Synthesizer V Studio 2 Pro 2.3.0 Beta 1

[Dreamtonics]

Synthesizer V Studio 2 Pro 2.3.0 Beta 1

Dear Users,

We are excited to share the much-awaited beta for the upcoming free update to Synthesizer V Studio 2 Pro.

Since the announcement of the Synthesizer V Studio 2 Pro 2.3.0 Technical Preview, the Dreamtonics team has completely overhauled the phoneme inventory and its underlying system. This new beta presents a near-final version of the system, resolving several longstanding phonetic inconsistencies and simplifying spelling conventions so it is overall easier to learn and memorize.

Importantly, the updated system provides access to a unified set of 150 standard phonetic symbols (X-SAMPA) shared across all languages. This offers a much more precise way to fine-tune a note’s pronunciation without relying on a large number of AI retakes.

Version 2.3.0 Beta 1 also brings to Synthesizer V several new features first introduced to Instrument X, for example, the option to synchronize parameters with note movement, and the capability to delete/duplicate multiple tracks at once.

Download Links

Windows: https://drive.google.com/file/d/1Oksyhqyc9b5nZ-VJ7n6bEo_-G0OLW6jF/view?usp=sharing

macOS: https://drive.google.com/file/d/1UlnBr_lYp-9A4mM7a4HLG17wJc-otB_c/view?usp=sharing

Audio Quality Feedback

Following reports of potential audio quality issues in the 205b1 voice update, we are conducting a blind listening test comparing voice versions 201 and 205b1.

The test consists of a brief calibration followed by 15 paired comparisons. We recommend using headphones or monitor speakers in a quiet room. You may leave optional comments on specific words or overall differences, and you may stop at any time.

The test will remain open until October 14. The test has closed and we are now analyzing the results. Thank you for your participation.

Phoneme System Update

Prior to version 2.3.0b1, Synthesizer V used a mixture of language-specific phonetic alphabets (such as Arpabet for English) and standard phonetic symbols (X-SAMPA) modified with custom additions. Based on feedback from user surveys over the past few months, we have introduced simplified, intuitive phonemes for everyday use in each language alongside X-SAMPA symbols for advanced fine-tuning.

To summarize the changes, English largely retains the Arpabet symbols with an additional ll to distinguish syllable-ending “l” from onset “l” and four experimental phonemes for a British accent. Mandarin, Cantonese, Korean and other languages previously based on X-SAMPA now default to symbols that mimic the spelling/standard romanization. For example,

Format	Example
Mandarin
Orthography (Hanzi)	芳草碧连天
Romanization	fang cao bi lian tian
Pre-2.3.0b1 phonemes	f A N tsh AU p i l iE :n th iE :n
2.3.0b1 simplified phonemes	f aa ng c ao b i l iae -n t iae -n
2.3.0b1 X-SAMPA	f A N tsh au p i l iE :n th iE :n
French
Orthography	L’herbe tendre
2.3.0tp1 phonemes	l E R b t A_~ d R
2.3.0b1 simplified phonemes	l e- R b t an d R
2.3.0b1 X-SAMPA	l E R b t A~ d R

Japanese also retains most of the romaji symbols, with the addition of -i, -u, -N for non-syllabic vowels:

For detailed changes per language, please visit this page that lists reference tables mapping old symbols to new symbols.

Language-independent X-SAMPA Support

Version 2.3.0b1 introduces the ability to use a unified set of 150 standard phonetic symbols (X-SAMPA) across all languages. To enable this, turn on “Use X-SAMPA symbols” under the Language section in the Notes panel. This automatically converts existing phonemes into their X-SAMPA equivalents while maintaining the original pronunciation. From there, you can override phonemes using a much broader selection of international phonetic symbols.

A similar option is available in the Dictionary panel. You can now define custom dictionary entries using either the simplified language-specific format or the standard X-SAMPA format. Using X-SAMPA allows access to the full inventory of phonetic symbols, which is especially useful for simulating languages not natively supported by Synthesizer V.

This page lists all X-SAMPA symbols currently supported by the system.

New Vocal Fry Phoneme

The new phoneme system also disentangles vocal fry from the existing cl phoneme for glottal closure. It introduces a new vf phoneme to all languages.

vf can be inserted before or after a note, or between a vowel and a consonant, enabling different singing expressions.

(Inserting vf at the end of a note, like in this example, often leads to a pitch drop effect.)

Italian Cross-lingual Synthesis

Italian is now available as a cross-lingual synthesis language in the 205b1 voice update.

Because Italian spelling aligns closely with X-SAMPA, Italian phonemes default to standard phonetic symbols (X-SAMPA). Enabling “Use X-SAMPA phonemes” in the Notes panel grants access to an even broader set of phonetic variations.

