Synthesizer V Studio 2 Pro 2.3.0 Technical Preview

Announcements
1

Dear Users,

We are excited to announce a technical preview for an upcoming update to Synthesizer V Studio 2 Pro, featuring several experimental changes. We welcome early feedback from the user community to help refine these updates.

About This Technical Preview

This release differs from our standard beta program. We are currently evaluating several major changes in parallel, which have not yet reached the feature maturity typically required for a beta. However, given the scope of these updates, early feedback is highly beneficial. This allows us to validate our direction and make informed decisions before committing further to specific development paths.

Download Links

Windows: https://drive.google.com/file/d/1aWf4_5jNEcqOw6CCsqBcA7hiDb-__Wk7/view?usp=drive_link

macOS: https://drive.google.com/file/d/1Y3BApiCuvH7NoeiYv4xsTkc7lTCv5XTh/view?usp=drive_link

To access beta voice updates, open Product Manager from within Synthesizer V Studio 2 Pro and enable “Beta Versions”.

Details

image
image2746Ă—854 150 KB

Highlights of the Technical Preview

Rendering Speed Improvements

Version 2.3.0 Technical Preview introduces a major backend refactoring designed to reduce multithreaded rendering overhead on modern CPUs. This addresses performance issues on big-little core architectures observed in previous versions. On processors with four or more cores, rendering speed has improved by 30%–50% compared to the previous version. Additionally, projects with multiple voices now load significantly faster, as voice loading occurs in the background.

Windows on ARM Support

Synthesizer V Studio now includes native support for Windows on ARM. Note that plugin availability may depend on your host application’s ARM support.

French, German and Portuguese Support

For the first time we present French, German and Portuguese cross-lingual synthesis support for Synthesizer V. The additional language support is enabled by updating voices to version 204b1 and updates are currently available for all version 2 Dreamtonics voices. Please note:

  • For Portuguese, we currently target a relatively neutral Brazilian Portuguese accent.

  • We are aware that the current phoneme set and notation system may not be optimal. They are subject to ongoing refinement and we do not guarantee phoneme compatibility between this technical preview and the final release.

Predictable and Reproducible Rendering

The voice update also resolves a long-standing rendering stability issue. Previously, minor changes to note positions could cause significant variations in pitch and timbre under certain conditions. This typically manifested as sudden volume dips across a subset of notes in a long sequence, or inconsistent vocal quality depending on the rendering start position. After carefully analyzing error propagation within the voice synthesis model, we implemented targeted design adjustments to ensure predictable and reproducible rendering results.

Korean Improvements

Korean word-to-phoneme conversion has been improved based on feedback from native speakers. These changes apply to existing voices and do not require a voice update.

Reporting Issues

Please submit bug reports through our support form. Please specify version “2.3.0tp1” 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 2.3.0tp1

New Features

  • Added experimental cross-lingual synthesis support for French, German and Brazilian Portuguese (v204 voice update required).
  • Added native Windows on ARM support.

Behavior Changes

  • Reset existing pitch and parameters when applying Voice-to-MIDI conversion to an ARA-mapped group.
  • Create the new track under the instrumental track when applying Voice-to-MIDI to an instrumental.
  • Merged “Import Audio…” and “Import Project…” menu items.
  • Renamed “Misc” menu to “Help” to enable native macOS command searching feature.
  • Save loop status in the project files instead of the editor’s settings file.
  • Group duplication by holding Alt/Option and drag now follows the default group pasting behavior (linked vs unlinked) settings.
  • Hide tools and mode options when the piano roll is collapsed.
  • When inserting lyrics, treat consecutive spaces as a single note separator.
  • When converting a version 1 project, split pitch curves at note boundaries.
  • Default to saving the project inside the host when importing a project into the instrument plugin.
  • Ignore file extension casing when importing files.

