I am using Synth V pro with Solaria voice bank, i noticed some of the phenomes have a crackle, the word “thing” has a small crackle near the end, the only way i have figured out how to get rid of it is to set Voice options Sythesis engine Mode to “prefer speed” but that takes too much away from the actual voice, Is this just something we have to deal with using this technology right now or is it just with certain recorded parts in sound banks, i am sure this is not very common right?
Also i am using it multi channel out in Reaper and i notice when you solo an aspiration track there is like a tin whistle sound, it is all over the frequency range so it cannot be removed using surgical EQ tricks, not that i can figure out anyway. Any tips for these things?
I am thinking i need to get more voice banks to see if the others have the same thing going on… Solaria still sounds awesome tho. I have just noticed those 2 problems, but probably nothing we can do about it right?
I have tried all the settings to see if there is any difference, only changing that one option did anything. And changing from prefer quality to prefer speed does not sound very nice.
I could not fix it on their original file tho. not sure why it was happening.
I remade the sentence and it is gone. It was happening on the word “something” in the Solaria Demo file they give you. The crackle was audio noise for sure, a bit strange, not sure why it happened on their demo but when i just remade it , it is gone, I put the same settings on the note too.
You really only notice the tin metallic whistle on the aspiration when you solo it in a DAW program like reaper, it is noticeable, maybe not on other voice banks.
But she still sounds great, there are lots of good covers out there using Solaria voice bank without multi channel routing, so i will try different ways. Thnks for the replies.
I’ve heard the “whistle” on a number of the English AI voices, and have reported it as a bug.
Since aspiration noise is the unpitched, non-harmonic portion of the vocal, this should never be happening. But it’s definitely here, and sometimes loud enough to be quite noticeable.
You can try to find the approximate frequency and place a notch filter on the aspiration track, but there’s still going to be some leakage.
The crackle sounds like a transient bug, as it disappears when you remake the track. I’ve heard it on a number of sounds, and I think it’s more than just a vocal fry, since you’d have to be doing bad things with your voice to get that result. Vocalists shouldn’t be singing with a gargle sound.
Fortunately, you can just just edit it out of the aspiration track without any real loss. Doing this in Reaper is trivial.
Yea i tried for hours using narrow EQ cuts to try to take it out, EQ will not work, I have never tried a Notch Filter plugin before, I found a couple free ones, Will have to watch a youtube video on what they do.
The transient sound makes sense tho, that is probably what it was making that crackle, not sure why. I have a few other questions now though
What does putting an apostrophe before a word do, i saw on the Solaria Demo files they used this often, examnple: 'As … they put the apostrophe first… what does that do… also does making a capital letter rather than lowercase change the sound or volume or anything, i kind of think it does not right?
I also noticed they did a few things. for breath notes… they had changed the phenome on a br note to “sil”… what does that do? it sounded like possibly a smallI lip-smack type noise. I saw them put hh as a phenome above a br note as well, just kind of uncertain what these things do… is there a list anywhere online for this kind of stuff. like a list for the extra note options like breaths and stuff?
oh thanks, i did now know there is a notch setting in the reaEQ, have never seen anyone use it yet for that. and thanks for the description for cl br and sil… but yes there needs to be an online dictionary for all these things and for alternate wordings and phemone changes.
For example I have been going over the Solaria demo files and they have done some neat re-wording of words so it sounds correct like for “child” they had “chai uld” not in the phenome but for the actual lyrics, instead of record they put wreck ord, i am sure for every voice bank it is a bit different and you can find little re-wordings to get better or different pronounciations.
Also I noticed they did a lot of custom phenome changes to words, I have been writing them all down in a list as I go over each track, taking my time doing this. I think if there was a collaborative effort to put together online library for these things that would be great.
I am still wondering why they had an apostrophe at the beginning of certain words in the lyrics like 'as or 'i’ll but then i saw them also write i’ll normally, is there a difference i wonder by putting the apostrophe before a word.
There’s a difference, but it’s not always apparent, depending on the phoneme.
For example, the stop in 'taco can be heard clearly.
I suspect that the glottal stop is merely being used to add a short silence, where sil doesn’t work as well.
As far as custom phonemes go, I find it simpler to just sound them out, because they depend on the context. Often you may add or drop a phoneme from a word because that’s the way it’s sung in a particular song. But that context may never come again, so why put it in the dictionary?
But whatever works for you!
If you do use phonetic replacements, be sure to write it into the word using the dot notation, like .k ae t, instead of replacing the phonemes above the note. Otherwise, you may change the word later, and not notice the phonemes have changed.
Just to add onto the apostrophe thing, I remember hearing that Solaria also has this thing where, depending on the context, adding that apostrophe gets you a little vocal fry at the beginning of a note. I can’t remember who I heard that from or where I heard it but I’m pretty sure it’s a thing, and that’s why you can find it in Solaria demos.
What if someone made a script to be able to mute single notes or groups of notes. I think that could be useful to mute things on and off for listening back and adjusting. Just an idea. I would not know how to make scripts.
I just had a look in the additional content for Solaria, and noticed in the scripts (of all places) that there’s a script called EnglishPhonemeReference_EclipsedSounds.js that explains a number of phonemes.
While I’m glad the list it exists, I would have expected to find it in a doc subfolder, not in the scripts folder!
Here’s the slightly edited version of that list for the “special” phonemes:
sil(Silence) - a phoneme you can insert into notes for a pause. Any amount of space between notes is also considered a sil.
cl(Stop/Glottal Stop) - a phoneme you can insert to have the vocal perform a stop, cl can be used to accurately input words like uh-oh or used for singing effect. In some AI voice libraries, a cl between or before vowels will sometimes cause the library to perform a vocal fry, this is context dependent, but can provide a unique effect.
-(Melisma) The - sign can be placed in the lyrics section of a note to extend the previous syllable over the next note, carrying the sound into the current note. If many consecutive notes have - in them, they will continue to carry the syllable over multiple notes
+(Syllable) A + sign can be inserted in the lyrics section of a note to split the word in the previous note into multiple syllables, carrying the next syllable into the current note. If many consecutive notes have + in them, they will continue to pass on the syllables.
I made it into a special hotkey, the 2 Eclipse sound Scripts. Phenome Refererence, and Replace Phenome. Both are very useful. That was very helpful. Thanks again, I went over every option and put all into hotkeys, then I am printing it off to a list to memorize. I do the same with Reaper for working in midi and editing audio using stretch markers. Using hotkeys makes things a lot easier. Thanks again for the help.
