nice. i'm not familiar with with reaper but i expect it's probably similar?
i've tried it with priming with some preamble to 'set the tone' with mixed results. best combo ive found is to choice a voice that will introduce more emotiveness and non-word noises (moans, grunts, etc), set the stability really low, like 5 to 10 and set the clarity fairly low like 20 to 40. if the stability is too low then you get lots of weird random noises and if the stability is too low then the voice can drift off to weird accents, other voice characteristics, even switch genders. its weird but not all voices will emote like i want and so thats just trial and error to figure it out. rachel seems to be best when getting started.
definitely have to keep with first person throughout. if theres a mix of narrator/third person and first person, even in quotes, it tends to make the whole thing sound more 'narrator-y'.
11labs appears to be token-based rather than character based and token can be used to introduce sounds. best tokens that i've found are mixed serieses of dots, commas and dashes like:
, ,,
. .. ... ...
- --
i pretty much put ellipsis between every sentence
non-word strings of letters trying to make sounds work sometimes but often have to be cut because sometimes they just sound really strange. i tend to use a lot of them in a bunch of strange combinations and knowing im going to be cutting at least half of them typically. so that's stuff like:
Haaa Aaah Oh oooh Mmmmh Uuuhng haaaaah
when the stability and clarity are low, the ai will try to create a sound for those tokens (usualy). also, writing in a style that introduces a lot of first person emotion seems to make it emote more, so mixing up sentence structure to be more everyday-speech like and introducing first person feeling words and exclamation points helps, too. so i really try to avoid inputting sentences that have a academic/journal/news/narrator/novel vibe.
word repetition can force more emotion the later words, too. and uppercase will sometimes increase emphasis and emotion in the voice.
So this is pretty plain vanilla narrator:
I'm so hugely pregnant. I can't believe I'm ten months pregnant with quadruplets. My belly is as big as a beachball and weighs 150 pounds. I can't even move. There's so much movement in my belly; the babies just won't stop moving, they're so ready to come out. Half my body is my womb. I think I must be the most heavily pregnant woman ever. But I'm so aroused by being this pregnant.
i'd input it more like this to get a lot of emote:
... Oohh, I'm so hugely pregnant! ... I -- uhh -- I can't believe I'm TEN months pregnant - with quadruplets!! ... My belly, my belly is as BIG AS A BEACHBALL .. -- and weighs ONE HUNDRED AND FIFTY POUNDS! ... oh, dear god, . I just can't even .. ,, I CAN"T EVEN MOVE! ... There's so, so much movement in my belly .. oooff! -- these babies just won't stop moving! ... They're so ready to come out! ... Ooh, oh god,, my body is HALF WOMB! ... I must - I MUST - be the MOST heavily, heavily, heavily pregnant woman - EVER! ... But,, but, I'm so - AROUSED - by being THIS pregnant! ... aaahhh ... mmmmh ...
i probably have 50 percent of the tokens as non-words
can't wait to hear what you make!
i've tried it with priming with some preamble to 'set the tone' with mixed results. best combo ive found is to choice a voice that will introduce more emotiveness and non-word noises (moans, grunts, etc), set the stability really low, like 5 to 10 and set the clarity fairly low like 20 to 40. if the stability is too low then you get lots of weird random noises and if the stability is too low then the voice can drift off to weird accents, other voice characteristics, even switch genders. its weird but not all voices will emote like i want and so thats just trial and error to figure it out. rachel seems to be best when getting started.
definitely have to keep with first person throughout. if theres a mix of narrator/third person and first person, even in quotes, it tends to make the whole thing sound more 'narrator-y'.
11labs appears to be token-based rather than character based and token can be used to introduce sounds. best tokens that i've found are mixed serieses of dots, commas and dashes like:
, ,,
. .. ... ...
- --
i pretty much put ellipsis between every sentence
non-word strings of letters trying to make sounds work sometimes but often have to be cut because sometimes they just sound really strange. i tend to use a lot of them in a bunch of strange combinations and knowing im going to be cutting at least half of them typically. so that's stuff like:
Haaa Aaah Oh oooh Mmmmh Uuuhng haaaaah
when the stability and clarity are low, the ai will try to create a sound for those tokens (usualy). also, writing in a style that introduces a lot of first person emotion seems to make it emote more, so mixing up sentence structure to be more everyday-speech like and introducing first person feeling words and exclamation points helps, too. so i really try to avoid inputting sentences that have a academic/journal/news/narrator/novel vibe.
word repetition can force more emotion the later words, too. and uppercase will sometimes increase emphasis and emotion in the voice.
So this is pretty plain vanilla narrator:
I'm so hugely pregnant. I can't believe I'm ten months pregnant with quadruplets. My belly is as big as a beachball and weighs 150 pounds. I can't even move. There's so much movement in my belly; the babies just won't stop moving, they're so ready to come out. Half my body is my womb. I think I must be the most heavily pregnant woman ever. But I'm so aroused by being this pregnant.
i'd input it more like this to get a lot of emote:
... Oohh, I'm so hugely pregnant! ... I -- uhh -- I can't believe I'm TEN months pregnant - with quadruplets!! ... My belly, my belly is as BIG AS A BEACHBALL .. -- and weighs ONE HUNDRED AND FIFTY POUNDS! ... oh, dear god, . I just can't even .. ,, I CAN"T EVEN MOVE! ... There's so, so much movement in my belly .. oooff! -- these babies just won't stop moving! ... They're so ready to come out! ... Ooh, oh god,, my body is HALF WOMB! ... I must - I MUST - be the MOST heavily, heavily, heavily pregnant woman - EVER! ... But,, but, I'm so - AROUSED - by being THIS pregnant! ... aaahhh ... mmmmh ...
i probably have 50 percent of the tokens as non-words
can't wait to hear what you make!
