Love, Rendered · September 9, 2026
Love, Rendered Uses Performance Capture and AI to Rebuild a 70-Year-Old Memory
Google DeepMind · Behind the Scenes
What happened
The documentary short follows Burt and Ethelle Shatz, an elderly couple confronting memory loss, and recreates the unrecorded day they first met. Directed by Liz Garbus and produced by Dan Cogan and Darren Aronofsky, the film was created with Google DeepMind and Primordial Soup. The team restored old photographs with generative models, then used performance-capture models to transfer the couple's present-day micro-mannerisms—including head movement, speech hesitation and facial expression—onto their younger likenesses. Ethelle actively corrected visual details throughout the process.
Why it matters
This is an unusually specific, filmmaker-led use of AI: source photography, human performance and directed iteration rather than a text prompt standing in for production. The workflow keeps the subjects' behavior as the creative source while using generative systems to translate it across age and incomplete visual records. It also makes authorship legible: the director, subjects and technical team shape the result together, and the technology is applied to a narrowly defined storytelling problem.
Antique Robots take
This is where the technology gets interesting. The performance remains the source; AI becomes another transformation layer. Garbus and the team are not asking a model to invent humanity—they are using it to preserve and translate human detail that already exists. That distinction matters.