Will AI finally take over the music in videos?

An overview of MusicGen, an AI-powered music generator by Meta.

AI of this, AI of that–Every time we cover a new AI breakthrough in the film and video industry, it seems like another AI innovation is released before we hit the submit button on the previous article. Good thing they’re all written by AI, aren’t they? (Just kidding, I’m a real person…at least for now.)

The concept of AI music generation has always seemed to be one of many logical next steps as artificial intelligence takes over all sorts of tasks in video production. Although there are arguments that AI may never materialize (or at least soon) actually replaces the more nuanced elements of creative artistic expression where cinematography, video editing, and music fall under, and at least that foray seems to be finally here.

Meta has released its new AI-powered music generator, MusicGen, to the public. It enables users to create high-quality, royalty-free music with simple text instruction prompts. Let’s take a look at this new AI music app and explore how it finally takes over the music of the video as we know it, or not.

MusicGen by Meta

Before delving too deeply into MusicGen by Meta, Meta’s model of music generation is open source project And it’s shared by both the AI ​​community and the music community.

Interested users and nerve-wracking investors can monitor its development, contribute to its design, and eventually even create their own variations based on early neural networks. But for those strictly concerned about which industries this will replace, or how it will affect music licensing for videos, those questions are still unanswered so far.

As you can see in the introductory video posted by AI researchers at Meta, MusicGen is designed to be a simple and controllable music generation model. Both text and melody let you create new renditions of your favorite pop classics, or create your own melodies using text conditioning and sampling (you can see it in action below ).

Usage

Without going too far into the complex world of behind-the-scenes machine learning capabilities, MusicGen provides a neural network that can work with basic text descriptions and audio input. In just a few seconds, MusicGen will receive any prompts you want and based on your cues he will create 15 seconds of audio.

MusicGen is also very customizable and accurate, as it can provide prompts based on beats per minute, or even provide your own audio clips for the purpose of creating different versions and loops. are also promised. And when you sign up for a project, hug face spacegives you access to clips up to 2 minutes long.

It’s all exciting and scary, isn’t it?

The future of music generation for video

Still, I have so many questions regarding this new release. First, issues of ethics and legality come to mind. However, Meta said MusicGen is only trained on his 10,000 hours of high-quality certified songs and about 400,000 instrumental tracks from his media library powered by Shutterstock and Pond5. That’s what I mean.

As anyone who has worked on video productions that require various types of copyright-free or copyright-free music can attest, the world of video music licensing is very complex and difficult to navigate. is difficult. There’s still a lot to see about how these AI-powered music loops can really be used in the world of video, or just the world of audio production and editing.

But rest assured. By the standards of this rapidly changing AI world, these new innovations and applications have long been unrealized. AI-powered music generators shouldn’t come as a surprise, but it will be interesting to see how they develop and how video professionals and the industry at large adapt to these new AI technologies. Definitely amazing.

How do you think AI-powered music works for movies and videos? Interested in trying out MusicGen?

Let us know what you think in the comments section below.