Music Gen, created by the social media company Meta, generates music. “An ’80s driving pop song with heavy drums and synth pads in the background” can be used to generate sounds. Music Gen has a 12-second demo online. Reference audio can now influence Music Gen. Music Gen attempts to match the melody and description of an existing song. This optional function lets users manage music generation.
Music Gen, Meta’s latest innovation, was trained on a large music library. The training dataset includes 20,000 hours of music, including 10,000 high-quality licensed tracks and 390,000 instrument-only files from Shutterstock and Pond5, two popular stock media libraries. The company now offers pre-trained models to the public. The vendor has not shared the code for training these models. These pre-trained models require a 16GB GPU.
Music Gen was just assessed. The speaker recently stated that while technology can generate music, it cannot replace human artists. A music review appreciated its melodic quality. The reviewer called the “ambient chiptunes music” melodic. The critic even compares the music to Google’s AI music generator, MusicLM. The team won’t win.
Using generative AI to make familiar-sounding music is getting viral. These tunes are convincing. In recent fights, music companies have quickly notified streaming partners of copyright infringement concerns and won. “Deepfake” music may violate artists’, labels’, and other rights holders’ copyright.
Must Read: Mobily Postpaid
Rules may be forthcoming. Music-making AI may be affected by several ongoing litigations. One complaint concerns artist whose work is used to teach AI systems without their consent.
Meta states that all music used to train Music Gen was “covered by legal agreements with the right holders,” such as a Shutterstock arrangement.