Speculative Thoughts on the Possible Evolution of the Engine (2026–2027)

Synthesizer V Studio
1

:warning: Disclaimer
This post is pure speculation.
It is not based on any insider information, leaks, or official statements from Dreamtonics.
Everything below is an inference drawn from current Synthesizer V features, especially the evolution of the engine since V2 and the introduction of Choir 1, 2, and 3.
The goal is to spark discussion, not to predict or suggest an official roadmap.

1. Established starting point: where Synthesizer V already is

It is important to start from a realistic assessment.

:backhand_index_pointing_right: Synthesizer V already operates with a global, contextual interpretation model — even for a solo voice.

Since V2:

  • a vocal line is no longer treated as a sequence of independent notes

  • expressive states propagate across phrases

  • interpretation depends on:

    • melodic direction

    • rhythmic context

    • continuous expressive continuity

In other words, even a single voice:

is already interpreted as part of a coherent musical system, not as an isolated signal.

2. Choir 1, 2, and 3: not effects, but windows into the engine

The Choir modes are often discussed as “chorus tools”, but they reveal deeper properties of the engine.

  • Choir 1: controlled temporal desynchronization

  • Choir 2: embodied timbral variation

  • Choir 3: correlated ensemble organization

In all three cases:

  • the user fully controls all parameters

  • nothing is imposed or opaque

  • but parameters are interpreted as global ensemble states, not as independent offsets per voice

Choir 3 does not remove control —
it demonstrates that the engine can handle correlated collective behavior in a deterministic, musical way.

3. A simple technical observation: voice is the hardest case

From a synthesis and interpretation standpoint, a singing voice combines:

  • continuous pitch

  • controlled micro-instability

  • non-linear transitions

  • dynamically evolving timbre

  • articulation driven by language

This is at least as complex — and often more so — than:

  • a solo violin

  • a saxophone

  • a trumpet

  • any expressive monophonic instrument

Yet Synthesizer V already handles:

  • expressive monophonic lines

  • long-range phrase coherence

  • ensembles of correlated lines (Choir)

This leads to a reasonable hypothesis.

4. Central hypothesis (speculative, but consistent)

Synthesizer V is already, in principle, a generic expressive musical synthesis engine — with voice being its most demanding application, already solved.

If we abstract away lyrics:

  • a solo voice

  • a monophonic instrument

  • or a group of monophonic instruments playing in unison

share the same musical foundations:

  • phrasing

  • human timing

  • attack behavior

  • controlled stability / instability

  • expressive coherence over time

The remaining differences are largely:

  • timbral models

  • physical constraints

  • not the core interpretation logic.

5. Key projection: interpretation guided by musical direction

A likely next conceptual step is not prompt-based generation, but interpretation guided by intent, similar to:

  • a choir conductor

  • an orchestra conductor

Plausible evolution (2026–2027):

Synthesizer V could allow users to direct interpretation, not just edit parameters.

For example:

  • interpretive directives such as:

    • restrained / assertive

    • tense / relaxed

    • fragile / massive

  • applied to:

    • a phrase

    • a section

    • an ensemble

The engine would translate these directives into:

  • timing behavior

  • stability

  • attack shaping

  • ensemble cohesion

Crucially:

  • everything would remain editable

  • nothing would be imposed

  • directives would guide parameter interpretation, not replace control

This is a change in musical perspective, not a loss of precision.

6. Natural implication: extension beyond voice

Within this framework, extending beyond voice becomes conceptually straightforward.

Not as:

  • a full orchestral library

  • a massive instrument catalog

  • an automated scoring tool

But as:

  • a single expressive monophonic instrument

  • driven by the same interpretive engine

  • controlled by the same musical-direction logic

Later on:

  • small instrumental ensembles

  • operating under the same principles already demonstrated by Choir 3

7. Plausible directions for 2026–2027 (non-predictive)

Summarizing possible, non-exclusive evolutions:

  • continued internal generalization of the interpretation engine

  • increasing importance of high-level interpretive direction

  • a symbolic, carefully scoped non-vocal extension

  • deeper work on co-interpretation (voice + other sources)

All without breaking:

  • editability

  • determinism

  • the current Synthesizer V philosophy

8. Open conclusion

This post is not about stating what Dreamtonics will do,
but about asking a broader question to the community:

Is Synthesizer V ultimately meant to remain an exceptional vocal synthesizer —
or to fully embrace its identity as an expressive musical interpretation engine, with voice being its most complex case rather than its sole domain?

