In Japan, there are female divers known as “Ama.” They dive without regulators, using only their breath to catch fish.

Depending on the region, they would sometimes dive to depths of several dozen meters. Although it was a life-threatening endeavor, in some areas, the income was so high that a single Ama could support a family of several people on her earnings alone.

While modern Ama generally wear wetsuits and other gear when diving, until the latter half of the 20th century, they mostly dived naked. In particular, the Ama of Hegura Island in the Noto region wore only “Saiji,” a loincloth-like undergarment. This was a “work uniform” that had been lightened to the absolute minimum to reduce water resistance.

The saiji was handmade underwear crafted from bundles of cloth, featuring a triangular patch in the crotch area and a completely string-like, thong-style design at the back. The waistband consisted of several strands of twisted cloth. The women would tuck a metal scraper called a “kaigane” (weighing about 3 kg), a metal weight called a “Hachiko,” other tools, and a lifeline connected to a boat on the surface into this area.

The Hegura ama divers would descend to the seafloor, dozens of meters below, and use the kaigane to pry abalone attached to rocks loose and harvest them. They worked as long as their breath held out, and once the harvest was complete, they would pull on the lifeline by hand. On the boat, their fathers, younger brothers, and others stood by. When they sensed movement in the lifeline, they would pull it up with all their might to help the ama resurface. It was difficult for an ama to resurface quickly on her own, and if the family members on the boat neglected their duties, there was a risk of drowning.

The men of Hegura went out to work in the deep-sea fishing industry upon reaching adulthood, but this was dangerous and physically demanding work, and they often lost their lives. Consequently, it was customary for the elderly—who could no longer go out to sea—and children to assist with the ama’s work.

This LoRA model was designed to recreate the ama of Hegura. It was trained using photographs taken in Japan by Western photographers in the 1960s, as well as images from the 1930s Japanese photography magazine Asahi Graph. However, due to the limited training data and image quality that would not meet AI learning, standards, the costumes cannot be considered a perfect recreation. The details of their traditional goggles are vague, and the distinctive ear flaps are missing. There are also some ambiguities in the details of the traditional swimming caps and Saiji they wear on their heads. Please consider the historical accuracy this LoRA can achieve to be around 50%.
Additionally, faces were excluded from the training data as much as possible. This LoRA is not designed to reproduce any specific individual, nor was it trained on the features of any particular person.

There are many materials I haven’t yet obtained, and I’m searching for them at the Library of Congress (LOC) and at auctions. If I manage to acquire better supplementary materials, I may update the LoRA.