Always something new or evolving in the world-meet Melisma, AI generated instruments from Music XML files. In other words, you have it read your notation and it interprets it according to the info you provide -articulations, volume etc.
The woodwinds in particular sound quite good, the strings can sound pretty good though it sounds to me like they need further development. One Japanese composer/developer.
No MID, btw:
https://kagura-music.jp/melisma-top
There's more than meets the eye
Register now to unlock all subforums. As a guest, your view is limited to only a part of The Sound Board.
Melisma
-
My name is Nobody
- Posts: 56
- Joined: Sep 05, 2019 5:31 am
Re: Melisma
The phenomena is interesting, AI interpreting sheet music.
But the result to me sounds like a Roland sound canvas from yesteryear. ( ok, a tad melodramatic)
If I think of any instrument and all that is involved in mastering not only the instrument itself, but also the many ways in which musicians interpret how to play any piece of music from sheet, I think for now it is worlds away before AI musicianship can become a reality, if even possible.
But who knows ;-)
But the result to me sounds like a Roland sound canvas from yesteryear. ( ok, a tad melodramatic)
If I think of any instrument and all that is involved in mastering not only the instrument itself, but also the many ways in which musicians interpret how to play any piece of music from sheet, I think for now it is worlds away before AI musicianship can become a reality, if even possible.
But who knows ;-)
Re: Melisma
significantly better than I expected, but I didn't expect much.
It is an interesting idea. but a long way from being useful to musicians.
Now the general public, sadly, will always go for "good enough"
It is an interesting idea. but a long way from being useful to musicians.
Now the general public, sadly, will always go for "good enough"
Re: Melisma
The interesting part to me is the lack of midi. I used to be somewhat better than I am now with notation, but I have avoided it for many years. I imagine the user base for this is going to be pretty specific.
Re: Melisma
The woodwinds sound very good. The strings don't. I find this approach very interesting. Sample libraries target the market of composers who compose in a DAW. If you compose in notation, you have twice the work. One step is the writing process, and then you have to create the mockup. If Melisma progresses so far that it rivals a good mockup, it will be a great tool for all of us who work in notation.
Re: Melisma
A french horn is the next addition to Melisma. I'm still thinking what the potential of this technology is if a single person can get such results. The basis for the training are recordings in a small room. I assume that's why the strings sections don't sound convincing. I can only assume that the results would be much better if you can use a proper string section, recorded in a great room, as the training material.
There's a project I need to finish first. When that's done, I'd like to give Melisma a proper test run.
There's a project I need to finish first. When that's done, I'd like to give Melisma a proper test run.
Re: Melisma
It needs to connect to the web right? I don’t want to have my materials be out in the space, plus I work with copyrighted music so that’s a no for me
Re: Melisma
Melisma doesn't use the uploaded music to train the algorithm. But there is also a desktop version. With that, your music stays on your pc and you don't have to upload it to a cloud.