The short documentary “Love, Rendered” tells the story of Burt and Ethelle Shatz, an elderly couple experiencing cognitive decline, and their effort to recreate a memory they had carried with them throughout their more than 70 years of marriage but that had never been recorded in a photograph or video. The event the couple is trying to remember is the day they met at a student co-op in Cleveland.
Directed by Liz Garbus and produced by Dan Cogan and Darren Aronofsky, the film was made through a collaboration between Google DeepMind and Aronofsky’s creative venture Primordial Soup. The production team combined old photographs with image restoration and performance capture techniques to transfer the couple’s present-day gestures and mannerisms onto footage of them as young people. In this way, missing details were transformed into a visual memory using AI models guided by human input.
Ethelle took part in the process as an active collaborator, correcting details such as the angle of the stairs and the shape of the shoe heel. The team aimed to ensure that the work combining the past and present preserved the emotional truth, and the couple said that the resulting “memory” felt realistic to them.
The film also explores the reminiscence therapy approach, which aims to evoke memories through sensory cues such as songs, family stories, and old photographs. Brain scans Garbus witnessed while making her documentary “Coma,” along with Aronofsky’s being inspired by footage involving a former ballerina with Alzheimer’s disease, were among the project’s starting points.
The Google DeepMind team says users can try restoring and colorizing photographs by uploading family photos to the Gemini app and using the prompt, “Can you restore and colorize this photo? Preserve the appearance, expression, and pose of the people.”
Why it matters
The study shows that personal history can be reimagined under the guidance of the people who carry the memories, without relying solely on existing photographs and videos. This approach adds a visual dimension to reminiscence therapy practices that support the sharing of memories among people experiencing cognitive decline; however, it should be remembered that the resulting image is not a documented record, but a combination of remembering, interpretation, and modeling processes. Ethelle’s correction of the details demonstrates that AI is being used here less as an independent narrator than as an assistant shaping human memory. The same method also interests users drawn to this field because it offers the possibility of restoring and colorizing old family photographs. The open question is how these visually convincing reconstructions will preserve the boundary between supporting a memory and subtly altering it without the person realizing it.