Advances in artificial intelligence in recent months have focused (with great impact) on generative models, capable of producing diverse content from simple instructions (prompts) in natural language.
A prompt is the instruction in text that, using natural language, generates a content
This makes it possible to describe a landscape, object or character and have the result be an illustration (with Dall.e or Stable Diffusion) or to request the summary, translation or composition of a text (with ChatGPT or Google Bard). Fortunately for music lovers, there is also the possibility of using the same premises. Thus, through a description made in natural language, you can obtain as a result songs “composed” by an artificial intelligence according to the instructions provided by the user.
Best AIs to generate music
Some examples of artificial intelligences that are capable of generating music following descriptions of the type of songs users wish to originate are:
-MuseNet: It is a model developed by OpenAI (creator of the popular ChatGPT) that can generate original music in various styles and genres from descriptive text. The user can enter a description or a text fragment and MuseNet will create a piece of music based on the interpretation of that text. One of its peculiarities is its ability to combine different genres in the same composition. MuseNet can generate music in styles such as jazz or pop, and can also perfectly combine different instruments. It is free to use.
-Jukedeck: Acquired by TikTok, this is an artificial intelligence-based platform that uses algorithms to compose personalized music. Simply enter a text or description and Jukedeck will generate a musical composition according to the content of the text. The resulting music is royalty-free, in addition to being unique songs. This is one of the reasons why TikTok acquired this platform, in order to be able to offer users of the video platform unique background music without having to pay royalties for it.
-Amper Music: This platform offers two modes of song generation, Simple and Pro. In this case, rather than allowing the user to type a description in natural language of the expected result, it offers the possibility of selecting the characteristics it will have by means of different buttons. Thus, from the first options (classic rock, modern f0lk, 90’s pop, movie soundtrack…), after which you have to select other features such as duration and “mood”. There are also settings that allow you to set tempo or key of the piece.
-Magenta: It is an open source Google research project that explores the intersection between artificial intelligence and creativity. Magenta includes AI models capable of generating music from prompts in different formats, including text, images and audio signals.
-MusicGen: Coming from Meta (parent of Facebook, Instagram, WhatsApp…), this artificial intelligence has been trained with 20,000 hours of licensed music and 10,000 high quality audio tracks, allowing unlimited possibilities of combinations and remixes to obtain the tunes the user is looking for.