Given:

  • Choir 3,

  • the current maturity of the engine,

  • and the already-present global interpretation model,

this feels like a reasonable topic for discussion.

I’m very interested in hearing other perspectives, agreements, or disagreements.

4 Likes
2

I hope it remains solely as a vocal tool. Otherwise we will see…

  • neglect of development of the vocal engine
  • bloat
  • breaking of things

I think SV has a long way to go to realize its full potential as a vocal engine, and I would hate it to go down the ACE route.

Dreamtonics have shown in the past that they have a lack of understanding with some musical contexts, and with recording technology. I think its a bigger leap than it seems for them to expand to other instruments, especially when considering genres beyond pop.

No problem with Dreamtonics creating a new product for instrumental work if they have the skills, desire and staff but I hope they retain SV as a vocal only engine and that this remains a primary focus for them.

4 Likes
3

I understand the desire to keep SynthV focused on vocals — that focus is part of why it works so well today.
That said, I think it’s worth separating two questions: what the product is positioned as, and what the engine is actually capable of.

From a technical standpoint, SynthV already solves problems that go well beyond “singing voices”:

  • continuous pitch control

  • phrase‑level interpretation

  • controlled instability over time

  • correlated behavior across multiple emitters (Choir)

Those are not voice‑specific problems. They are general expressive monophonic interpretation problems. In fact, adding lyrics and phoneme articulation arguably makes voice the hardest possible case.

Once you strip text away, many instruments (violin, sax, trumpet, solo winds) are simpler from an engine perspective than a sung voice. The core behaviors SynthV already models — phrasing, timing, attack shaping, ensemble cohesion — apply directly.

So when people talk about SynthV “doing more than vocals”, I don’t see that as a radical shift in direction.
I see it as a question of whether the same interpretation engine remains voice‑exclusive by design, or whether its generality is acknowledged.

There’s also a market reality here that’s hard to ignore. Tools that present themselves as unified expressive engines (voice + instruments) are starting to define expectations, even if their depth is lower. If SynthV stays framed purely as a vocal island, it risks being seen as excellent — but narrow — while other platforms own the “full expressive workflow” narrative.

That doesn’t mean SynthV should become a catch‑all tool, or chase scope at the expense of quality.
But it does mean that, long term, staying vocal‑only is not just a technical choice — it’s a strategic one, with real risks attached.

Personally, I think the safest path would be:

  • keep SynthV deeply vocal‑focused

  • avoid bloating it

  • but not pretend the engine itself is limited to voice, when clearly it isn’t

Whether that leads to a separate product, optional extensions, or something else is an open question.
But as the market moves, doing nothing also has consequences.

4 Likes
4

I personally would like synth V to stay focused on vocals for now.

For one, there is still is a lot to do. For example

  • In terms of expressing vocal fry and growls are still beyond the engine or the available voices
  • There are still cases where the engine produces very unnatural connections between phonemes
  • I would like it to be able to whisper and speak as part of vocal performances
  • I would like to see more languages, such as German and Portuguese
  • I’d love to see deeper integration with the daws.
  • Synth V still struggles with non-step tempo changes
  • I would love to be able to have more control over viele pronunciation

Secondly, I don’t want Synth V to try to replace my DAW. I think that the current division of labor between daws and instrument plugins is a good one in principle. And that does not necessarily limit the business opportunity for Dreamtonics, I think - I spent more money last year on synth V add-ons than I spent with my DAW vendor.

Third, I think that the sort of high-level ai interpretation you advocate for is not something I personally find very desirable - too uncomfortably close to generative AI. As a producer and composer, I want expressively to be mine :slight_smile:

5 Likes
5

I’m very much looking forward to the addition of new languages and the improvement of XLS.

