Synthesizer V Knowledge Resource

Workflow for Complex Projects

While working with the plugin version of Synthesizer V (SV) in Cubase on a fairly complex song containing several tempo and time signature changes, I realized that it was better to create one SV project for each song section (verse, chorus, bridge, etc.) rather than put them all in the same project. I then bounce each section to an audio file and drag it into the appropriate place in my DAW project. There are several reasons for doing it this way.

First (as of the time of writing this post) SV cannot follow time signature changes and so you have to enter these manually, and this can become tedious if you have a lot of them.

Second if you change your song structure by shifting sections around (as I do), you'll have to move the time signature changes in SV as well to keep it in sync with your DAW.

Third creating one project for each section will reduce CPU usage because the program only has to render one vocal phrase rather than all of them in an entire song. This also saves you a bit of time.

Fourth, the above approach greatly reduces the amount of time spent scrolling in SV to find the phrase you want to edit if, for example, you want to change your lyrics or vocal melody.

Note: It's helpful to place the first note of your vocal phrase on measure 2 in SV so that when you bounce it you'll have one measure of silence before the waveform. This will enable you to adjust waveform timing in your DAW. (I'm using Cubase, so this may not be necessary in other DAWs.) Similarly I place a short note two measures after the last note in my melody to ensure that I'll have a full measure of silence at the end of my bounced phrase (in my DAW I cut the final measure containing the short note leaving one measure of trailing silence).

Cheers...

you should definitely check and install some script to enrich the keybinding possibilities. I have scripts (sorry, i did not marked the download link) to assign keys to note creation, moving notes with arrow keys, shorten/increase length, add breath, etc.
in the next days i will provide a zip with all the scripts i found most interesting

another trick: i saw in some youtube video that, if you don't like the instant mode vibrato or simply if you want to delay it, you could split the note, using "-" to hold the syllabe, and so the starting of the legato will not happen until the very end of the very last note. I never tried this one yet.

(Minor contribution discovered today working in a real project)
Grids - Triplets: If you need triplets in a project (for swing or "shuffled" style), "adaptive" grid does NOT work. Set manually to 1/6 quarter (or 1/12 or 1/24). I could not see these options, I was expecting the more common naming 1/4T, 1/8T or 1/16T respectively:

I find that the vertical bars indicating the measures are very difficult to see at some piano roll magnification, they are confusing and often I end up in wondering where the bar or the quarter starts.

so I used this trick:




I put short notes every quarter on antoher track (which I mute) and I have a handy reference for the track i'm working on

I just found this in my feed:

doctoremmet said:

I just found this in my feed:

Click to expand...

Soundformore tutorials are great - thanks for posting this. He does loads for iOS - I hadn’t realized he did other stuff as well.

This may be helpful here. Some of the script repositories collected from the "big" thread and other places:

https://github.com/claire-west/svstudio-scripts/tree/main/hotkey-scripts
https://github.com/hataori-p/real-voice
https://github.com/Dreamtonics/svstudio-scripts
https://github.com/dcuny/synth-v-scripts
https://gist.github.com/so-c/d2ce848b3b209c9b496effacfaf37246 (change track color, handy)

A search on "SynthV" on GitHub brings up several more.

Also attaching a script I wrote that helps implement bosone's idea above. (Edit 2/25/23: Updated script to fix a bug and enhance it some.)

SPANISH language.

Some hint on using Synth-V in Spanish: Japanese phonetics are closer to Spanish than Chinese or English. For example, formants on vowels "a-e-i-o-u" are almost a perfect match. So the vowel phonemes are there. Some of the more exclusive phonemes (a good way to remember them is the consonants in word "Rejoncillo") are POSSIBLY not present, but I am not sure yet.

I came across the idea to try to find a phonetic dictionary to use Spanish lyrics with Japanese phonemes. To my surprise there is no such a dictionary created by users (!), but there is one created by Eclipsed Sounds (!!!!).

The result is: If I use Solaria or Kevin, set them to "Japanese" and then use their dictionary JAP-->SPA, I can type lyrics in plain Spanish and get a very decent result without looking manually for phonemes.

Vowels are perfect in my opinion. Some consonants (like strong "RR") simply don't exist in the original (English) material recorded by Solaria. I have not tried yet Spanish "J", "Z", "Ñ" or "LL". But to me the result sounds close to "una guiri cantando en perfecto castellano". Perfectly usable. Vowels are OK because Synth-V Studio can move formants in AI voices (to "try" to sing in a different language).

I suspect that this could work better or even much better if the original voice is native Japanese, but I do not have any other AI voice to test, and I don't know if Eclipsed Sounds dictionary will work with voice banks by other vendors.

I hope to find some time to experiment with Solaria and Kevin in the week-end. Anyone with Japanese voices who wishes to try Spanish?

Ah, I found possible -and easier - solutions for German, Italian and French, but I have no idea how they sound (and I have no way to judge if results are decent or excellent).