Feature Enhancements

  • Improved Korean grapheme-to-phoneme conversion.
  • Improved multithread rendering efficiency.
  • Reduced project load time by loading voices on a separate thread.
  • Allow project data to be copy-and-pasted across standalone and plugin, or across separate plugin instances.
  • Improved MusicXML import:
    • Added support for repeats.
    • Added support for mid-measure tempo changes.
    • Fill notes with the default lyrics if the score contains no lyrics.
  • Added menu commands and key shortcuts for toggling loop mode and metronome.
  • Automatically extend a group during MIDI recording.
  • Allow individual automation points to be deselected with Ctrl/Command + Click.
  • Show files overwrite warning when exporting audio files to the same destination.
  • Improved the error display in the case of file/voice loading failure.

Bug Fixes

  • Fixed position-dependent/non-deterministic rendering results (v204 voice update required).
  • Fixed “Prevent Group Overlap During Dragging” settings option affecting note creation.
  • Fixed plugin loading failure due to Garageband’s app sandboxing.
  • Fixed project not rendering after reopening a project based on the ARA plugin.
  • Fixed piano roll keyboard’s preview tone being routed to the wrong plugin output channel.
  • Fixed crashes when closing the program during playback.
  • Fixed crashes when executing “Remove Short Silences”.
  • Fixed crashes after removing the ARA plugin from REAPER for the second time.
  • Fixed crashes when deleting the group/track/project during MIDI recording.
  • Fixed ghosted dialogs after removing the plugin.
  • Fixed occasional unsynchronized playback when using the instrument plugin in Cakewalk.
  • Fixed occasional unsynchronized playback between synced tracks and local tracks when using the ARA plugin in Cubase.
  • Fixed pitch control points/curves deviating from the intended positions when the note is clipped by the group’s boundary.
  • Fixed an audio group affecting the length of a note group from a different track when dragging it in the Arrangement view.
  • Fixed dialogs and voice selector not conforming to the HiDPI scaling on Windows.
  • Fixed menu commands getting stuck in a disabled state after canceling a dialog on macOS.
  • Fixed several feature regressions introduced in version 2.2.1:
    • Fixed broken compatibility with certain audio devices.
    • Fixed crashes when reopening certain Studio One projects using the ARA plugin.
    • Fixed phoneme conversion regressions when encountering certain Korean/accented latin characters.
  • Fixed voice preset under Voice panel not resetting Unison parameters.
  • Fixed changes to application settings not being saved when multiple plugin instances are open.
7 Likes
2

is italian on the roadmap for the future!?!?!? :slight_smile:

4 Likes
3

The Korean feedback I’ve been waiting for has been reflected. I’ll quickly test it out. The added language is interesting, too!

1 Like
4

How and where to get v204 voices?

Edit: Sorry, I’ve found it.

1 Like
5

French is far from perfect… It sound more like an English person trying to sing French with a strong English accent. Word endings, “r” consonants, “an” nasal vowels… Nothing in this update sounds like authentic French language, sorry.

1 Like
6

I already have the Dreamtonics voices working, but I guess Eclipsed Sounds ones aren’t available yet, so I’ll be eagerly waiting. I have a project with Asterian in the works and I want to try new languages ​​- mainly German, to see if it will improve my Czech. Of course, I can try it for now, for example, through Liam, but I have to transpose the melody higher, because Asterian can handle the contra-octave (from F1 to C2) without any problems… I use those notes quite often in songs.

1 Like
7

the new languages do not yet appear in the phoneme reference. Until this is done, it’s not likely to be much use as we can’t do corrections, other than through a lot of trial and error. But I’m really excited about having German when it reaches the useable stage! In the meantime, i’l probably just stick to Spanish for my German.

