Tools / Tech: Generative AI image synthesis (Stable Diffusion / Midjourney), 180° stereoscopic VR rendering (Unity), Gaussian splatting, Domestic Data Streamers Synthetic Memories pipeline, USC MEML spatial imaging workflow

Overview
What if you could step inside a memory?

A synthetic memory is a digitally reconstructed representation of a personal experience generated from verbal recollection using AI. Unlike biological memory, which is inherently unstable and fragmented, synthetic memories are computationally produced through iterative prompting and visual synthesis. 

Developed in collaboration with Domestic Data Streamers and USC’s Mobile & Environmental Media Lab, this project transforms spoken recollections into immersive, navigable environments, allowing users to move through AI-generated interpretations of real memories.

Research Question
How might immersive environments:

• transform how users reconstruct and interpret memory
• enable AI-generated imagery to function as a proxy for lived experience
• shift memory from a visual artifact into a spatial, embodied system

Approach (Research + System Design)

• Conducted 30+ semi-structured interviews to study how individuals recall and describe fragmented memories
• Translated interview data into structured prompts for generative image systems
• Iteratively refined outputs with participants to align generated visuals with subjective recall
• Developed a pipeline that converts AI-generated 2D images into 180° stereoscopic VR environments
• 180° VR rendering for spatial immersion
• Gaussian splatting for volumetric, fluid reconstruction of memory spaces

UX / Research Insights

• Users do not recall memories as complete images, but as partial, evolving fragments
• AI-generated visuals are often accepted not as “accurate,” but as emotionally plausible stand-ins
• Immersion significantly increases:
  • perceived realism
  • emotional engagement
  • sense of presence within reconstructed memory

Participants consistently responded more strongly to immersive environments than static outputs, suggesting that spatial context plays a critical role memory interpretation and meaning-making

Outcome
A working system that transforms personal recollections into immersive VR experiences, enabling users to:

• navigate reconstructed memories in first-person
• engage with AI-generated environments as lived spatial experiences
• reinterpret memory as dynamic and co-created with computational systems

Design Insight
This project positions generative AI as a tool for:

• memory reconstruction and augmentation
• human–AI co-creation of subjective experience
• spatial computing interfaces for personal and cultural memory

It demonstrates how memory can shift from: 
static documentation → interactive, embodied environments