Just hum a melody to turn it into notes. VocalizeR, an application developed by graduates of Applied Computer Science, transforms the user’s voice into an editable MIDI file. The project was awarded in this year’s CreatiWITy competition in the team project category.

The application was created by Applied Computer Science graduates: Mateusz Dejewski, Michał Kałuski, Tymoteusz Kołaczek and Michał Maksanty.

The inspiration came from a topic proposed by the Wrocław-based startup Futurient. Although the team explored other concepts, this idea proved the most compelling – largely due to the musical interests of the authors and the lack of strong solutions in this space.

As the winners admit, VocalizeR initially seemed like a straightforward project. In practice, it turned out to be a complex technical challenge.

Na na na…

The team focused on simplicity. VocalizeR is designed primarily for humming. As the creators highlight, the application works both for trained vocalists and for those with no musical background.

– We went with the simplest form of humming – the classic “na na na”. It also works with singing or any kind of monophonic sound – explains Mateusz Dejewski.

The intended workflow is simple: hum a melody, record it, send it to the server, analyse it and receive musical notation. What matters most, however, is not the audio itself, but its transformation.

– Raw audio is just a sound wave – a one-to-one recording that’s hard to edit. VocalizeR converts it into a symbolic representation in the form of notes, which is far more suitable for music editing – explains Michał Kałuski.

This symbolic representation is key for further creative work. It allows users to edit melodies, shift notes, correct mistakes and experiment with sound.

The output includes MIDI files that can be imported into external music software or played on keyboards, as well as a PDF score ready for live performance.

A tool that doesn’t require perfection

– Many people can hum or sing something, but without any knowledge of music theory. Almost no one can write it down. Even if we make mistakes, the app allows us to adjust and refine the sounds – says Mateusz Dejewski.

The recording can also be translated into different instruments, allowing users to hear how the melody would sound when actually performed. VocalizeR does not require perfect execution – what matters is the idea, which can be refined later.

One of the project’s key features is the ability to choose between transcription models. VocalizeR offers eight models, including two developed by the team. These combine existing approaches to maximise output quality.

– Each model has its own strengths. They perform best with specific types of sounds – adds Mateusz Dejewski.

For beginners and professionals

The application is designed both for beginner music creators, unfamiliar with notation or professional tools, and for experienced musicians, for whom manual transcription by ear can be time-consuming.

– I often find it hard to judge whether a note should go higher or lower just by ear. That’s why we added features like pitch contour – it shows where the melody rises and falls, even when it’s not obvious – explains Michał Kałuski.

VocalizeR can also be useful in technical environments, such as game or film production, as a fast tool for prototyping melodies.

The application is available online and designed as a free browser-based tool. It shows how technology can directly support creativity.

– Want to quickly capture your music? Just visit www.vocalizermusic.com and hum it into our app – concludes Mateusz Dejewski.