Artificial intelligence in perfumery – a brief overview

AI is making big waves these days, more and more influencing different industries, businesses and people every day. Perfumery, the art of making perfumes, certainly becomes one of them. In this blog post we will give an overview of current and possible future applications of this disruptive technology in the interesting world of scents.

Introduction

The youngest and probably most disruptive field of artificial intelligence is generative AI. It is the most creative type of AI out there – it allows us to create texts, images, music, speech and very likely – scents. All this is possible due to some recent groundbreaking advancements in the field of natural language processing, which allow us to interact with these AI systems in a convenient and natural way – by chatting.

With couple of sentences, also known as prompts, we can give these systems instructions to create different artifacts such as images and texts. The applications of such content generation are endless.

Visual design and marketing

Perfume brands can use AI models to create visual materials for their scented products. By describing the idea, purpose and fragrance notes, generative AI tools such as Midjourney, DALL-E 3 or Stable Diffusion can create visually appealing images which can be combined with images of perfumes making the unique visual experience for their customers. In this way generative AI tools can help brands by providing a new and innovative aspect to their overall marketing strategy.

Those tools can be further adapted and tailored to match brand’s own visual identity. By applying such modifications, brands can create personalized generative AI tools.

One nice example of such an approach is the famous Bee Bottle by Guerlain, incorporated into a series of AI-generated visual environments. This leads us to another possible application of generative AI tools in perfumery – bottle design itself. Designers of bottles could use these systems to get inspiration for new bottle designs never seen before.

We also experimented with the DALL-E 3 system to create new perfume bottles. Here's one above. We were amazed by the level of details on the bottle itself. The only input we used was a short textual description of a bottle.

Fragrance composition creation

Another example of using AI in perfumery is the generation of perfume compositions and formulas. Given enough and meaningful data, (generative) AI systems can learn general patterns of perfume creation, combine them with the consumers’ needs and help perfumers to make new and innovative fragrances. In our opinion, this approach can serve only as a supplement in the perfumer’s creative process, but definitely not as a replacement.

The Brazilian perfume house O Boticário already pioneered with the AI in the process of perfume creation, by producing the Egeo On You and Egeo On Me – probably the first fragrances created with the help of AI and a perfumer.

Emotions and colors

One of the recent trends in AI perfumery is predicting the emotions that people feel when smelling a perfume. Fragrances, colors and emotions are very interconnected and some fragrance producers are using AI to predict colors and emotions evoked by a certain fragrance.

The other way around could also be possible - by learning associations between colors, emotions and scents, AI models could suggest new fragrance compositions.

Customer Experience

OpenAI’s ChatGPT is the first publicly available chatbot based on the large language model (LLM) technology. It is a very versatile model able to conduct natural conversations on various topics and solve various medium-difficult tasks.

Perfume brands could utilize LLM models to create engaging experiences for their customers. By using own data about their products and services, as well as consumers’ wishes and preferences, they could make customized assistants which would be able to recommend brand’s perfumes and other beauty products to users.

In this way, each customer would benefit from personal approach, especially on web shops where customers often don’t have the opportunity to talk about different products directly to sales representatives. We believe that LLM-based virtual shopping assistants could provide personalized and engaging user experiences.

On the other side, perfume makers could use large language models to create new perfume names, styled descriptions or marketing slogans. They could also benefit from using such systems for improving the efficiency of internal processes, by letting document summarization and analysis to LLMs or by creating various knowledge-based conversational assistants for their employees.

Our approach

Artificial intelligence is one of the key technologies which Scentalytics utilizes in its products and services. There are two main areas of how we develop our AI-based services for the perfume industry.

AI scent profiling, sentiment analysis and more

Our machine learning models are already capable of predicting different perfume characteristics, such as accords and genders, based only on the input list of perfume notes. They can also predict user sentiments and ratings, allowing the brands to evaluate a potential commercial success of their future perfume products.

Moreover, our AI models can predict a type of a fragrance, as well as the seasons and times of the day when it's most appropriate to wear it. The hidden value of such a multi-purpose model is to use its internal knowledge for our advanced scent profiling service.

This service will allow brands and fragrance producers to visualize each fragrance as a point in the unique 2D virtual space, enabling an innovative and a never-before-seen way of product positioning and comparison with different groups of fragrance products, as well as discovery of new perfume compositions.

Described set of models will be further upgraded in a reversed way – to be able to generate perfume compositions from the input characteristics such as desired type, season suitability and accords.

Custom language models

The second area of Scentalytics’ focus in the AI field are customized large or small language models, tailored to the specific customer needs. Such models could be used for the internal, as well as the external purposes, such as assistants for perfumers or shopping assistants on e-commerce platforms.

Conclusion

Artificial intelligence is here to stay. As the time goes by, it will surely have an increasing influence to the perfume industry as well as others. It already allows a wide variety of possibilities which will further be improved more powerful. If you have any questions, comments or interest in our AI services, feel free to reach out. We will be more than happy to help you.