8

send to dreamtonic

As a french speaker, the french crosslingual is far from perfect

the “O_~” often phonemes sound like an “A_~” depending the combination (often on a single note) and the “E_~” sound like an weak “eh” or “aa” sound

the consonant “s” is very sibilant almost harsh sounding on some voicebank (especially kevin), the “R” sound like an “hh”

and the french phoneme conversion often ignore liaison

for example, the “t” on vingt-ans is ignored, making it

Capture d'Ă©cran 2026-05-22 204145
Capture d'Ă©cran 2026-05-22 204145351Ă—239 8.83 KB

sound like “vin-an”

Uploaded and audio sample about the wrong pronouciation https://drive.google.com/file/d/13EAfhWMEGQFRjJjFRrrqwX38z2csCCjP/view?usp=drive_link

1 Like
9

True, that R is awful. Then you have all the nasty final letters (that are not always pronounced), words beginning with an H (you sometimes pronounce the final consonant of the preceding word, but sometimes you don’t)… All those exceptions and particular cases are real nightmares in French… It would require the Dreamtonic team some time to deal with exceptions and particular cases. And it’s certainly the same for other languages, as each one has its specifications and characteristics. As far as I’m concerned, I’ll happily volunteer if they need help, and I’ll gladly wait for them to spend more time (and take even a whole month or two if needed) and have something way more convincing !

1 Like
10

I just tried a new version of my latest pop song, this time in French.
It’s far from perfect, but it’s a first attempt.
I hope that with the final version and a few improvements, it will be interesting.
Pop song in french

1 Like
11

It’s fine listening work (it has a kick) and it sounds like it’s for a Czech who only knows French by listening without understanding, so he definitely likes it…

12

I’m not sure, which user or music are you talking about?

13

I downloaded the Mac OS version of the update, but now the software crashes whenever I try to load a vocalist pls fix this I’m dying to try French…

1 Like
14

Have you also updated the voice banks?

15

I’m getting DAW crashes almost every time I do basic stuff in the VST3 plugin. Live recording, for instance, either stops and crashes halfway, or crashes after recording.

1 Like
16

I updated Liam and Felicia but nothing changed, I’ll try to reinstall the software update file hoping it was just a minor bug

17

There’s another inconsistency in French. Some two-syllable verbs (or nouns) are pronounced with two syllables in the plural form and only one syllable in the singular form (i.e. in “ils chantent”, the verb will be sung “chan-tent” vs “il chante” it will be sung on one syllable), even if you want to force it with an adjacent note with the phoneme “+”.
And overall, each voice has to sound more French - not French with an English accent. Vowel phonemes in French are more straight, not as “relaxed” as in English.
On further testing, I think the “R” consonants are not that catastrophic, but they’re still not real French “R”s. They sound more like an imitation, like a raw “hhhh” that resembles an authentic French “R”.

Another problem (and another typically French nightmare) is the accented “E”, such as “é” and “è”. A word like “les” (plural “the”) is usually pronounced “lé”, but it’s pronounced like “lè” in SynthV, which is the way some people pronounce it (usually pronounciations differ according to regions where people live).

Speaking of “les”, I’m trying other words to check “linkings” between words (like in the example given by @FLMMDShare). If you try for instance “les uns et les autres”, SynthV will link words correctly (“les zuns et les zautres”), so it seems like they did program it correctly for some words.

1 Like
18

I’m talking about myself, who knows about ten French words (+ internationally known words) and I don’t care when listening to your song whether the singing is in perfect French. When I work on my songs in Czech, no one here has criticized the Czech sung by AI voices. Because no one who knows Czech at the level of a native speaker has heard it. However, I have native speakers around me and I prefer not to let them do it listen too much. On one song, I spend 2/3 of the time on Czech, and that’s out of the entire time it takes to create it. Your song is easy and pleasant to listen to.

1 Like
19

Many thanks! This wasn’t a song intended for the showroom. Just a simple translation of my English version to see how it turned out.
But I was just wondering if there was another message regarding your comments that I might have missed.

1 Like
20

I don’t know which voices you’ve tested but the majority of English speakers speak much worse French than what I’ve heard from the song posted and my own (admittedly limited so far) tests. Although there are certainly problems – including the ones described – I’m reasonably happy with the initial attempt at French to be honest. German is arguably worse, though even here for new text, it will certainly save time once we have the phonemes dictionaries available to make corrections as required.