Syllable Breaking Styles

Aside from phoneme-related changes, version 2.3.0b1 also enhances lyrics-to-phonemes conversion. For non-CJK languages (e.g., English, Spanish, …), a hyphen at the beginning or end of a note is accepted as a syllable breaking indicator, in addition to the existing “+” mark supported since the beginning of Synthesizer V.

Reporting Issues

Please submit bug reports through our support form. Please specify version “2.3.0b1” in your ticket:

https://dreamtonics.zendesk.com/hc/en-us/requests/new?ticket_form_id=44558189476505

While we welcome feedback on the forum, we may not be able to address the replies promptly. It’s also difficult to request necessary details (e.g. project files) for bug reproduction in a public forum. Therefore, we kindly request that you report issues via the support ticket system.

---

Synthesizer V Studio 2 Pro 2.3.0b1 (Oct 8th, 2026)

The following are changes since 2.3.0tp1.

New Features

• Overhauled the phoneme system: simplified/X-SAMPA switch, new vocal fry phoneme, and language-specific phoneme additions.
• Added cross-lingual synthesis support for Italian.
• Added a new settings option to synchronize parameters with note movement.
• Added key shortcuts (default: up/down arrow keys) to move selected notes up/down by 1 semitone.
• Added real-time phoneme spell checking to the phoneme input box above notes.
• Added ARA support for LUNA.
• Added to plugins a “Reset Window Size” command under View menu.
• Implemented a render cache rotation strategy and settings option to limit the disk usage.

Feature Enhancement

• Made the settings option to disable MIDI Input Device also available in the standalone.
• Allow multiple selection of tracks and editing track-level properties together.
• Optimized render cache to reduce the space consumption.
• Improved MusicXML support.
• Updated macOS app icon.

Behavior Changes

• Changed the key shortcut for navigating to the previous note to Shift + Tab on macOS.
• Added “*” as the default shortcut for starting a recording.

Bug Fixes

• Fixed a display-only issue showing interrupted pitch curves.
• Fixed audio not rendering when reopening an ARA project on some DAWs.
• Fixed a memory leak in ARA plugins on certain DAWs on macOS.
• Fixed data loss in macOS REAPER after setting up ARA and saving & reopening a project for the first time.
• Fixed Ctrl key unexpectedly triggering playback when using the ARA plugin on Windows.
• Fixed audio stutters when using the instrument plugin with a large buffer size.
• Fixed pitch control points/curves going beyond the piano roll’s vertical limits after note transposition.
• Fixed displaced notes after finishing MIDI recording inside some note groups.
• Fixed cropped Settings and Product Manager windows when running inside ProTools with high DPI settings enabled on Windows.
• Fixed the bottom part of the piano roll getting blocked by automation/phoneme panels.
• Fixed several UX issues with the preset selector.

[pearp]

honestly id prefer it if phenomes were just letters instead of symbols and letters since it’s hard to memorize it. but i love the changes!

[dbeaugra]

It seems that the Choir voices collection are not available in version 205b1 !!!
Is it planned soon ?
Otherwise are there any version of the choir that works with this beta 1 ?

[Pat]

You can use the version 204b choir voices with SV 2.3.0b1.

However, the X-SAMPA phoneme symbols are not yet supported by the version 204b voices.

[raidergale]

I’ve been testing Italian, and while it’s a promising start, it definitely needs some more finetuning.
For starters, since it only uses a single “r” phoneme for both tapped r and rolled r, it just defaults to rolling Rs everywhere, which is not accurate. It can be fixed by shortening the duration of the r phoneme, but it really should be handled automatically in the final version. For reference, generally (and I emphasize generally, there can be exceptions depending on accents and regions where you’re sampling the sounds) you get a rolled R only when you have a word with a double r, for example “porro” (leek), or when the r is preceding a consonant, for example “porto” (harbor). If you have a single r in the middle of a word, for example “pera” (pear), you get a sound that’s more similar to a tap. Same goes for a word starting with r, although this can vary a bit depending on the accent of the speaker.

Another weird thing I’ve noticed is that the current dictionary adds a lot of double consonant sounds, I’m not sure if this is meant to imitate a longer consonant sound or a glottal stop (but even then, the “cl” phoneme exists for glottal stops), but with the current implementation you can often hear the doubled consonant very clearly, which is extremely jarring.

[buckycore]

Oh thank God!!! THANK YOU THANK YOU THANK YOU!

[ZLY_CZLOWIEK]

Hey, raidergale! I love your YouTube covers.

[ZLY_CZLOWIEK]

Request: separate close mid and true mid vowels

e - Cantonese, German, Portuguese, Italian, French
e_o - Japanese, Mandarin, Korean, Spanish

o - Cantonese, German, Portuguese, Mandarin 中, Korean, Italian, French
o_o - Japanese, Mandarin 喔, Spanish

also separate Japanese dental n from Spanish alveolar n
and separate Voiced alveolar approximant (like in “red”) from Voiced non-rhotic alveolar approximant (like in 此) as these sounds do not sound at all similar.

these changes would greatly reduce foreign accent spillovers

and I feel like the lisp from 2.3.0tp1 worsened… except LIAM who was the best then is still the best now and perfected the german R sound. LIAM is still Trash on other uvulars like X

[ZLY_CZLOWIEK]

BUG found - dz phoneme doesn’t work in xsampa mode

[Mrhehon]

Couple of Questions
I gather this won’t work with my third party HXVOC and Solaria voices?

