Listen: We Tested Google’s AI Text-to-Music Generator—and It’s Really Good

[ad_1]

Google rewrote the symphony of creativity with MusicLM, its groundbreaking AI text-to-music model.

According to a academic paper published in January, the tech giant said it had “no immediate plans” to share the experimental AI tool, which allows users to type in descriptive prompts to generate unique song. But they reversed course last week, when they announced in a blog post that MusicLM’s waitlist was open.

I’ve tried several AI music programs and haven’t found one with real promise, so I’m suspicious. But after just a few minutes of playing around with MusicLM, any scintilla of doubt disappeared. This is the most intuitive and dynamic AI music generator today.

After opening MusicLM, you meet a modal with instructions on how to create the most effective prompts, such as mentioning the “vibe, mood or emotion” you want to evoke. However, the most surprising is this line: “Some queries that refer to specific artists or include vocals cannot be processed.”

That seems to be a veiled response to the recent increase in scam AI songsthe most explosive of which used AI deepfake tech to mimic the voices of Drake and Frank Ocean, and other major contemporary artists.

Many consider copyright issues to be the cornerstone of the connection between AI and music, but there is currently no silver bullet on the issue—and the countdown is just beginning. The cases piled up by the day. However, Google seems to be aware of the implications.

Google’s MusicLM, an experimental text-to-music model that allows users to type descriptive prompts to create unique songs.

Google

Clicking from the pop-up will take you to MusicLM’s user-friendly dashboard, where you can enter a prompt and have it fed by AI.

The tool then quickly releases two downloadable 19-second audio clips, which are grainy and muddled—but that’s to be expected. If Google integrates AI-powered mixing and mastering services, it will be over.

I started with a simple prompt, “Deep, grimy dubstep at 140 beats per minute.” The program, which Google says was trained on a dataset of roughly 280,000 hours of music, did not disappoint.

In terms of the relationship between a prompt and the resulting audio, MusicLM is most effective when writing clearly with descriptive words. William Faulkner told us killing our loved onesbut he is not a musician.

Here is the surprising result of my prompting, “Ethereal, lush lo-fi with a melody evoking a warm summer day sipping rosé.” You could say I’m ready for summer.

I got another strangely intricate, phantasmic song after the prompt, “Soft, fluttery house music I could hear in a happy dream.”

You can sign up for the MusicLM waitlist HERE.

Editor’s Note: The music included in this article is uploaded to SoundCloud privately and for exploration purposes only. The songs will not be released on streaming platforms.

[ad_2]

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *