"A light electric dance music", with this description, what is the music that flashes through your head?
Google released a MusicLM, can use a simple text description or even an image to generate music, can be said to be the composition world ChatGPT.
Of course, if you can go directly to this link to listen to feel the natural best, but I can still try to point out some very interesting places, such as the above with textual narrative content, in fact, textual narrative is far more complex than "a light electric dance music" also, as is easy to remember, the music does sound very consistent with the description.
Here there is also a simple description made 5 minutes of the tune.
I even heard some suspected lyrics of Indian songs, which was consistent with the scenario change of continuous music.
There is even a music library with a mix of multiple tunes.
The most powerful thing is to compose music based on the picture and the narrative of the picture, from the more realistic style of Napoleon to the more abstract and unrealistic mood of the above picture can be expressed, although I listened to more or less still feel and their own imaginary music discrepancies, but the mood of the painting can still be felt.
The great thing about this AI tool is that it can create music that fits the scenario, the style, the instruments and even the era.
It is understood that this MusicLM used 280,000 hours of music data to train, it can create a good few minutes of 24kHz music, as for the creation of music quality and creative ability, you can look at the various cases above as well as to the site to listen to the field once.
MusicLM's capabilities are undoubtedly powerful, but there are still some problems, such as the inescapable material infringement controversy, and the problem of incomplete technology, such as the part of the vocal sounds like a very good thing, but the meaning of the lyrics is not clear at all, and the AI can also be "lazy", creating 1% of the library is directly The AI is also "lazy", and 1% of the library is taken directly from the material it was trained on (but this is much better than many human composers who plagiarize). In any case, the AI's ability to compose is worthy of our continued attention.
What do you think this AI? Is it a game changer for music industry? Please leave a comment below!
