Meta unveiled a new artificial intelligence model for generating songs from text

Meta company along with Audiocraft research team unveiled a new open source artificial intelligence model. This model named MusicGen, like ChatGPT, can generate songs using input in the form of text. It will even be possible to receive a song sample and change it according to the user’s descriptions.

According to Tekna technology and technology news service, users can inform MusicGen of their desired music style and then give a part of a song to this model. In just a few seconds, you can find a song according to the user’s description or changes. Currently, the demo of this program is on Hugging Face Facebook site. Users can produce a similar song by describing the desired song, for example songs from the 1980s.

In the development of this artificial intelligence, 20,000 hours of licensed music have been used. Also, Ade Meta database and other sites have been used for training. Also, the training process using the 32,000 kHz audio tokenizer related to EnCode has also been used.

The MusicGen model is open source and it is possible to use it to produce commercial songs. This software shows the speed of artificial intelligence development in different fields. Also, machine learning can be a threat to the music market in the future.