Ivona Tau
Family Album, My Grandmother’s Memories
Mar 20, 2025
Overview
“I trained several Stable Diffusion models in Python using the Dreambooth framework, allowing AI to transform these images and videos into new iterations of memory.“
- Ivona Tau
Key Information
Exploring family albums through the lens of memory loss, this project is derived from my personal photography archive spanning the 1960s and 1980s. Later in life, my mother experienced dementia, and her ensuing memory loss forms the heart of this work—my interpretation of how she perceived reality, often reliving her childhood. In this narrative, generations blend as her children’s memories intertwine with her own, allowing past and present to collapse into one.
Building on themes from my previous series, UnBeautiful, which challenged the pervasive misrepresentation of women's bodies in AI-generated imagery by embracing imperfections and questioning a narrow, male-dominated perspective, Family Album reclaims personal history. Both projects invite viewers to reconsider established narratives—whether by reinterpreting body imagery or by exploring the fragile, intertwined nature of memory and identity.
This project is a reconstruction of a past that exists somewhere between reality and imagination. It began with scanning dozens of old negatives that my grandfather shot in the 1960s and 1970s on Soviet cameras like the Zenit—many of which were never printed. I took hold of them after he passed away, as my family was cleaning his apartment, and uncovered a forgotten archive of everyday life: my grandmother with her two children (my mother and my aunt), vacations, skiing trips, quiet moments at home, and cityscapes of Vilnius, Riga, Zakopane, Nida, and Trakai. Alongside these, I used video recordings from his Japanese Jelco Zoom 8mm camera as additional training material, creating a dataset that became the foundation for this work.
In a way, this archive is not just a record of family life but also a recording of my grandmother’s memories—of her, through her, and around her. She was always present, either as a subject or a witness, living through each of these moments. But in the last decade of her life, after a car accident, she suffered from memory loss—most likely linked to dementia or Alzheimer’s. She no longer recognized us, mistook my mother for her sister, and lived in a version of the past that only she could see. I was both heartbroken and fascinated by the world she inhabited.
Now, through AI, I have reinterpreted this archive. I trained several Stable Diffusion models in Python using the Dreambooth framework, allowing AI to transform these images and videos into new iterations of memory. The results are strangely familiar—faces that resemble my family members but are never exact, moments that feel real but remain just beyond recognition.
AI is the perfect medium for this project. Its generative glitches—the fluid, shifting morphs, and unreal distortions—serve as an approximation of a fractured perception, mirroring the way memory erodes and reshapes itself over time. I find beauty in these imperfections, in the way AI hallucinates people and places that almost exist but never fully do. These digital artifacts echo the way my grandmother’s mind reassembled her past, creating a world that felt real to her but was built from fragmented recollections.
This project is a homage to her and her memories—both those that remained and those that faded. It explores how AI, like memory, distorts and reconstructs the past, offering glimpses into a world that is deeply personal yet ultimately unknowable.