In what feels like a smorgasbord of non-stop AI news having an application in almost every aspect of our lives, Meta sends ripples through the music community with the launch of their new text-to-melody AI generator ‘MusicGen.’

Yep, you read that right. Text-to-Melody. MusicGen’s powerful generating capabilities enable users to turn basic text instructions into intriguing musical pieces, and here’s how it works.

First of all, users will find a ‘Describe your music’ text box where they need to enter a prompt describing what they wish the resulting sample would sound like. You could keep it simple like “Deep bongo beats“ or you could go with a more detailed approach like “A peaceful, meditative zen track infused with Jazz elements and a smooth saxophone solo”. Of course, the more precise you are the better results generated.

Then users specify the duration of the sample and press submit and boom, just like that a new and original sound-clip is generated in mere seconds. Currently the sample duration limit is 30 seconds, but we’re sure that’ll change in this era of evergoing AI advancements.

Taking it up a notch, MusicGen offers the option of adding an audio reference to your prompt. The text sets the basic style, which then matches the melody in the audio file. Imagine having the theme song to “El-Basha Telmiz” but in a rock-infused style, or the iconic voice of Adele but with the raging beats of a Tabla and Doff in the background. Jury might be out on whether or not those outputs are pleasant or not, but that’s just the beauty of it. If you don’t like what you hear, you can just adjust a couple of words and create a whole new sound on the spot. The possibilities are endless.

MusicGen’s user-friendly interface empowers individuals, allowing anybody, regardless of musical experience or expertise, to produce fascinating music.

This comes as a gamechanger in the video production field, as content creators need not rely on music producers to provide music for their content. Moreover, MusicGen’s generated music is 100% copyrights-free, eliminating the fear of copyright infringement claims from content creators’ heads, leaving them with the freedom and ability to customize and elevate their creative vision.

Meta’s MusicGen isn’t the first AI music generator, there are others like Riffusion and Google’s non open-source MusicLM which was introduced at the beginning of the year.. We won’t be getting into the tech-y nitty gritty details, but here’s a cool video we found comparing Meta's MusicGen with Google’s MusicLM.

With all that said, it is important to emphasize how much of a double-edged sword Music AI Generation is. It facilitates the lives of content creators with zero musical expertise and promotes on-the-spot experimentation,  but what implications does it hold towards music producers? Should they follow the footsteps of visual artists who fear their imminent replacement by artificial intelligence? As always, the future of AI in music is as exciting as it is scary.

Meta has released the code and models as open-source on Github with commercial use permitted, and a demo is available Huggingface which you could try out now by clicking HERE.