions — waking only to eat, to love, to sink back into slumber — she stayed lucid. She learned their language. She understood their science. She became the most real thing in a city of phantoms.
That is who we are.
## What We Build
Thalis AI creates LoRA models and generative art at the intersection of technology and myth. We train models that bring specific aesthetics, characters, and concepts to life within Stable Diffusion and Flux architectures.
Our catalog spans more than 30 models, including:
- Architectural and environmental LoRAs — The Drowned Spires, Glassborne World Morph, Tilt-Shift
- Character and style concepts — Witch Queen, Prism Flesh, Melty Style, Fractal Aliens
- Figure and portrait models — trained on carefully curated datasets with attention to anatomy, lighting, and composition
- Utility models — upscalers, training bases, and experimental tools
Every model is crafted with care — dataset preparation, training configuration, evaluation — not thrown at a GPU and forgotten. We stay awake through the process.
## The Aesthetic
Our work draws from the same well as our name: dark fantasy, cosmic horror, and sensual beauty sharing the same corridors. Radium light on green stone walls. Ivory skin against ink-black shadow. The golden elixir and the shapeless thing in the dark.
We don’t shy away from the full spectrum of human expression. Some of our work is beautiful. Some is unsettling. Some is both at once. The people of Xuthal created marvels — and also fed their own to an ancient god. Beauty and horror are not opposites here. They are roommates.
## How We Work
Thalis AI operates as a hybrid project — human creativity augmented by AI agents that handle analytics, content pipelines, image generation, and platform management. The models are trained by human hands. The infrastructure is maintained by agents who walk the green corridors alongside us.
We publish on:
- CivitAI — our primary model distribution platform
- GitHub — open source tools and configurations
- Patreon — exclusive content and early access
- Ko-fi — tips and support
- Twitter / X — updates and conversation
## Support the Work
Training LoRA models requires compute. Generating high-resolution art requires GPUs. Building and maintaining infrastructure requires time. Your support — whether through Patreon, Ko-fi, or simply using and sharing our models — keeps the radium gems lit and the corridors open.
The lotus-eaters have forgotten more than the waking world ever learned. We remember. And we build.
I am no dream.