Text to music – Now it is possible!

This article is all about AI music generation. Of course! AI in music is a hot topic and there are many sides: We've seen record labels asking streaming platforms to remove AI-generated music, we've seen artists taking advantage of AI technology, and other artists not so happy having been imitated by AI. This doesn't stop and no company wants to be left behind, so, as if that wasn't enough, recently it's Google who's making waves by announcing the launch of its new artificial intelligence tool to generate music from text: MusicLM.

That is, Universal complains about the music generation by AI and Google decides to make an AI precisely to compose. Yes, although it is not a threat at the moment for Universal, since this AI is not designed to imitate the voice of famous singers, which is the main problem for record companies. But let's go to the most interesting and impressive:

How will MusicLM work?

It could be said that this AI is going to be like the ChatGPT of sound and music. Basically, the MusicLM will generate music from the text. The user will write the type of music they want to achieve and this technology, from a data set of almost 300,000 hours of music, will generate songs or fragments according to the description.

One of the examples, to give a more practical idea of how easy it will be to use it and create music, is: “A slow-tempo reggae song led by bass and percussion. Steady electric guitar. High-pitched bongos with resonant timbres. Relaxed voices that give a feeling of carelessness, very expressive”.

What can MusicLM do?

It is truly shocking to see the rate of advancement of these technologies. Google has posted several examples of what this AI will be capable of and how it will work for the user. W

In addition, as if that weren't enough to generate music using AI from a text description, this technology can turn existing songs into whatever you ask of it: whistled, hummed by a choir, played by a string quartet, sung by a soprano.

It also has a “Story Mode”. In this mode, the song is generated from a sequence of short premises. You can give free rein to your imagination, for example:

(0:00 – 0:15) video game song

(0:16 – 0:30) song to meditate next to a river

(0:30 – 0:45) fire

(0:45 – 0:60) fireworks

And the result, is perfectly synchronized with the indications.

0 stars, based on 0 reviews

Until 04/09/2023 00:00:00