Meta has released a new AI-powered music generator platform “MusicGen” for anyone out there to compose music with help of AI.
Likely as Google’s MusicLM, Meta has created an AI-model for music-composing and called it “MusicGen”. MusicGen is competingly cleverer in generating music than MusicLM, the composed music shows. Besides, text, image, and 3D model generation, AI is now tuned to compose music on its own with just a text description of what type of rhythm you want. Meta had also open-sourced MusicGen.
Felix Kreuk, an AI research engineer at Meta, revealed MusicGen in a Twitter thread last week adding that pre-trained music-model is available for open research.
Meta’s MusicGen
Meta trained MusicGen on 20,000 hours of music, including 10,000 high-quality licensed music tracks and 390,000 instrument-only tracks from ShutterStock and Pond5, a large stock media library, it said.
On feeding in a text description of the music you intend, MusicGen can create you a 12-second-long audio in just no time. These AI-Music models will not certainly replace human composers, but its pieces of work are reasonably good and convincing for just a simple text-prompt. Like you type “jazzy elevator music” and the AI will create a treat for your ears.
Adding to texts, you can also suggest a reference audio, like a track from an existing song to the AI by uploading it, for guiding the AI to compose a similar melody blended with your text prompts. MusicGen, however, is not capable enough to add chorus, or human voices in between the audio, like Google’s.
“MusicGen is built on top of the EnCodec audio tokenizer and is a single-stage transformer LM with efficient token interleaving patterns, which eliminates the need for cascading several models,” the research-paper writes. Means – MusicGen is solely developed without incorporating other language models into it and can generate high-quality audio samples, while being conditioned on textual description or melodies.
AI-generated Music – Examples
People had started experimenting the AI-music-generator, and here are comparisons of Meta’s MusicGen with Google’s MusicLM generated outputs.
Text prompt: ‘Lo-fi slow BPM electro chill’ with organic samples. Result:
Text prompt: The main soundtrack of an arcade game. It is fast-paced and upbeat, with a catchy electric guitar riff. The music is repetitive and easy to remember, but with unexpected sounds, like cymbal crashes or drum rolls.
Cons of AI-generated Music
Generative-AI is expanding its capabilities and, in the go, has concerned professionals involved in a realm either directly or indirectly. As far as music is considered, as these AI-generated music relies on its trained already existing tracks for composing audio, AI like MusicGen, MusicLM might produce music that sounds similar to an original soundtrack, leading to copyright issues.
Further, AI-generated music might infringe composers and albums, as the technology could possibly pull off famous artists’ voices without their consent, as it did with ‘Drake and the Weeknd’ case.
As AI in other fields demand for regulation ultimately to avoid risks for humans, music realm also urges for an ordinance to set things smooth.
Meta has made available its MusicGen to everyone. You can access it here.
Kindly add ‘The Inner Detail’ to your Google News Feed by following us!
Hope you find the page useful!
(For more such interesting informational, technology and innovation stuffs, keep reading The Inner Detail).