mind&magic
Proud Member
- Joined
- Apr 22, 2021
- Messages
- 49
- Points
- 18
We'd all love to hear new original MJ content, but unfortunately the next posthumous release of unheard music is still a matter of wishful thinking and speculation.
So how about creating our own new MJ songs using AI?
I think most of you are probably familiar with Jukebox AI https://jukebox.openai.com/. This AI analyzes published songs by an artist (instrumental and vocal together) and creates new songs in the artist's style. Even if the style of the music sometimes sounds like MJ, the songs are not really enjoyable because the voice sounds weird.
But what if you took it a step further by having an AI analyze the instrumental and the voice separately? It would also be possible to analyze only the acapella and reproduce MJ's vocal. This is not science fiction, YAMAHA has done something like this with Hibari Misora, a (dead) Japanese singer in 2019:
"Yamaha Corporation announces that it has succeeded in reproducing the vocals of the late Hibari Misora, a legendary Japanese singer, using its proprietary VOCALOID:AI vocal synthesis technology, in technical collaboration with a NHK (Japan Broadcasting Corporation) television broadcast*1 that aired in Japan on September 29, marking both the public debut and the first real-world implementation of VOCALOID:AI."
You can listen to it here:
Read more: https://www.vocaloid.com/en/news/vocaloid_ai_news_release/
In my world, it would be creepy and like christmas at the same time to use this technology to create thousands of "new MJ" fan songs. What do you think?
So how about creating our own new MJ songs using AI?
I think most of you are probably familiar with Jukebox AI https://jukebox.openai.com/. This AI analyzes published songs by an artist (instrumental and vocal together) and creates new songs in the artist's style. Even if the style of the music sometimes sounds like MJ, the songs are not really enjoyable because the voice sounds weird.
But what if you took it a step further by having an AI analyze the instrumental and the voice separately? It would also be possible to analyze only the acapella and reproduce MJ's vocal. This is not science fiction, YAMAHA has done something like this with Hibari Misora, a (dead) Japanese singer in 2019:
"Yamaha Corporation announces that it has succeeded in reproducing the vocals of the late Hibari Misora, a legendary Japanese singer, using its proprietary VOCALOID:AI vocal synthesis technology, in technical collaboration with a NHK (Japan Broadcasting Corporation) television broadcast*1 that aired in Japan on September 29, marking both the public debut and the first real-world implementation of VOCALOID:AI."
You can listen to it here:
Read more: https://www.vocaloid.com/en/news/vocaloid_ai_news_release/
In my world, it would be creepy and like christmas at the same time to use this technology to create thousands of "new MJ" fan songs. What do you think?