I ordered an AI-generated perfume to learn how it interprets the world

No Ordinary Scent can create personalised, unique perfumes that turn images into fragrances

By the beginning of this year, I moved to Sweden. And since Zuckerberg is always improving his techniques, one of the first ads that popped out on my Instagram was an exact combination of some of my favourite topics: artificial intelligence, innovation and fashion. This is how I learned about No Ordinary Scent or NOS, a Swedish perfume maker founded in 2019 which uses artificial intelligence to create custom fragrances after the analysis of images sent by customers.

I spoke with one of the founders, Sandra Kinnmark, who said that the idea of creating NOS came after the realisation of how hard it is to find companies that make custom fragrances, especially online. “Scents are very personal to us and we wanted to build an online experience where we could create personal scents based on emotions, feelings and visions”, she says.

Both founders Sandra and Amelie Saltin Thor already had experience working in tech startups in Sweden and abroad. This was certainly very useful when they came with the idea of using AI to analyse images that would lead to customised fragrances. But in spite of the feeling of complexity that such technicalities spike, purchasing the product is very simple.

You can name your own perfume and pick up to three images that will give the “tone” of your fragrance. It is recommended that you choose more than one single image, so that the perfume ends up with a more complex scent, but in the end, it’s your creation. Technically, you can pick any image, but it’s encouraged to choose images that bring memories, so that the perfume has other layers of meaning.

In my case, I decided to name my perfume “Sehnsucht”, which means something like “to miss something or someone” in German, and it’s the name of a song by Einstürzende Neubauten that I really like. So I chose photos of people that I miss, on occasions that seemed trivial at the time, but now I miss even these. The analysis is performed live, so you can have an idea of the different notes like the base and the accord.

Each perfume is crafted with up to twenty ingredients. In the case of the base, which carries the deepest and most persistent scents, there are options featuring wood, earth, vanilla and coffee. In the case of the accords, which are a combination of top and middle tones, you will find options such as floral and fruity fragrances, spices and aldehydes.

In order to make the association between these ingredients and the images sent by customers, the algorithm created by NOS analyses the colours, seasons, lightness, and whether there is the presence of nature elements, people, pets and more. While each analysis leads to a specific result, Sandra explains that their AI is always learning from user feedback: “We might have started out with ‘lightness’ equals fresh scents, but the algorithm is way more specific now.”

Sandra believes that the beauty of the technology resides in the fact that the AI is constantly learning and updating itself. “Everything is being co-created by the people who are actually using the scents rather than marketing strategists at big corporations.”

In my case, it was interesting to see how the images that I submitted resulted in a perfume with a base of amber, tonka grain, and roasted almonds. The middle tones are composed of peaches, salty trunks and metallic musk, while the top tones carry pink pepper. I felt that the perfume was a little bit sweeter than what I am used to (I prefer fresh scents), but its woody base balances the scent in a way that makes me think of the word “bittersweet” — a term that I would say that fits very well with Sehnsucht.

Sandra told me that the NOS team never has access to the images sent by customers — these are all stored in an encrypted platform. Through that, the algorithm can use aggregated data obtained from the images to improve the results and present more options for the users. In other words, similarly to other cases of data analysis, the more data the algorithm is fed with, the better it gets.

In the meantime, what their AI is doing is something that was once reserved to semioticists. Both in the fields of communications and arts, there are researchers that do precisely this kind of image analysis by considering the composition, colors, elements, technique and other features that might enable the interpretation of a work.

Curiously, my master’s degree was in semiotics and I made a similar analysis of Gottfried Helnwein’s works. While there are some consensus (for example, that the black colour carries a more grim feeling), these are conventions that can very easily change through time and from one culture to another — as much as pink clothes were once associated to boys or how Hindi brides wear red instead of the white gown that is commonly used in the west.

With that considered, NOS creates fragrances that are agender. While some brands still sell female and male perfumes, it is widely known how the fashion industry is currently leaning towards more neutral creations and the subversion of gendered associations. However, the algorithm attempts to categorise what colours, elements, and shapes might mean and how they could be translated or represented in scents.

This is an approach already seen in the work of the neuroscientist Paul Bach-y-Rita, who created “sensory substitution” devices for people with disabilities to acquire the affected senses from other ways (for example, sound through vibration). It is in a similar way that the artist Neil Harbinsson’s implant works, by converting colours into sounds.

While this conversion is made in terms of frequency of sound and light, the algorithm used by NOS is more subjective. Still, it makes accessible to people a common technique used in marketing and in the consumer goods industry — that is, the creation of fragrances that trigger specific memories and, consequently, sales. It is the idea behind the specific smell that McDonald’s restaurants have, and how it is fabricated in a way to make it a standard in all franchises distributed around the world. Same goes for other shops that use air fresheners with a specific fragrance.

With the customisation of fragrances offered by NOS (who also works for brands that are searching for this same kind of product), anyone can create their own scent and “trigger” for a memory. Taking this to the extremes of science fiction, in “Antiviral”, Cronenberg proposed the confection of fragrances or even meat after the DNA of celebrities. However, in the real world, what we are actually aiming for is to find ways on how these emerging technologies can help us reconnect with ourselves instead.

Disclaimer: This is the translation of an article originally published at Tilt UOL.

Did you like the post? Would you like to Buy me a coffee? :)



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Lidia Zuin

Lidia Zuin


Brazilian journalist, MA in Semiotics and PhD in Visual Arts. Researcher and essayist. Science fiction writer.