This Wednesday, Meta released an open-source AI solution called AudioCraft that lets anyone make music from text prompts.
Just like ChatGPT lets you write stories or summaries, the Meta AudioCraft lets you input text and spit out a song.
The AudioCraft bundle contains three bundles, AudioGen, EnCodec and MusicGen, and works for music, sound, compression and generation.
How did Meta build AudioCraft?
Trying to avert questions about what they trained their algorithms on, Meta specifically said that they created MusicGen using only company-owned and specifically licensed music.
For those who haven’t kept up with the generative AI boom, many artists are raising the alarm over copyright violations, saying that AI software like ChatGPT is trained by scraping copyrighted work.
For writers and students, things got so bad, people now have to do creative workarounds to prove they didn’t use ChatGPT in their works. Let’s see what happens on Bandcamp next!
The current actors and writers strike in Hollywood is partially due to fears of how AI will be used in entertainment, partially because of how actors do not get residual payments from streaming apps, unlike with traditional media.
Of course, there is also this lawsuit from comedian Sarah Silverman, who says ChatGPT was able to summarize her work so well because it was trained on a pirated data set.