The House of the Einsteins

Overview

An AI-generated historical episode exploring Albert Einstein’s private life in Neusatz between 1905 and 1913, during the years of his marriage to Mileva Marić. Rather than focusing on Einstein’s scientific achievements, the piece reconstructs the quieter part of his life: family visits, streets, rooms, children, routine, and the city that knew him before history turned him into an icon.

I developed the episode end-to-end: historical research, narrative structure, script, visual direction, shot design, period reconstruction, voice, sound, video generation, editing, and final delivery.

The production combined Veo 3, Sora, Runway, Midjourney, ChatGPT, Photoshop, Premiere Pro, and After Effects, using different tools according to the strengths of individual shots and stages of production. Veo was used for atmospheric cinematic movement and environmental realism, while Sora and Runway were tested across character moments, historical streets, interiors, transitional footage, and different approaches to camera behavior and continuity. Midjourney was used primarily for out-painting and extending source imagery beyond its original frame, allowing archival and generated images to be reformatted into cinematic compositions without losing their visual character.

The central challenge was reconstructing an intimate period of Einstein’s life for which almost no moving-image material exists. The historical record tells us where he stayed and what happened, but very little about what those moments looked or felt like. I treated that absence as the creative problem.

Each sequence was developed from historical references to architecture, clothing, street layouts, interiors, bicycles, family photographs, churches, and the visual character of early twentieth-century Neusatz. The goal was not to invent historical events, but to construct a believable visual environment around documented ones.

The visual language deliberately avoids the heroic image usually associated with Einstein. He walks through streets, rides a bicycle, sits with family, moves through rooms, and occasionally disappears from the frame entirely. Children, domestic objects, architecture, bells, windows, and empty streets carry as much narrative weight as the central characters. The camera remains slow and observational, creating something closer to recovered memory than conventional biography.

Every generated shot was evaluated for historical plausibility, character consistency, composition, performance, camera movement, lighting, continuity, emotional tone, and its function within the edit. Generative models created possibilities, but filmmaking decisions determined what survived.

The project represents the kind of AI filmmaking I am most interested in: using generative tools not simply to recreate production techniques that were previously expensive or inaccessible, but to visualize parts of history that were never photographed or filmed in the first place, and turn those missing images into a coherent cinematic memory.