Does this overwrite the current version or can you run this beta along side?

Due to the HXVOC and Solaria I thinking best leave this alone.

[sparklep]

With the exception of the three choir packs, the only ones that got beta updates to 205b1 are all the Dreamtonics voices, including the most recent ones (Kohki, Nanami and Rochelle). And yeah it does overwrite the current version BUT you can always download the current stable version if you feel like this beta isn’t working for you

[noway]

French is still far from perfect.
New languages are missing in Voice-to-MIDI Conversion.

[tady159]

There aren’t enough letters in the Latin alphabet, even uppercase and lowercase combined, to cover the entire International Phonetic Alphabet. There are over 100 consonants and 28 primary vowel symbols.
X-SAMPA is one among a few approaches that map out most of the IPA using only ASCII characters. Digits and non-alphabetic characters (symbols) inevitably had to be used.

[C0rrupt3dSp1r1t]

The features for this update are excellent… unfortunately I’m sad to say it’s not pleasant at all to use. Voices have a scrapey metallic noise to them and their base timbres and accents are quite strange, especially when using XLS. We either need XLS tuned to be like SynthV Studio 1, or need to be able to control accent to sound more like their default or more native, v200 sounded too extreme in one direction (trying for heavy accents but getting numerous bugs as a result) and v201 and v204b sounded too extreme in the other. 205b1 is even worse in this regard. Vocals are a soup that blend into each other. Kevin is too smooth, raspy voices dropped some of it, Yi Xi is too youthful and nasal, Mai and Yuma are too deep, Feng Yi Soft sounds lore like Ling Wan than Ling Wan does. On top of the noise, it’s very unpleasant, and I can’t imagine how much worse this will make third-party vocals. Them not having betas is quite debilitating and makes updates hard to judge. Japanese third-parties especially suffered heavily during v201 with loss of numerous vocal traits. Autopitch is also nearly the same across vocals, even when using the modes on the expression panel that are meant to bring the differences out. Vocal mode issues seem to be reduced in v205, but they also seem to have weaker effects. The best I can say about v205 models is that the range prediction seems to be improved over v201, where it suffered serious bugs. I want to put this on a non-bug feedback form, but I can’t because it’s not out yet.

[IUnknown]

English /p/ should be in the aspirated section (ph)

I would argue that english /b/ should be in “voiceless bilabial plosive” rather than “voiced bilabial plosive”

also

please separate close-mid and true-mid /e/ and /o/.

and for any Finnish or Castilian Spanish speakers here, the Korean “s\_t” is a really good approximation for the retracted /s/

[Stanlee]

I don’t know about the sv2 latest update. Don’t wanna put it on because the voice update that I did install seems to have made all tested voices sound worse than ever, with weaker sound and more noise. Disappointed. It may be my imagination but it seems like companies are breaking their products around the board.

[buckycore]

What do you mean by weaker and more noise? And what voices have you noticed this on. I’m just curious…I didn’t notice any change.

[Stanlee]

With Marvin and Rochelle also when using choir 1 with 1 singer there is a mid frequency shush type sound blended with voice that almost no other voices ( choirs 2 and 3, liam etc. )doesn’t have present. Also certain words seem lower in dynamics than previous to 2.5b

[cschormann]

First of all, thank you to Dreamtonics for all the new language support. This is very exciting.

I did my first experiments using German. I decided to just try out the words in the phoneme guide table here:

Phoneme changes

My first impression is that this does definitively not yet sound like a native German speaker at all - it sounds (at least with all the voices I have tried so far) like a foreigner singing German. Sounds like the “ch” in “ich” sound very much non-native, many of the vowels have the wrong tone, and the automatic spacing of some consonants is not what a native speaker would do.

What can we (the community of native speakers) do to help to improve this as we move towards the full release? Is it even a realistic expectation that the existing voices will be able to master a native-like pronunciation, or will we have to wait for voices from native speakers?
In the past, I was quite impressed how well cross-language use worked, at least with English to Spanish, where I have limited ability to judge (and access to native speakers for some deeper checks) - but as far as I can tell, for German we are much further away from that level of native sound.

Again - what can we do to help?

And many thanks again to Dreamtonics for embarking on this road. Very exciting!

[ZLY_CZLOWIEK]

@IUnknown

and for any Finnish or Castilian Spanish speakers here, the Korean “s_t” is a really good approximation for the retracted /s/

Very useful tip! ᅟ⠀⠀⠀⠀