A team of researchers from the Tokyo Institute of Science unveiled a proprietary AI model capable of capturing, duplicating, and creating new odors. The technology could impact several industries, from fragrances to healthcare and beyond. Here’s how automated AI scent design will unlock a range of new fragrances in the coming years.
Why the Sense of Smell Matters More Than You Think
If you were to ask someone to pick which of the five senses they could live without. Most people would likely choose the sense of smell. While this may seem obvious, evolution differs in that opinion. It prioritizes the sense of smell over sight and sound.
Unlike your other senses, olfactory sensory response skips the thalamic relay. Instead, they are sent directly to your piriform cortex, amygdala, and hippocampus, the parts of your brain responsible for memories. That’s why the smell of mom’s cooking instantly brings back poignant memories.
According to recent research, your sense of smell adds to your ability to remember other characteristics of a moment. This study revealed that by stacking odors with scenarios, patients could remember visual cues more accurately. The study determined that olfactory sensory input, in addition to other senses, improves memory functionality.
Fragrance Market Growth Driven by Demand for Unique Scents
All of this data sheds some light on the surge in the fragrance market. The fragrance sector is more than just perfumes and colognes. It includes a massive array of products like air fresheners, cleaning products, and cosmetics.
Even car manufacturers have begun to add fragrances to their cabins. When you combine the size of the sector with the rise in luxury products, like $50 scented candles, it’s easy to understand why analysts predict the market will grow from $30B to $50B by 2030.
Challenges Facing the Traditional Fragrance Industry
Fragrance designers are now under immense pressure to develop and refine odors that continually differ from their last attempts. After centuries of fragrance design, it’s hard to consistently develop new and pleasing odors. As such, the process has become very labor-intensive and time-consuming.
Why Perfumers Still Rely on Manual, Time-Consuming Methods
Additionally, the process is expensive as it requires several specialized perfumers who undertake trial-and-error attempts during the process. This approach lacks sustainability and uniformity as each perfumer has a varying level of skill and expertise. Consequently, the results vary per person.
High Costs and Raw Material Challenges in Fragrance Production
In addition to the need to hire experienced professionals, fragrance manufacturers have to get all of their materials safety tested. This step, coupled with fluctuations in the price and availability of raw materials used to create the fragrances, can result in rising costs to manufacturers and consumers. According to some reports, you can expect to spend anywhere from $100 to $2000 to acquire high-end fragrances nowadays.
How AI is Changing Fragrance Design Forever
Recently, fragrance engineers have turned towards AI to help streamline the process. The creation of generative diffusion networks represents a major step for AI and the fragrance industry. These systems use AI to recreate odors based on predetermined model data. Notably, these models are commonly used in text and image generation systems. Now, they can create scent profiles.
Limitations of Current AI Scent Diffusion Models
Unsurprisingly, the new AI systems have helped engineers to streamline the process and better understand scent profiles. However, today’s generative diffusion networks suffer from a variety of issues that continue to limit their capabilities.
For one, the majority of these models rely on proprietary data such as preloaded scent profiles. These models are limited in their depth and also don’t allow for maximum performance. Additionally, the current process still requires a lot of human interaction to complete.
Tokyo Tech’s AI Scent Design Breakthrough Explained
Recognizing the limitations in today’s diffusion networks, a team of scientists from Science Tokyo introduced a fully automated AI scent model. The study1 “Generative Diffusion Network for Creating Scents” delves into the purpose-built AI model and how it creates new fragrances by reversing a noise process based on mass spectrometry data.
Source – Aleixandre et al – IEEE Access
How OGDiffusion AI Creates and Recreates New Fragrances
The OGDiffusion model was built from the ground up to streamline fragrance creation. The model can duplicate an odor or create new scents based on a description. The new aromas are mixed using a blend of essential oils distributed from an automated system, eliminating the need for human intervention.
Using Mass Spectrometry to Power AI Fragrance Models
At the core of the OGDiffusion approach is the use of spectrometry data profiles. Specifically, the engineers performed a spectrometry scan on 166 essential oils. Their odors were divided into 9 odor descriptors, ranging from floral to woody or citrus.
PCA Analysis Helps AI Understand and Replicate Scents
The mass spectrometry scans were set up utilizing a Pine Scotch essential oil as a reference. When scanning new aromas, the system will create a new profile that includes modified odor descriptors, enabling the system to create a target representation of the smell based on its overlapping odor characteristics.
AI-Generated Odor Descriptors Tailor the Final Fragrance
The model mixes essential oils to get the desired results. The corresponding mass spectrum data tells the mixing protocol if it aligns with the desired descriptors. Notably, the AI automatically calculates the precise mixture of essential oils needed to recreate the scent.
Testing AI-Created Fragrances on Human Participants
The engineers put their system through several testing phases to ensure its accuracy and capabilities. The testing procedure involved 14 participants. It included double-blind human sensory and scent validation tests.
In these steps, the subjects were required to distinguish between aromas. Interestingly, the team decided to use one odor originally, before adding a second odor descriptor and asking the patients to differentiate between the two.
