Earlier this year, we introduced the world to Osmo’s most ambitious project: Scent Teleportation, a technology that captures a smell in one part of the world and diffuses it in another. That blog post started a lively discussion. Some of you tech-forward people were curious to know more about how Scent Teleportation works and how far along the project is in its evolutionary cycle. And we’re happy to oblige, glad for you all to accompany us on this epic journey to digitize scent. 

‍

We have already achieved Scent Teleportation in our labs and documented this process in a Scent Teleportation video. Our first successful attempt involved a simple slice of coconut. One could describe what happened that day as us sending the scent of that coconut from one side of the lab to the other. But that description elides a much more complex underlying process of digitizing and reconstituting the molecules that make things smell the way they do.

If you are an olfactory expert, you will likely be familiar with most of the technologies mentioned here. What is novel about our approach is the way we’ve gotten every piece to work together in the weekly experiments we’ve conducted since our first Teleportation, using modern software and artificial intelligence. The process requires no human intervention at this point, except at the input and output level.

We select a scent to teleport and introduce it to a machine called the GCMS (Gas Chromatography-Mass Spectrometry). If it’s a liquid, we inject it directly; if it’s a physical sample, like a plum, we use headspace analysis, which means trapping the scent in the air around the object and absorbing it through a tube. The GCMS identifies the raw data, which can be interpreted as molecules, and uploads it to a cloud. There it becomes a coordinate on our Principle Odor Map — a novel, advanced AI-driven tool that can predict what a particular combination of molecules smells like. This formula is sent to one of our Formulation Robots which treats it as a recipe, and mixes different scents to replicate our sample.

Finally, we compare the original sample and the facsimile. Our goal is to achieve a high degree of similarity, ensuring that recreated scents align with their original odor families. With every trial we get a bit better at capturing every nuance. These nuances are inherent to the world of odors. Some molecules are so subtle that they barely register on sensors but still significantly impact the overall scent. Sulphurs in tropical fruit are notoriously hard to detect and reproduce. Some molecules involved in everyday scents remain unidentified. Our ongoing data collection process is an effort to reduce this number of mystery molecules, and find new ways to recreate them. Every time we solve one of these tough cases, it feels like an important breakthrough.

This process requires a lot of humility, because the challenges along the way can be immense. We are doing something that no one has managed before. On a daily basis, we collect more and more data. We have already accumulated the biggest AI-compatible scent data bank in the world. This data is essential for training our algorithms; refining their noses, as it were. Converting all that data into usable formulas still requires some human guidance — training wheels that we hope can soon come off.

Luckily, we have a lot of help. We use simple algorithms that use existing data to decode scents; Machine Learning Deformulation models to predict formulas from GCMS data; and of course our Principal Odor Map, a sophisticated model that maps odors into a multidimensional space for more accurate recreations. We put these tools to constant stress tests in monthly Scent Teleportation trials that help us refine our processes.

Soon, we will be opening select demos to members of the public. Hopefully some of you will come by to experience scent teleportation first-hand. We will ask you to pick a flower or a fruit from an extensive bouquet, and then let you reflect on its smell, while we analyze a sample of it in our lab. In a single sitting we will recreate your special scent and diffuse it back to you. And then, the moment of truth: your review. How faithful was our reproduction?

In the six months since we first announced scent teleportation, we have made an astonishing amount of headway. Every month, we make Scent Teleportation a little bit quicker, a little more automated, and challenge it to transmit more complex smells. And so, we get a bit closer to that ideal we’re working towards: a world brought closer together in one of the most essential ways, where a beautiful scent that you smell on a hike is as easy to send as a picture or song.

‍