In general, the developers would need many voice providers to achieve that, but I feel like someone who’s knowledgeable about phonetics and phonology can be particularly helpful. Like, a linguistic generalist.

1 Like
6

That, right there, would be my number one request

3 Likes
7

This is a bit old, but I did some whisper vocals on this one - https://youtu.be/TZ1dM52wS08?si=y3qEORX3xJIqxYq_.

3 Likes
8

That whisper is really good, in fact all five parts of the EP are well worth a listen :headphone:
I will continue to try to get something like a ‘talking’ voice but the ‘musicality’ of speech is so different to singing it really does push the parameters avaiable in SynthV.

1 Like
9

They’re really not. Instruments have a lot of different articulations that must be controllable on a per note basis. Using the instrumental version of vocal modes for these would be massively clumsy, and what other way would there be, using something close to the current engine, to realize these articulations? Interpretation of intention is something that requires interaction, not automation. Take a horn section - imagine the same line played by an orchestral brass section, then by a modern big band, and then by a rock band. The way the same line would be played, is very different, even if the notes on the page are the same. So then maybe you have genre “vocal modes”… jazz could go from Dixieland through to Bop and beyond… That would create some interesting results, but it still wouldn’t put the control in the hands of the composer, the way you’d want it to. Okay, so lets forget about ensembles… how about it just applies to solo players… well, same thing goes. Notes on the page are only half the story, and even within a single sub-genre, there would be different ways to interpret a phrase depending, somewhat on context, but moreso on history. So how far do they go with sample data for modeling? With samples (and physically modeled instruments) the user decides, per note, what the articulation is going to be. The exact value of the note played. That is not how Synth-V is set up at all.

Then there’s the sound. Synth-V is largely physically modelled. Seems like there may be some samples in there (maybe consonants?), but the bulk of the sound is modeled. There are physically modeled instruments too - Audio Modeling’s SWAM for example. Not only expensive per instrument, but also, constantly criticized for its sound. We may be getting better, and a merger of Audio Modeling with Dreamtonics might yield some interesting results, but largely, the expertise to create instruments is specialized. Can you hand that off to a LLM? Maybe. You still need to record the instruments being interpreted… that in itself requires expertise that I don’t think DT has. If you watch their video on the making of the choirs, it is evident that their expertise in recording is very limited. If you think about the different ways instruments are recorded - even the same instrument, in different genres… they’re certainly not ready to take that on.

All of the issues can be overcome in the long term, and with enough money, but I think we’re a long way from it at this point. But who knows… maybe this is already in the pipeline. I personally hope not, but if it is, good for DT.

1 Like
10

New user - just wanted to chime in with agreement to a couple comments/requests from others:

  • spoken/whispered. I wanted a “shout” or “chant”… seems a common interest. (instead, I had the “rest of the band” sing the lines a little sloppy (with Vocoflex) and loved the sound so much, I dropped the shouting bit… But still - I’m a rock and roll guy, and need a shout or scream now and then.

  • Timing. Oof. What a hassle I had with a song that changes tempo. PLEASE get DAW integration further developed, including this aspect.

  • More options in voices. Again - I’m a rock and roll guy mostly, and there’s not a lot to choose from. I’ll keep fiddling with Vocoflex to see if I can “make one” - but I don’t have samples to start with, so it’s not an easy answer.

3 Likes
11

Well, I think we now have the answer to our questions…

Instrument X, released yesterday (Instrument X - Neural Acoustics Modeling Plugin), is very close to what I was describing back in April, and it makes perfect sense as the next step in Dreamtonics’ development roadmap. In doing so, they have effectively stolen a march on the ACE STUDIO / EASTWEST partnership, which has also been working on this very topic.

As expected, samples are being replaced. An automatic performance interpretation is provided (which is still better than the default playback offered by sample-based VSTs, isn’t it?), yet users retain virtually unlimited possibilities to shape and control every aspect of the performance with the highest level of precision.

It may be time to start writing a REQUIEM.

A Requiem for sample-based VSTs.

The computer buried the typewriter. The same thing will happen here.

3 Likes