This project was a collaboration with Domestic Data Streamers on their Synthetic Memories project, which reconstructs lost or fragmented memories through Generative AI. While their existing pipeline produced 2D representations of memories based on interview data, our mission was to make these moments immersive, allowing participants to step inside their memories in 3D space with audio. 

Our research focused on two key goals: 

The process begins with a one-on-one interview, where participants describe a vivid memory in detail. Using AI-generated visual prompts, these recollections are iteratively refined until they align closely with the participant’s perception.

Through experimentation with different ways to spatialize memory, we developed a pipeline that leverages 3D stereoscopic 180° video and spatialized audio to evoke a strong sense of presence and emotional resonance. This immersive format allowed interviewees to not just recall, but relive their memories in a comforting embodied way.


The experience integrates two complementary visualization methods:

To evaluate both emotional impact and technical feasibility, we tested the immersive experience with participants during the installation. Each person received a 2D Polaroid reconstruction of their memory before entering the headset. In nearly all cases, the immersive version, whether rendered as stereoscopic video or through Gaussian splatting, evoked a noticeably stronger emotional response, reinforcing the potential of spatial immersion in deepening memory recall.

Although speed was not the primary focus, we successfully generated 15 immersive memories within just a few hours. Compared to DDS’ original pipeline, which typically produces 3–4 memories per day, this streamlined approach demonstrates how immersion can be delivered efficiently without compromising emotional resonance.