Participants Accurately Identify AI-Generated Scents
The AI scent design tests results indicate that the model automatically generates accurate and identifiable scent profiles. The system creates these custom fragrance recipes utilizing AI descriptors, allowing it to produce diverse and original profiles that participants consistently matched with the appropriate descriptor.
Benefits of AI-Driven Fragrance Production
There is a long list of benefits that this research brings to the market. For one, the entire system is fully automated. As such, anyone will be able to operate one in the future, meaning that you could be able to customize aromas as needed.
Faster Scent Creation Through Automated AI Systems
When compared to traditional fragrance development, this system is much faster. By automating the generation of mass spectra corresponding to desired odor profiles, fragrance manufacturers can cut out time-consuming steps and personnel. As such, the system is far more efficient.
Essential Oils Offer a Safer, Natural Alternative
Another major plus for the system is that it utilizes essential oils as the main ingredients. Essential oils are safe to apply to the body and the environment. As such, it allows fragrance manufacturers to avoid stringent synthetic ingredients and allergen regulations.
AI Fragrance Design Offers Efficient Scalability
Scalability is another benefit that the AI protocol has over older production methods. The automated system can be scaled up to meet demands without the need to create massive production facilities. Impressively, the system can also fit into a compact space or linked with others to meet industrial needs.
Real-World Applications for AI in Fragrance and Scent Tech
There are many applications for the AI scent design system. A multitude of industries, from hospitality to healthcare, rely on scents to make your stay more pleasant or their product more desirable. Here is a list of some primary applications of this technology.
AI in Perfumery: A New Era of Custom Scents
Creating new and unique aromas for perfumes and cologne is the obvious use of this technology. A combination of improved social media marketing and celebrity endorsements has helped propel the market forward. This billion-dollar industry continues to see rising prices, as scents become a crucial part of grooming for younger generations.
Smell Engineering in Food: AI Replicates Natural Aromas
When you think of smell, you probably think of a delicious meal. The odor emanating from the restaurant. The food industry is built on fragrances. As such, this technology could allow engineers to create unique dishes that could replicate the aroma of ingredients, without the need to add them.
Customized Scents in Home and Cleaning Products
From scented candles to cleaning goods, home products rely on scents to entice their clientele and leave your location smelling fresh. Companies like Pine-Sol have built their brand on their aroma and product capabilities. Future options could let you customize the odor to fit your needs.
AI Scent Detection Could Help Identify Counterfeit Goods
An odor detection system could be a powerful tool to determine if a product is a counterfeit or original. Certain products have a particular scent profile based on their materials and composition. The AI system introduced in this study could utilize a pre-scanned profile to ensure that the products shipped are the originals.
AI Scent Design Researchers
The AI scent design study was led by engineers from the Institute of Science, Tokyo. Specifically, Professor Takamichi Nakamoto worked with other researchers to ensure the study’s success. The project received additional support from the Laboratory for Future Interdisciplinary Research of Science and Technology (FIRST) and the Institute of Innovative Research (IIR). Notably, this study is not the first time that this team has worked with diffusion networks. The team’s previous work was crucial in creating the new AI algorithm class.
The Future of AI-Driven Smell: From VR to Healthcare
There are a lot of possibilities for AI scent design technology. For example, recent breakthroughs in VR have opened the door for olfactory sensor inputs. These systems allow designers to add smell to their virtual worlds. Notably, this technology is already being tested as a way to help fight dementia and cognitive decline in elderly patients.
Investing in AI and Fragrance: A Growing Market Opportunity
The fragrance industry is a billion-dollar market that has seen considerable growth since the pandemic. Today, there’s a mix of high-end and emerging fragrance firms pushing the boundaries of tech to the next level. While not involved in the study, here is an example of one company that has the leverage and network to utilize its findings with great effect.
The International Flavors & Fragrances (IFF +0.26%) company enteredd the market in 1889 as Polak & Schwarz (P&S). The company was founded by Josef Polak and Leopold Schwarz after the two decided it was time to combine their knowledge into a manufacturing powerhouse.
In 1958, Polak & Schwarz (P&S) merged with van Ameringen-Haeble. This move created the International Flavors & Fragrances company as it’s known today. Since then, the company has seen considerable growth. Today, it is one of the largest fragrance manufacturers in the world.
International Flavors & Fragrances Inc. (IFF +0.26%)
Those seeking a proven fragrance market leader should do more research into IFF. The stock has strong support from the institutional and commercial sectors. This support is evident by the fact that DuPont shareholders own 55.4% of the combined company.
IFF Stock News and Latest Developments
AI Scent Design Set to Revolutionize Fragrance Creation
Automated AI scent design is just another example of how artificial intelligence continues to expand mankind’s capabilities. This system will eliminate much of the waste in the fragrance industry and enable a new level of customization, reduced production costs, and accessibility. For these reasons and many more, you have to commend this team on their efforts.
Learn about other cool AI projects here.
Studies Referenced:
1. Aleixandre, M., Romain, L., Otsuka, H., Inamura, Y., & Nakamoto, T. (2024). Generative diffusion network for creating scents. IEEE Access. https://doi.org/ACCESS.2025.3555273