New Delhi (India), 13th June 2023: The Meta Audiocraft team has released a demo of their MusicGen AI, which can generate short audio snippets based on a text description. You can also start using samples of existing melodies.
It works quite simply; simply go to the demo website and describe what you wish to hear in the audio. They gave the algorithm a basic instruction, “80’s synthwave with catchy hooks,” and then hit the produce button. It took some time, but eventually delivered a 12 second clip.
After that they added another command to our prompt, instructing the machine to make “80’s synthwave with catchy hooks and jazz melodies.”
MusicGen was trained on 20,000 hours of music, including 10,000 “high-quality” licensed music files and 3,90,000 instrument-only songs from the ShutterStock and Pond5 stock media libraries, according to the business.
Felix Kreuk, Research Engineer at Meta AI research tweeted, “We present MusicGen: A simple and controllable music generation model. MusicGen can be prompted by both text and melody. We release code (MIT) and models (CC-BY NC) for open research, reproducibility, and for the music community.”
Google also released its own text-to-music AI system, MusicLM, to the public earlier this month, and it operates similarly to MusicGen.
