What AI model do you use?

Our model is a fine-tuned diffusion model built on an open-weights base architecture, with additional training on a curated corpus of fine-art imagery that's either licensed or in the public domain. We don't train on scraped social media, copyrighted living-artist works, or opt-out-required data.

The model is paired with a human art-direction layer. Every order passes through at least one of our designers, who iterates on prompts, checks composition, and makes refinement decisions before the image goes to print. This is where the human creative authorship enters the pipeline — and it's what lets us stand behind the final piece as a work we're proud of.

We don't name a specific model version publicly because we rev it frequently, but we'll tell you what your specific order was generated with if you ask — we log model version, seed, prompt, and designer on every order.