How the Eranshahr team is thinking about and using AI

Eranshahr is a research and production studio. Our goal is to make Iranian culture, history, and mythology accessible to a global audience. Through various projects, our team uncovers and reinterprets cultural knowledge, sharing it as free online resources and through physical books and games available in our shop.

How we work

Eranshahr operates as a volunteer-based research collective. Teams gather virtually from around the world to document, visualize, and translate different aspects of Iranian culture, including mythology, folklore, history, language, and art. Over the years, hundreds of contributors have helped build and refine our open-source cultural archives and digital projects.

Our use of AI

Our first major project focused on describing Iranian Mythical Creatures. In 2021, a small team of researchers and illustrators met daily with the goal of publishing one creature every single night.

To keep that pace, we had to find a way to make the illustrations look uniform. The first 150 were done in Adobe Illustrator and textured with a shared Photoshop template. When we decided to turn these illustrations into a book, we wanted a more cohesive look that felt like a medieval manuscript, specifically that mix of Persian miniature painting and old astrological diagrams.

At the time, generative AI was just starting to pop up. We were excited to experiment with it as a cutting-edge tool. We tried uploading styles from open-source libraries, but standard techniques (like LoRA) didn't give us the results we needed. We eventually used "IP adapters" because they were better at understanding the style of our reference photos and combining them with the forms we had already hand-drawn.

It was far from a "one-click" process. Since the tech was so early, the images were full of noise and mistakes, especially on hands, paws, and faces. We spent thousands of hours in Photoshop cleaning, collaging, and reworking each image. Even though it was extra work, we felt the final output was much more beautiful and fit the theme of the book better.

We also tried more user-friendly tools like Mid-Journey and DALL-E, but they were never able to understand "Persian" art styles. They likely weren't trained on Persian miniatures, so they never gave us a satisfactory result.

Ethics and Perception

We’ve talked a lot about the ethics of using AI. We weren't worried about "stealing" from other artists because the forms were our own original drawings and the styles were taken from open-source medieval and Mughal manuscripts.

The bigger worry for us was perception. We feared people would think this was a cheap generation anyone could do with one prompt. In reality, it took 1000s of hours of human work to modify, stylize, and clean up these images. We concluded that using AI as an "accelerator" allowed us to create a better, more unified experience for the reader.

Looking back now, with a better understanding of AI’s repercussions, we might not have gone down this path. We talk regularly as a team about ethics and what responsible creative output looks like as we move forward.

Stay connected

Get updates on new projects and products.