Music is a constant companion that helps us express our identity, connect with others and form communities, just like fashion, where fashion shows, advertising campaigns and even shopping experiences wouldn’t be the same without its soundtrack.
This relationship between the two is reflected in how we associate certain styles such as leather jackets with rock fans, or how Taylor Swift fans were inspired by one of the singer’s ‘eras’ when choosing what to wear to the concert.
A research group at LCI Barcelona has explored this connection to develop an artificial intelligence tool capable of generating playlists based on a person’s look. This system is called Groovify and was presented at the Sónar+D festival.
This AI-powered tool “analyses a simple photograph of a person to extract data such as the clothes they are wearing, their facial expression and their posture. With this information, a playlist is created with songs personalised for that individual”, a spokesperson at LCI explains.
A unique rhythm for every style
The algorithm behind Groovify analyses clothing styles by detecting visual codes, correlating this information with a vast music database from the Sonar Festival, spanning three decades. “A person’s look is not only made up of trousers and an upper garment, but also accessories and complements,” Alessandro Manetti, head of Europe at LCI and recently appointed vice president of LCI in Europe, told FashionUnited.
“Our professor specialised in artificial intelligence proposed to develop an algorithm and from the fashion area we proposed to associate this algorithm to a reading of the different looks, of the different styles of the garments through the detection of visual codes,” adds Manetti.
The application then creates up to five different prompts that generate a unique single for each person, generating a rhythm and sounds that represent the look of the photographed user.
“We wanted to go beyond simply generating suggestions based on clothing style, we sought to create an unreleased track generated by artificial intelligence and pay homage to vinyl with an imaginary cover.”
The result includes not only the suggestion of playlists based on a person’s clothing style, but also the creation of a mobile application that could recommend music tailored to the user’s look, as well as the creation of a completely new song, created by artificial intelligence and based on the database of Sonar artists. This unreleased music would also be accompanied by an imaginary cover designed using image-generating applications.
The process of associating style and music is based on an exhaustive research of existing databases, analysing styles, music, covers and visual codes in discographies and photographs of concerts and audiences. Manetti explains: “We wanted to go beyond simply generating suggestions based on clothing style, we sought to create an unreleased track generated by artificial intelligence and pay homage to vinyl with an imaginary cover”.
The response to the project at the Sónar Festival was very positive, both from the public and interested investors. The technology used is a proprietary algorithm, which has increased interest due to its exclusivity and originality. Looking ahead, the project’s developers plan to launch an improved version, 2.0, in October.
Groovify 2.0 will be released in October
This new version will expand the database to include additional sources such as Spotify, YouTube and Beatport, with the aim of making the app more relevant and useful in the market. In addition, the team is preparing to participate in the Web Summit in Lisbon in November, where they will present the beta version of the app and look for partnerships and business opportunities.
The project, led by Alessandro Manetti and coordinated by Mariele Violano, is supported by area managers Pedro Coelho, Estel Vilaseca, Anna Pallerols, David Carretero and Salvatore Elefante. The team is made up of 19 people, including six students working in graphic design, interior design, user service and setup and teardown.
This article was originally published on FashionUnited.ES. Translation and edit from Spanish into English by Veerle Versteeg.
