Add Weight to Keywords (instruments, genre, mood, themes etc.)
under review
Bybby Creme
It would be nice if you could determine the role of the keyword you enter. A bit like in Image generators such as Midjourney. For example "alternative rock:80" (as in 80%), "synth:30" etc. Then the result will be more rock with a bit of synth rather than keywords being completely ignored or those dominating which shouldn´t play much of a role.
So overall, the result is more what you imagined to be.
U
UdioAdam
under review
Ian Braunstein
Merged in a post:
An extra prompt window purely dedicated to genre, so we can guarantee the genre we want!
L
Lucid Fir
Go to ishkur's web of music or the enclyopedia metallum - genres have specific sounds. Ultimately you could even make your own AI powered genre definer. The point is that I want to be able to control the strength with which the prompt define genre to a far greater extent, currently it is difficult to reliably define genre.
Steven Williams
This would be helpful for instrumental music as well. For example, I tried making a trumpet and flute duo and I wanted to try to use weights in the Custom Lyrics box to try to tell Udio what instrument to play at what time, or both, such as.
First Generation:
[Trumpet:0.9]
[Flute:0.0]
Second Generation:
[Trumpet:0.0]
[Flute:0.9]
Third Generation:
[Trumpet:0.5]
[Flute:0.5]
Rimfrost
Steven Williams prompt it, and most accurate it will be doing it with the tags in lyrics field. It does everything almost perfect if telling ut to!
Steven Williams
Rimfrost when I use the prompt box, I am unable to specify the order or timing as done in the custom lyric box.
I can already call a Trumpet and Flute duo in the prompt box but that alone doesn't give me much control over the sequence or combination or the two.
Would you have any suggestions?
Rimfrost
Steven Williams I mean as you will never get any 50/50 or 30/70 more than telling the AI "weight" it like this with your suggestion. I mean that could you prompt in the way that suits you. Try to just write it in your native language. You can have some control but it's the AI that decide to last whatever system, to give it instructions, but it will not get better of understanding in other ways. [instrumental solo break flute accompagnato trumpet] or [a due flute e trumpet] but in Swedish I would just have typed like [Endast flöjt och trumpet som spelar där flöjten ska ha lite mer tid] (If you only do instrumental then "solo break" seems wrong, but just write as to a buddy that should play. I have been surprise of how good it was just writing completely normal". But there are for sure a lot of music terms to use to reach what you look for.
Steven Williams
Rimfrost I just figured since weights could be added to stable diffusion models that create AI Graphics, that maybe the same could be done with Large Models that create music.
Here is an image of the ComfyUI tutorial that talks about adding weights to words in the prompt to tell the model how much it should influence the outcome.
(ignore the anime, I didn't make the tutorial)
iNever
It wouldn't do anything other than prompt it. There is no guarantee of getting the genre you want right now, that's just due to how the model works.