On the other hand I can confirm something I found in the forum: Adding a few dictionaries in the proper directory makes loading of Synth-V EXTREMELY slow. I recommend to move to other place dictionaries that are not required.

parapentep70 said:

I suspect that this could work better or even much better if the original voice is native Japanese, but I do not have any other AI voice to test, and I don't know if Eclipsed Sounds dictionary will work with voice banks by other vendors.

Click to expand...

parapentep70 said:

I hope to find some time to experiment with Solaria and Kevin in the week-end. Anyone with Japanese voices who wishes to try Spanish?

Click to expand...

Mai is a native Japanese AI voice, and is free to download for anyone with a SynthV Pro license.

mallux said:

Mai is a native Japanese AI voice, and is free to download for anyone with a SynthV Pro license.

Click to expand...

I tried, but for some reason this voice does not take the dictionary.

Maybe this voice is "Standard" (not "Lite", but not necessarily "AI" either? I think it maybe older than "AI" families).

Anyway "SOLARIA_JPN-_ESP" dictionary works with Solaria and Kevin. Not with Mai. To my surprise the dictionary appears as valid for Tsurumaki Maki AI ENG Lite and Also Eleanor Forte Lite. English is the only option I tried. But in a short test I saw results absolutely useless, all phonemes are corrupted, which makes sense because the original language is English.

I think the simple explanation for the dictionary not appearaing with Mai is that this voice is older than AI extensions (cross-lingual capabilities). I'd like anyone in the forum with a Japanese voice to run a short test

Next step for me will be to download another "Lite + AI" voice but in Japanese instead of English. After work!

parapentep70 said:

I tried, but for some reason this voice does not take the dictionary.

Click to expand...

The dictionary doesn't show up because it's originally made for English voicebanks.
Since now AI voicebanks support cross-language synthesis, you can "cheat" the system by copying the same .json dictionary in "Documents/Dreamtonics/Synthesizer V Studio/dicts/japanese-romaji" (or the equivalent path to your SynthV dictionaries folder), so it can also show up with Japanese voices

Mai is definitely a full AI voicebank, it got released with the 1.8 update of the program last November.

raidergale said:

The dictionary doesn't show up because it's originally made for English voicebanks.
Since now AI voicebanks support cross-language synthesis, you can "cheat" the system by copying the same .json dictionary in "Documents/Dreamtonics/Synthesizer V Studio/dicts/japanese-romaji" (or the equivalent path to your SynthV dictionaries folder), so it can also show up with Japanese voices

Mai is definitely a full AI voicebank, it got released with the 1.8 update of the program last November.

Click to expand...

No, the dictionary is for JAPANESE banks. It makes some sense for Spanish.

The other custom dictionaries provided by Eclipsed Sounds for German and Italian are all for English banks. You can see in my list of dictionaries. In fact these dictionaries are in folder ....Documents\Dreamtonics\Synthesizer V Studio\dicts\english-arpabet and they work with Solaria and Kevin ONLY if set to Japanese.

But thanks, now I understand the structure of folders (I only have /english-arpabet under /dicts, not any /japanese-romaji).

The dictionary is made for cross language Japanese, yes, but it's made for Solaria, an English voicebank, so it was put into the "english-arpabet" folder because that's where native English voices look for dictionaries. If you create a "japanese-romaji" folder and put it there it'll show up for Japanese voicebanks as well, because that's the folder where native Japanese voices will look for dictionaries

Writing this out, I realize this might be something that can be improved in future versions of the program. Make it so it will also look in the folder of the currently selected language, instead of only the native language of the voicebank. This would only work for whole track cross language synthesis, of course, and not for note based one, but still.

@LenKiMo (I hope you don't mind tagging you), do you think this would be a good addition?

parapentep70 said:

SPANISH language.

Some hint on using Synth-V in Spanish: Japanese phonetics are closer to Spanish than Chinese or English. For example, formants on vowels "a-e-i-o-u" are almost a perfect match. So the vowel phonemes are there. Some of the more exclusive phonemes (a good way to remember them is the consonants in word "Rejoncillo") are POSSIBLY not present, but I am not sure yet.

I came across the idea to try to find a phonetic dictionary to use Spanish lyrics with Japanese phonemes. To my surprise there is no such a dictionary created by users (!), but there is one created by Eclipsed Sounds (!!!!).

The result is: If I use Solaria or Kevin, set them to "Japanese" and then use their dictionary JAP-->SPA, I can type lyrics in plain Spanish and get a very decent result without looking manually for phonemes.

Vowels are perfect in my opinion. Some consonants (like strong "RR") simply don't exist in the original (English) material recorded by Solaria. I have not tried yet Spanish "J", "Z", "Ñ" or "LL". But to me the result sounds close to "una guiri cantando en perfecto castellano". Perfectly usable. Vowels are OK because Synth-V Studio can move formants in AI voices (to "try" to sing in a different language).

I suspect that this could work better or even much better if the original voice is native Japanese, but I do not have any other AI voice to test, and I don't know if Eclipsed Sounds dictionary will work with voice banks by other vendors.

