AI-powered Search Engine That Finds Songs Through Emotion

Every piece of music that captures the heart encompasses worlds of musical production, rhythm, melody, lyrics, and style. Even the simplest of songs unfolds incredible complexity that activates entire areas of our brain and evokes subtle feelings of joy, sadness, heartbreak, or longing.

MyPart, according to its entrepreneurs, was established to "crack the genome" of songs with the help of the most innovative technological tools available, primarily: artificial intelligence. The company has trained AI models that understand songs in depth – from genre and style, through lyrics and writing style to the rhythm, melodies and chords that make up each tune. "These capabilities do not currently exist as a whole in any search engine or streaming platform, which focus on trends, genres and historical listening data," MyPart executives explain.

MyPart has built an intuitive song search engine that complements the music streaming experience, where you can search with reference songs as well as words that express intuition or emotion – such as "an angry breakup song, but one I can dance to, to get over my ex," or "a pop song like Ariana Grande’s latest release – but that deals with self-empowerment without a lot of curse words."

Behind the startup, which opens new horizons of content for music lovers, is a group of musicians who happen to be engineers, artificial intelligence experts and data scientists, along with senior executives from the music, film, and advertising industries.

Matan Kollenscher, the company's founder and CEO, has been drawn throughout his life to two fields of expertise that sometimes seem contradictory: music and computer science. " Since my teenage years, I was a programmer on one hand – and a songwriter on the other. When I finished my 7-year tenure at an elite intelligence agency, I studied electrical and computer engineering at the Hebrew University. But I couldn’t leave music behind, so I majored in songwriting and composition at the Rimon School of Music and continued to academic musicology studies," he says, adding that his favorite teachers at ‘Rimon’, Dr. Ari Katorza, was one the first to join MyPart. “I thought it would be amazing to teach a computer to deeply understand songs: their mystical combination of words, melody, and musical production. This is in fact what makes music so special – it has multiple dimensions that affect us independently and simultaneously. Until recently, our vision sounded like science fiction – but today we see what the models we developed achieve and simply can't believe our own eyes," shares Katorza.

The Songhunt interface

The company's CTO, Amir Graitzer, an experienced data scientist who is also a professional guitarist, explains: "There was a significant technological challenge here. Most of the AI tools we see on the market focus on one dimension, such as text, images, or sound. But because music is so multifaceted, we had to devote vast research and development resources to each dimension separately and, in addition, build a computational model that would factor them all in, in a way that corresponds with the human ear."

A new standard of song search

A few weeks back, the company launched Songhunt.ai, a song search and playlisting engine. "Imagine having a musical compass that directs you to songs that connect to your feelings and thoughts at any particular moment," says Kollenscher. "Songhunt can take a nostalgic or contemporary song that you love, as well as descriptive words about its music, lyrics and sound and suggest similar songs that are worth getting to know. Based on that, it can create playlists for Spotify, Apple Music and YouTube Music for a relaxing afternoon on the beach or a fun evening with friends."

The world of search has undergone a massive revolution in the past year. "In recent months, artificial intelligence has had an impact on a variety of fields. Most people are exposed to language models such as ChatGPT or image creation like Midjourney – but at the same time, a profound change has been happening in search," explains Kollenscher. "For years Google dominated the field, with a superior technology and interface that have become the standard. Now AI-powered search engines are changing the rules of the game. Microsoft, for example, promptly adopted the chatbot into their search engine, Bing, and Google followed with their own LLM, Bard. MyPart plays in a parallel field of song search, which is relevant to more than a hundred million people a day, who use search engines predominantly to find song titles and artists. Think about it – Google and Bing have tabs for searching text, images, videos, hotels and flights – but there is no tab for searching music, one of the top search categories."

Find the exact song

Along with Kollnescher, MyPart is led by a team of professionals that come from of music and technology. The company's president, for example, is Ronny Vance, former president of Geffen Music and Interscope Music, who signed mega-songwriters like Tupac Shakur and Gwen Stefani. The company's Advisory Board includes executives from the global music industry, as well as entrepreneurs like Tristan Jehan, one of the key figures behind Spotify's recommendation engine, that served as its director of research for years.

Kollenscher explains that the goal of Songhunt is to provide music lovers at large with access to what has thus far only been available to a closed group of music executives and creative professionals: to use AI to find relevant music in a simpler and more intuitive way than ever before: "When a song moves you, without thinking too much, you want to find more songs that will have a similar effect. Songhunt is that 'co-pilot' or personal assistant that based on what you are thinking and feeling now – can find the most appropriate matches for you. If you want to surprise your friends or partner with a playlist tailored to the occasion – the search engine we developed will find just the right songs for you."

Other company products include Songmine for music executives and creative professionals, and SongSource for indie artists and songwriters. "Since the platform’s launch, we have reached leading clients, including the largest music companies in the world, advertising, film and television companies, and tens of thousands of indie artists who are looking for a way to break through," Kollenscher concludes. "These people work hard to refine and convey a message or express a subtle human experience, and create or search for the precise song that will fit an artist, scene or brand. The technology we developed has until now only been available to them – but we are very excited about the opportunity to open it up to the general public as well."

Songhunt is currently available for free as a beta at the link >>

In collaboration with MyPart