Dreamtonics Synthesizer V Studio 2 Pro

Synth V’s UI has been reworked in a number of ways in v2, to reflect some significant refinements in the software’s toolset.

Dreamtonics take their session singer in software to the next level.

I love it when a music technology product leaves me believing in magic. Working with Dreamtonics’ Synthesizer V — which I reviewed in the March 2023 issue of SOS — was undoubtedly one such moment. Dreamtonics have now released version 2 of Synth V. This brings a whole raft of enhancements and new features that, on paper at least, seem very intriguing. To be on the safe side, I placed a pillow on the desk under my chin before exploring. Did I need it? Let’s find out...

Sing It Again

I covered the basic features of Synth V in the original review so, aside from restating that v1 was already capable of generating some amazingly realistic sung vocals (realistic to the point that many listeners will not realise that they are hearing a synthesized voice until you tell them), I’ll focus on what’s new here. If you need to catch up, then the original review (SOS March 2023) is available to read in full on the SOS website and the underlying principles still apply even if some operational changes have been made.

That said, I should also point out some highlight additions that arrived as updates to v1 of the software after that original review: a rap engine designed specifically to create rap vocals, and a voice‑to‑MIDI conversion option allowing you to automatically extract note/lyrical information from your own recorded vocals and turn them into the data need to generate a Synth V vocal. Spanish language support was also added. And, while running as a standalone application (which is how I did much of my own testing here), Synth V is also available as an ARA plug‑in within suitable hosts.

For those new to the product, it’s also worth emphasising one key concept; Synth V is a virtual instrument. You create the melodic (and lyrical) content of the performance and Synth V’s engine then renders that information into a synthesized vocal based upon your selected virtual singer. The engine utilises AI/machine learning to process detailed samples of the actual singer from which the selected voice library was built and to resynthesize those samples into the performance you have constructed. This includes the pitch, lyrical content, the choice of singing ‘modes’ (styles such as clear, muted, belt), characteristics (such as loudness, tension and, breathiness) and phoneme duration and pronunciation. Make an edit and Synth V resynthesizes — not in real time, but very rapidly — the required changes.

The principle is, therefore, no different from the way you might create a virtual solo violin or guitar part; a sample/synthesis engine interprets the MIDI and sound parameter information you have supplied into a realistic performance. It’s the same here but with the human voice. Dreamtonics now explain very clearly on their website how the process works and how they have collaborated with (and compensated) the original human singers upon which each Synth V voice library is based. This is not AI vocals generated from a text prompt.

New For Two

So, what’s new in version 2 of Synth V? There are some obvious headlines. On a technical front, the rendering of edits (to notes, pitch curve, voice parameters, lyrics, etc) is now considerably faster. I did much of my testing on an M1‑based Mac and re‑rendering after any edit was pretty much instant. It’s a very slick experience.

The overall UI remains familiar from v1. The track‑based Arrangement overview dominates the top (where you can house both audio tracks and Synth V vocal tracks), and the detailed view of the currently selected vocal track sits underneath. The latter is, itself, split into two zones: the piano roll, where you can edit note, pitch curve and lyrical content; and then optional lanes for automating the various vocal modes and parameters. In addition, a series of six panels — Voice, Notes, Dictionary, Render, License and Settings — can be toggled open/closed as required to access further parameters and settings.

However, digging in to all these elements reveals numerous changes, additions and enhancements. For example, in the Notes panel, the synthesis engine options have now been consolidated into two main modes of operation — Sing and Rap — and this panel now includes the Retakes options that used to have their own panel. The Retakes feature lets the engine generate alternative versions (takes) of either a whole vocal or just selected notes. This has been rather neatly reworked with new options for just varying timing, pitch, timbre, or all three together. There is also a new X/Y panel allowing you to quickly vary the ‘expressiveness’ used by the engine and the amount of vibrato modulation applied to longer notes. This is all very easy to use and you can swiftly switch between your various virtual takes to see which you prefer for any given phrase.