I hope to find some time to experiment with Solaria and Kevin in the week-end. Anyone with Japanese voices who wishes to try Spanish?

Ah, I found possible -and easier - solutions for German, Italian and French, but I have no idea how they sound (and I have no way to judge if results are decent or excellent).

On the other hand I can confirm something I found in the forum: Adding a few dictionaries in the proper directory makes loading of Synth-V EXTREMELY slow. I recommend to move to other place dictionaries that are not required.

Click to expand...

In my opinion, that Solaria's Spanish dictionary is not worth it as it's based on a (very limited) list of words. As in the Spanish language there is a very high phoneme-grapheme correspondence, I would bet on making a syllabic dictionary since it is very small (only 2316 possible syllables). The introduction of the lyrics in the editor would be in syllables instead of words, obviously.

However, not all syllables will sound good with Japanese phonemes and in some cases I would opt for English phonemes, having to change the language at the note level (which is currently possible).

One of the most problematic phoneme would be the Spanish J, but with a lot of Breathness and the growl script you can add a bit of grit to the /hh/ or /h/ phoneme.

I would like to try some of this but lately I have not found time for this hobby!.

Blancanegra said:

In my opinion, that Solaria's Spanish dictionary is not worth it as it's based on a (very limited) list of words. As in the Spanish language there is a very high phoneme-grapheme correspondence, I would bet on making a syllabic dictionary since it is very small (only 2316 possible syllables). The introduction of the lyrics in the editor would be in syllables instead of words, obviously.

However, not all syllables will sound good with Japanese phonemes and in some cases I would opt for English phonemes, having to change the language at the note level (which is currently possible).

One of the most problematic phoneme would be the Spanish J, but with a lot of Breathness and the growl script you can add a bit of grit to the /hh/ or /h/ phoneme.

I would like to try some of this but lately I have not found time for this hobby!.

Click to expand...

This dictionary is not that limited or syllabic! I opened it and it contains some 160K words. Comparable to a GOOD Spanish dictionary (perhaps less real words, but includes all verb variations). Perhaps the reason for long loading times. English --> German is even larger : 278.000 words!

"Solaria" dictionary has nothing to do, it is an ENG--> ENG dictionary intended to correct pronunciation problems or homographic words like wind (a clock) vs wind (from the North). This one is about 1250 words.

Small update on Synth-V in Spanish.
In summary I don't know the reasons but... to get some results the recipe is this:
-Use Solaria or Kevin set to Japanese
-Put dictionary SOLARIA_JPN-_ESP in folder ...Documents\Dreamtonics\Synthesizer V Studio\dicts\english-arpabet (not a typo! putting it in japanese-romaji does not work!!!).
-Select THIS dictionary.
-Enter lyrics in plain Spanish, separated by (normal Spanish) syllables.
-Results have some "mild English accent", but not bad at all in most cases:
-Strong "RR" sound --> definitely not native!
-"J" --> Not native, not bad.
-"Z" --> Not from Spain, but latin accent ("seseo"), but other than this it works in 95% of the phonemes.

In a short test I used a random text:
"Lo siento, pero como modelo de lenguaje entrenado no tengo acceso a información en tiempo real sobre personas específicas"

Some phonemes have some accent (that can be corrected later up to a point), but they are ALL correct with the exception of "personas". The dictionary says it shoudl be pronounced as "p er s ax n ax z" but for some reason the last "z" is missing (!!), and Kevin sings "persona" in singlular. Why? No idea. But it is just 1 exception (see below)


And this is Kevin singing the stupid melody with the non-sense lyrics. Without ANY correction at all. I think it is a good starting point.

Veredict: It works... up to a point, I have no idea about the behaviour respect the location of the dictionary and I do not have high hopes to improve accent automatically because japanese does not contain the missing phonemes. I hope I can improve manually the missing / wrong phonemes.

parapentep70 said:

Small update on Synth-V in Spanish.

Click to expand...

Hmh, sounds reasonable. I have a Salsa song lying around here and can't get it to work with English lyrics.... maybe it's time to learn some Spanish

Script to import lyrics and build notes

I've written up a script that might be of interest if you are as lazy as I am. It will read lyrics from a text file, place a bunch of corresponding notes on a track, then populate those notes with the lyrics. So it just gives a rough layout -- without needing to go in and count lyrics and place notes. The idea is you will then go in and move things around to make a song.

This is the first iteration. More info in the comments inside the file. Backup anything before using!

(Of course strip off the .txt before using.)

Updated 3/4/23. New version:

• gives greater control over the placement of notes
• changes out the random pitch variation for one that will run up and down a scale
• should support Mac and Linux

When I go to run script I get message input.txt not found. I’m on Mac (Ventura) and your details look like it’s for a pc system. I’ve tried putting the input.txt file in every place possible (all within /library/application support/dreamtonics/synthesizer v studio) with same results. Any suggestions what I may be missing?