Google recently released a research paper titled “MusicLM: Generating Music From Text“.
What’s notable is that experiments show that MusicLM outperforms previous systems both in audio quality and adherence to the text description.
MusicLM has been trained on a dataset of 280,000 hours.
A bunch of samples produced by the Model have also been shared on the following site.
There are still ethical challenges that needs to be dealt with.