Reimagining the Artists: Using AI to Create Historical Portraits of Ukiyo-e Masters

September 14, 2026

When we study Japanese woodblock prints, we become very familiar with the faces depicted by artists such as Hiroshige, Kunisada, Kuniyoshi, and Yoshitoshi. But what did the artists themselves actually look like?

It is an interesting question because for many of the great ukiyo-e artists, our visual record is quite limited. Most lived before photography became widely available in Japan. Others were photographed late in life, while many are known to us primarily through portraits created by fellow artists.

Over the past several months, I have been experimenting with artificial intelligence as a way of approaching this problem from a different direction: Can historical portraits, prints, drawings, and early photographs be used as source material to create a plausible photographic reconstruction of an artist?

Katsushika Hokusai - Photo Portrait

Why Create a Photographic Portrait?

There is something immediate about a photograph. A woodblock portrait can tell us a great deal about how a person was perceived, but a photographic image creates a different kind of connection. Suddenly, someone who lived 150 or 200 years ago can feel contemporary.

That is especially useful when teaching or writing about ukiyo-e. When I introduce an artist such as Hiroshige or Kuniyoshi, being able to put a human face alongside the artwork helps transform the artist from a historical name into a person.

But there is an important distinction: these images are not photographs of the artists, and they should never be presented as such. I think of them instead as AI-assisted historical reconstructions.

Utagawa Kunisada - Photo Portrait

Working With AI Rather Than Simply Generating an Image

The process is not as simple as uploading an old portrait and asking AI to “make this look like a photograph.”

For each artist, I try to locate as much reliable visual evidence as possible. That may include portraits made by other artists, self-portraits, and – in the case of later artists – actual nineteenth – or early twentieth-century photographs.

I then work back and forth with AI, comparing the generated face with the historical sources and making adjustments. Facial structure, age, hairstyle, expression, clothing, and other details may go through numerous revisions. AI provides the image-making capability, but I am continually evaluating the result against the historical material.

The goal is not to create the most attractive or dramatic portrait. It is to create an image that, based on the evidence available, seems as plausible as I can make it.

Tsukioka Yoshitoshi - Photo Portrait

Not All Reconstructions Are Equal

For artists for whom historical photographs survive, the process is relatively straightforward. A photograph gives us direct evidence of facial proportions and features. AI can be used to restore detail, improve clarity, or create a more natural photographic presentation while remaining closely anchored to the original image.

Reconstructions based entirely on woodblock prints, paintings, or drawings require considerably more caution. Artists did not necessarily intend their portraits to function as precise records of physical appearance. Ukiyo-e was a stylized art form, and portraits could simplify, exaggerate, or conventionalize facial features.

When several independent portraits show similar characteristics, however, they begin to provide useful clues. The shape of the face, nose, eyes, hairline, facial hair, and other recurring features can help establish a visual framework.

These reconstructions should also be considered works in progress rather than definitive portraits. Our understanding of an artist's appearance can change as additional visual evidence comes to light. A newly discovered photograph, drawing, or portrait – or simply a better-quality image of a source we already know – may reveal details that weren't previously apparent. When that happens, I can revisit the reconstruction and adjust it to reflect the new evidence.

I would also welcome contributions from collectors, researchers, museums, or anyone else who may have photographs or portraits of these artists that I haven't seen. The more independent visual sources we can compare, the better chance we have of creating a reconstruction that more closely approximates the artist's actual appearance. If you have an image that might help, please feel free to contact me through Mie Gallery.

Kawanabe Kyosai - Photo Portrait

The Artists

So far, I have created or worked on photographic reconstructions of Hiroshige, Kunisada, Hokusai, Kunichika, Yoshitoshi, Torii Kotondo, Kuniyoshi, Imao Keinen, Kawanabe Kyōsai, Natori Shunsen, Utagawa Yoshiiku and others.

For each artist shown in this article, I have included some of the historical source images alongside my reconstructed portrait. This allows you to see the evidence and, perhaps more importantly, decide for yourself how convincing the reconstruction is.

Portraits Have Never Been Entirely Objective

It is also worth remembering that altering or enhancing a person's appearance in an image is hardly something that began with artificial intelligence.

Today, filters and automated enhancements are commonplace, particularly in photographs posted to social media. With a few taps, we can smooth skin, remove wrinkles, brighten eyes, change lighting, and make dozens of other subtle – or not so subtle – adjustments.

But photographic retouching itself is almost as old as photography. Long before Photoshop, photographers and retouchers manipulated negatives and prints by hand to remove blemishes, soften features, improve complexions, and otherwise present their subjects in a more flattering way.

I remember one of my first jobs as a young designer in the 1990s. I spent about six hours manually retouching a photograph of Vanna White from Wheel of Fortune fame – removing wrinkles, brightening her smile, and even adding a little glint to her eyes. What can now be accomplished almost instantly required hours of painstaking work.

And the practice goes back much further than photography. Portrait painters and sculptors have always interpreted their subjects rather than simply recording them. Renaissance artists and those who came before them could idealize a face, improve a complexion, adjust proportions, or emphasize the qualities they – or their patrons – wanted a portrait to convey.

Japanese artists were no different. A portrait in a woodblock print is not the equivalent of a passport photograph. It is an artist's interpretation of another person, filtered through the conventions and aesthetics of the period.

That becomes particularly relevant to this project. When I use an ukiyo-e portrait as the basis for an AI-assisted reconstruction, I am not starting with an objective record of someone's appearance. I am starting with another artist's interpretation and asking a modern image-making tool to help translate that interpretation into something resembling a photograph.

That is one reason I believe it is important to describe these images as historical reconstructions rather than historical photographs. They can help us visualize the past, but they should never be confused with the surviving evidence itself.

Utagawa Kuniyoshi - Photo Portrait

AI as a Tool for Visualizing History

AI-generated historical imagery raises legitimate questions about accuracy. It is becoming increasingly easy to create a convincing image of something that never existed, which makes transparency especially important. My intention with this project is therefore not to replace the historical record, but to visualize it.

The original portraits, photographs, prints, and drawings remain the evidence. The reconstructed photograph is an interpretation based upon that evidence. We may never know exactly what Hokusai, Hiroshige, or Kuniyoshi looked like. But by bringing together the surviving visual evidence and using new technology carefully, we may be able to get a little closer.

A Note About Using These Images

The reconstructed portraits created for this project are available for others to use for educational, editorial, or personal purposes. If you use one, I would appreciate a credit such as “AI-assisted historical reconstruction by Mie Gallery.” Please contact me if you'd like a larger version of any of these portraits.

Most importantly, please identify the images as reconstructions rather than historical photographs, particularly when no original photograph of the artist is known to exist.