"When you have a lot of things, but nothing to wear." Ukrainians have created an AI application that selects looks from photos of your wardrobe: how smart.outfit.ai works

How the idea came about

The idea for smart.outfit.ai arose from everyday experience, not as a technological experiment. «In communicating with different people, we noticed a typical problem: many women are not satisfied with the price-quality ratio of stylists’ services, and independent selection of clothes often does not give the desired result — neither in terms of variability nor in terms of convenience,» says Dmytro Balyuk, founder of smart.outfit.ai.

That’s how the idea to create came about «Pocket Wardrobe Manager» is an application in which the user can store items from their closet and experiment with new clothing combinations using AI.

The founder and chief engineer of smart.outfit.ai is Dmytro Balyuk, who has over 10 years of experience in software development. The core team also includes Kateryna Kravtsova, a business analyst and tester with about eight years of experience. Additional tasks are performed by part-time specialists and contractors — in particular, in front-end development, design, and marketing. The team physically works in Kyiv.

The main problem that the product solves is well known to many: when there are a lot of things in the closet, but nothing to wear.

«People often use the same combinations. Our application helps to see new options and combinations that might not have occurred to you before,» adds Kateryna Kravtsova.

How the technology works

The development of the MVP, says Baluk, took about a year of part-time work. According to his team, at least $30,000 of their own funds have already been invested in creating the product. «A significant part of this amount is actually the team’s time invested: development, testing, and business analysis,» Baluk explains.

Smart.outfit.ai is based on a combination of image recognition models and recommendation systems. The system analyzes photos of clothing and automatically determines key characteristics of items — type, color, and other parameters that are used to create outfits.

«Different tasks arise at different stages — from uploading a regular photo of clothing to generating outfits. That’s why we use a comprehensive approach using several models,» explains Dmytro Balyuk.

The product combines proprietary AI solutions and open-source models integrated into the internal system. The team does not use third-party AI APIs.

At the beginning of development, the team created its own clothing recognition model trained on open datasets. However, later on, data quality issues arose.

«Most open fashion datasets are quite outdated and have a limited variety of clothing types. Because of this, we had to test different open models or return to our own solutions and refine them,» says Balyuk.

Data for AI training

The team uses open fashion datasets to train the models. The total amount of data, according to the team, is hundreds of gigabytes of clothing images.

«It’s difficult to calculate the exact number of images, but the total amount of data is hundreds of gigabytes,» he explains. Baluk. At the same time, the team did not manually mark up the data.

As the developers note, the volume and quality of the datasets directly affect the capabilities of the system. For example, the future feature of realistic display of fabrics and clothing fit will depend on what material data will be available to the system.

Another design decision was that the team refused to use images of people. «We didn’t want to be tied to gender or body type, so there are no photos of models or other images of people in the app,» Kravtsova explains.

Business model

The app operates on a subscription model. Users have access to basic functionality for free, while the Pro version costs around $3 per month (depending on region).

«The services of a stylist often cost from $200 for a limited number of outfits. In this sense, the application pays for itself almost immediately after signing up,» says Kravtsova.

Interestingly, the team does not plan to attract venture capital investments yet.

Users and metrics

The app currently has approximately 200 users. The largest number of downloads is in Ukraine. The average activity rate is 5–10 daily active users (DAU).

«Interest in the product strongly depends on the season, so the indicators can fluctuate significantly,» explains Kravtsova.

The main channel for user acquisition is currently App Store search, as well as TikTok and Instagram. The team calls user retention a key metric.

What’s next?

The niche of AI solutions for the fashion industry is already quite saturated. However, smart.outfit.ai believes that they can compete thanks to a simple interface and a large amount of free functionality.

«Our AI helps you not to create dozens of outfits manually, but only to adjust elements. This removes the monotonous part of styling,» says Baluk.

The team is also not afraid of competition from large technology companies. «Users will always find a product that suits them, regardless of whether it is created by a large corporation or a small team,» adds Kravtsova.

In the coming years, the team wants to develop the product as a tool for ordinary users that helps organize their wardrobe and quickly select clothes. Possible areas of development include AR clothing fitting, integration with marketplaces, and new AI styling tools.

«There is no better time to launch a product than now. Postponing the creation because the market is competitive is definitely not an option,» concludes Balyuk. He adds that the team will add the Ukrainian language to the service in the coming days, and they plan to release a version for Android within the quarter.

Lviv startup creating AI stylist iLea has launched a $41,000 Kickstarter campaign. There is less than a day left until the end of the campaign, and so far no one has supported the project.

On the topic

Lviv startup creating AI stylist iLea has launched a $41,000 Kickstarter campaign. There is less than a day left until the end of the campaign, and so far no one has supported the project.

Read the country’s main IT news in our Telegram

On the topic

Read the country’s main IT news in our Telegram