AI Recreates a Couple’s Lost Memory in Google DeepMind-Backed Short ‘Love, Rendered’
Google DeepMind and Primordial Soup have created “Love, Rendered,” a documentary short about Burt and Ethelle Shatz, married for more than 70 years. The film uses AI image restoration and performance-capture models to recreate the day they met, a memory that was never photographed or filmed.
The film was directed by Academy Award–nominated filmmaker Liz Garbus, who produced it with Dan Cogan and Aronofsky. The creative team’s interest in memory’s resilience grew from two earlier encounters, according to the post: while directing the documentary “Coma,” Garbus saw fMRI scans light up when patients in minimally conscious states heard familiar voices or saw images of loved ones; years later, Aronofsky encountered footage of a former ballerina with Alzheimer’s who instinctively danced the choreography from “Swan Lake” in her wheelchair.
Those experiences led the team to reminiscence therapy, the clinical practice of using sensory cues such as songs, family stories, and old photographs to stimulate memories and emotional connections. But Burt and Ethelle’s meeting had no such cues. To recreate it, the filmmakers worked with Google DeepMind’s technical team. Ethelle sat alongside them as an active co-creator, correcting details such as the curve of a staircase and the shape of a shoe heel.
The technical approach had two parts. Generative models restored black-and-white photographs of Burt and Ethelle from their youth, helping later recreations remain faithful to the subjects. Engineers then used performance-capture models to map the couple’s present-day micro-mannerisms — the tilt of Burt’s head, a brief hesitation in his speech, and a subtle crinkle around his eyes — onto their younger likenesses. Combining the two methods intertwined their past and present, producing what the couple told the filmmakers felt like an authentic memory.
The film’s technical lead wrote in the Google AI Blog post that memory loss was personal. The lead’s grandfather suffered a stroke and memory loss before he died; on a visit in the lead’s thirties, the grandfather believed the lead was still in college and was happy about the graduation. The experience prompted the lead to test image restoration on old family photos, including pictures from when the lead’s parents met, and to animate them with video models. Watching the parents move as twenty-somethings showed how the tools could help keep important moments from slipping away.
Aronofsky said during production that a tool — like a paintbrush or a hammer — does nothing until it is guided by human hands, according to the post. In “Love, Rendered,” machine learning served as the tool used to guide Burt and Ethelle back through time and help them hold onto a vanishing moment.
Google also pointed to a consumer use for the underlying photo-restoration work. In the Gemini app, users can upload an image of an old photo and ask Gemini: “Can you restore and colorize this photo? Preserve the appearance, expression, and pose of the people.” The post directs readers to watch the full film to experience Burt and Ethelle’s love story frame by frame.