Detailed pitch editing now includes the ability to define nodes on the pitch curve that can guide the synthesis engine.

The Voice panel, which contains parameters for your selected voice library, has also been refined. For example, while the Parameters section is as before, the Vocal Mode section — where you can adjust the level of the virtual vocalist’s various singing styles — now allows each mode to be popped open to provide individual control over the Pitch, Timbre and Pronun(ciation) associated with each mode. This provides you with a much greater degree of control over the character of the voice. This panel also used to house controls to vary vibrato and pitch slides into and out of notes, but this functionality is now replaced by the additions to the Notes panel and by additional options to edit the detailed pitch curve in the piano roll and automation lanes (more on this in a minute).

Changing Lanes

In operational terms, v2 also brings some key changes in the editing and automation panels. Editing notes and lyrics remains very much as before. However, you now also get a Phoneme lane located directly beneath the Piano Roll editing panel. This allows you to adjust both the strength and duration of individual phonemes in an entirely graphical environment. It’s a much easier process than previously provided within the Note panel and allows you to finesse the pronunciation of words in a much more intuitive fashion. Detailed editing of the pitch curve was possible in v1 but the process has been significantly refined in v2 by the ability to add (and edit) Pitch Control Points directly on the curve. You can also draw pitch curve sections in freehand. The way both of these elements are then used is also very clever as, once a point of manual curve segment is placed, the engine will then ensure that the synthesized pitch curve does its best to pass through these points/curves while still generating a natural result. This system provides powerful control over pitch transitions between notes or vibrato in sustained notes, for a much‑improved workflow over v1.

The new lane for phoneme editing makes it much easier to edit emphasis and timing of individual phonemes, letting you control pronunciation and, usefully, snag sibilants before they might hit a de‑esser.

Practice Makes Perfect

The toolset changes do mean that existing users will have to make some adjustments to their established workflows, but Dreamtonics have undoubtedly streamlined and improved the overall process with these changes. Editing is much faster, and the toolset feels much more refined. The (very welcome) improved rendering speeds contribute to the enhanced workflow.

Of course, just as (or even more?) important is whether the synthesis engine delivers improvements in the actual vocals created. The short answer is yes; once you have familiarised yourself with the changes to the toolset, the results can be incredibly realistic. I got to test a number of the updated native English voicebanks: Dreamtonics’ Kevin, Natalie, Hayden and the new Liam, for example, and Eclipsed Sounds’ (one of the main third‑party companies creating voicebanks for Synthesizer V) Solaria, Saros and Astarian. As in v1, Solaria was particularly impressive but the additional details within each voice’s Vocal Modes really do let you conjure more ‘human’ in the resulting vocals, providing you are prepared to automate the available parameters. And, if you need even more voice options, Dreamtonics’ Vocoflex (see the December 2024 review) provides plenty further potential; it works just as well with Synth V vocals as it does with those of human singers.

If you want to hear some examples there are a few short clips I created during the review period at the start of this review. The most obvious use cases for this virtual instrument is for songwriters and/or music producers developing song ideas ‘in the box’. However, whether you eventually use Synth V’s vocals as a guide to a real singer, or they actually make it to your finished mix, for writing and editing vocal parts, Dreamtonics have created a brilliant creative tool.

Synthesizer V Studio 2 Pro is amazing technology with a surprisingly modest price tag.

I’ll leave a discussion about the desirability of synthesized lead vocals for another day (but it’s undoubtedly a very interesting debate). Did I need that cushion under my chin? Well, Synthesizer V Studio 2 Pro improves on what was already an impressive product. With a little practice, it’s perfectly possible to generate lead vocals that will jump well over the quality threshold required for a number of very practical use‑case scenarios. That in itself is a remarkable thing to be able to say. This software makes me smile every time I use it. Synthesizer V Studio 2 Pro is amazing technology with a surprisingly modest price tag.