Context
What does an instant memory generator look like?

Emerging research in synthetic memory systems suggests that externalizing recollection through media can reshape how individuals process, retain, and revisit experiences. However, most AI image generation tools rely on manual text input and remain confined to screen-based outputs, creating a gap between ephemeral recall and tangible artifacts.

This project builds on prior research into synthetic memory systems, exploring how memory can be reconstructed, shared, and externalized through computational media.

---

Design Challenge


How might we design an interaction system that translates spoken memory into a tangible artifact in real time, preserving the immediacy and ambiguity of recall while creating a persistent physical output?

---

Approach (UX + System Design)


Designed and prototyped a voice-to-object interaction pipeline that transforms spoken input into a physical print within seconds.

• Developed an end-to-end interaction flow:
  • Voice input → transcription → image generation → physical output
• Integrated:
  • OpenAI Whisper (speech-to-text)
  • DALL·E 3 (image synthesis tuned for Polaroid-like degradation)
  • Instax Mini Link printer (Bluetooth protocol reverse-engineered via packet analysis)
• Built backend infrastructure (Node.js, Express) for real-time processing and secure API handling
• Iterated on system architecture through prototyping, including an earlier AR-based version that was discarded after evaluating its inability to produce persistent outputs

---

• Explored how users:
  • Translate ambiguous memory into language
  • React to AI-generated visual interpretations of personal recall
  • Engage differently with ephemeral (screen-based) vs persistent (physical) outputs
• Investigated the role of latency, fidelity, and materiality in shaping perceived authenticity of generated artifacts
• Used iterative prototyping as a research method to evaluate how interaction modality impacts meaning-making

---

Outcome


A working system that generates physical, Polaroid-style artifacts from spoken input in real time, reframing AI image generation as a material interaction rather than a purely digital one.

The project demonstrates how:

• Physical outputs can extend AI systems into lived environments
• Interaction design can shape how users interpret and assign meaning to generated media
• Voice interfaces can enable more intuitive, less mediated forms of creative input

---

Impact
Positions generative AI not just as a tool for creation, but as a system for externalizing internal states into tangible artifacts, opening new directions for:

• human-AI interaction
• assistive memory generation for cognitive support