Google taught AI to create music from a text description – video
News » Science & Technology
Google is behind the development, creating the MusicLM neural network service. The task of the service is to create musical works. You can listen to samples of the work on the demo site.
MusicLM creates music based on text requests. The user can set the genre, mood, tempo and other desired nuances.
“We present MusicLM – a model that generates high-quality music from text descriptions, like “soothing violin melody backed by a distorted guitar riff”, – say the developers on the site.
MusicLM considers the process of conditional music generation as a hierarchical sequence modeling task and generates music at a frequency of 24 kilohertz, which remains unchanged for several minutes. The developers are confident that MusicLM is superior to previous systems both in terms of sound quality and compliance with the text description.
The neural network can also combine several different works into one common one. To do this, in the instructions you need to specify descriptions for each of them. In addition, you can choose the level of proficiency in – for example, in order for a composition to be supposedly played by a novice musician or, conversely, a virtuoso.
Instructions can also be presented in the form of sounds. For example, the user hums or whistles a melody, and the neural network adapts it to a full-fledged track, taking into account the style.
Google представили новый ИИ, который генерирует музыку по текстовому запросу
MusicLM – так называется сервис, способен генерировать песни по текстовому описанию, по вашим образцам музыки или по загруженному изображению pic.twitter.com/2tX8lH09Ec
— STARLINK (@lgj170) January 28, 2023
Follow us on Telegram