The experience begins by 'uploading' an analogue photograph into the machine.

A brief narrative is generated by a vision-language model (VLM) which can be shaped using a novel dial-based interface. 

Three tactile dials control perspective, time, and mood. Together, they guide the framing of a narrative, shape the immersive memory that unfolds, and influence the character of the scent that will emerge.

The dial inputs are structured and passed to a large language model (LLM) to produce a prompt. This prompt becomes the semantic foundation for a translation from image-to-text-to-scent.

Drawing on a curated knowledge base that links narrative cues to structured scent profiles, the model interprets the emotional and thematic contours of the generated story.

An algorithmic blending process then combines up to four fragrances from a 44-scent library, to create a refined olfactory composition—a scent that synthesises an unfamiliar memory.

Register Your Interest

Register Your Interest

A brief narrative is generated by a vision-language model (VLM) which can be shaped using a novel dial-based interface. 

Three tactile dials control perspective, time, and mood. These guide the framing of a narrative, immerse the user in a memory, and determine the character of the scent that will emerge.

The dial inputs are structured and passed to a large language model (LLM) to produce a prompt. 

Drawing on a curated knowledge base that links narrative cues to structured scent profiles, the model interprets the emotional and thematic contours of the generated story.

An algorithmic blending process then combines up to four fragrances from a 44-scent library, to create a refined olfactory composition—a scent that synthesises an unfamiliar memory.

Read the paper (published in NeurIPS 2025) – The Anemoia Device: A Tangible AI System for the Co-creation of Synthetic Memories through Scent

Register Your Interest