Google AI turns text into music

Google released a document about MusicLM on Jan. 27, stating the developer trained the AI ​​using a database of 280,000 hours of music to create complex tracks across genres.

MusicLM not only combines instruments, but can also create musical works based on abstract concepts that are difficult for computers to understand. He is even able to compose music or paint a picture based on a whistle. MusicLM can combine different lyrics to create complete musical works.

However, this artificial intelligence still has some limitations, such as: B. not understanding singing. Many of his songs are also duplicates, as opposed to works written by humans.

The developer has not released MusicLM for copyright reasons. About 1% of AI-generated music is copied directly from songs in the training database.

Since the middle of last year, a number of AI applications have also been born that create images from text, such as Lens, Midjourney or Dall-E, which are causing a sensation. With just a few keywords, users can ask AI to create stunning works, from landscape photos to paintings of French nobility to fake vintage photos. Artists, however, dismiss these tools, claiming that AI is copying artists’ art styles without permission, encouraging piracy and de-beautifying art.

dive app (in accordance Engadget)