Review: Dreamtonics Synthesizer V Studio Pro

The computers have been talking to us for a while now – think Siri and Alexa, Google Translate, the voice announcing the next train over the tannoy, or the infuriating gatekeeper robot you need to defeat in a game of riddles before you can talk to a real person after you’ve been on hold to your bank for an hour and a half. Thankfully, perhaps, none of these voices could sing to you. But singing computers have been around for decades – the Bell Labs brainiacs first made an IBM 704 computer serenade them in 1962. They walked so Yamaha’s Vocaloid could run: one of the earliest and best-known examples of a true singing voice synthesizer, Vocaloid was first released in the early 2000s and tailored for use in music production.

The workflow was famously laborious, and the results were, while impressive for the time, not particularly natural-sounding. Until relatively recently, this was about all you could expect from a vocal synthesizer – you’d get away with it for backing vocals, or perhaps even for a lead in certain EDM genres, but you’re not tricking anyone.

Read more gear reviews here.

This is no longer the case. The AI boom saw huge leaps forward in the realism of synthesized vocals; by feeding voice data to an AI engine, you could prompt it to perform lyrics in a given style. This gave rise to controversy – in April 2023, an AI Drake song featuring The Weeknd was streamed millions of times before labels issued a takedown. 

Dreamtonics

Synthesizer V, launched by Japanese company Dreamtonics in 2018, suffers from none of these ethical concerns whilst generating similarly realistic-sounding results. Built around a hybrid of “concatenative synthesis”, which essentially relates to synthesis based on huge sample libraries of recorded voice, and “a deep neural network-based synthesis engine”, which is way too far above my paygrade to understand, Synthesizer V runs entirely offline and impressively renders almost in real-time. 

Where other AI models have run into drama by training their voice engines on copyrighted media (like Drake’s back-catalogue), Dreamtonics take a more ethical approach by employing voice actors and singers to create different voice databases. The inventory of voice databases is constantly expanding; “characters” with peppy names like “Natalie”, “Ninezero”, and “Stardust Infinity” are developed by several third-party companies based all around the world, as well as Dreamtonics themselves. 

If you’ve spent any time working with MIDI or Melodyne in a DAW, you’ll be familiar with the basic layout of Synthesizer V. Upon opening a new session in the standalone app or a new instance of the plugin, you’re met with a DAW-like Arrangement panel, with tracks, a bar-based timeline with tempo mapping and looping functionality, and a basic mixer (volume, pan, solo, mute). You can add multiple voice tracks which use different voice databases, and you can also import audio onto “Instrumental Tracks”, which is handy for backing and reference material. 

Below this is a standard-looking Piano Roll panel, overlaid with lyrics and a familiar pitch curve/waveform bubble situation, which allows you to add MIDI, lyrics, and fine-tune pitch. At the bottom of the Piano Roll, hidden by default, is the Parameters subpanel, which functions like an automation lane and allows for tweaking of settings like pitch deviation, vibrato envelope, and even breathiness, the combination of which can produce some incredibly expressive results.

To make Synthesizer V sing to you, you’ll need to start by adding a voice track and assigning a voice database – voice tracks without an assigned database will play MIDI files as sine waves. Next, you’ll need to add some MIDI either by drawing MIDI in manually, importing a MIDI file, or recording it in with a MIDI keyboard. A handy feature is the ability to import audio and, by right-clicking and selecting “Extract Notes From Audio”, convert a monophonic melody into MIDI with startling accuracy.

The lyrics default to “la”, so unless you’re recreating the chorus of “Brown Eyed Girl” you’re going to want to input some new ones. The workflow is surprisingly intuitive: double click the first MIDI note, type a word or syllable, and hit the tab key to move to the next note. Almost instantly, you’ll notice the pitch curve and waveform update as the vocal engine works its magic. 

Having completed this process for the first time, I was pretty blown away by the results. Don’t get me wrong, you’re probably going to need to tweak things. Depending on the voice database you’re using, you might need to play around with different phonemes to get the pronunciation right, otherwise it’s going to sound like a talented session singer performing in a second or third language they’re a bit rusty in. A degree in linguistics definitely isn’t necessary but it’d almost certainly help!

You might also want to tweak the tonality – the options for this vary depending on the database you’re using, but you’ll encounter parameters with names like “Soulful”, “Warm”, “Bold” or “Soft” as well as “Breathiness” and “Tension”. Subtle tweaks can change the character of the voice dramatically, allowing a single voice like the Natalie database I demoed to go from punchy power-pop to soft, breathy acoustic ballad with remarkable ease. 

Another intriguing feature is the “AI Retakes” panel, which allows you to prompt the voice engine to record different takes with varying pitch curves and timbres, as if you were working with a session singer. This is can be a useful way to generate a few different options to get you closer to where you need to be if the first attempt doesn’t quite hit the mark.

The sheer scale of tweak-able settings defies inclusion in this review, but have a listen to this Adele cover to get an idea of just how far down the rabbit hole you can go:

I’ll admit to some scepticism going into this review. I’ve never used a vocal synth before – I barely use virtual instruments, and I don’t produce electronic music. Frankly, I’m astounded at how realistic Synthesizer V sounds – I’d almost feel comfortable using it to generate backing vocals for a project and telling a client I’d hired a session singer. For contemporary pop sounds, where the conventions around vocal sounds allow for heavier processing, I could absolutely imagine Synthesizer V taking centre stage.

Sure, it’s capable of sounding like someone’s put Siri through pitch-correction software, and it might feel like learning a whole new DAW before you really get the hang of things, but once you get used to the idiosyncrasies of the voices and develop a feel for what each is suited for, the results speak (sing) for themselves. The future is now, old-timer.

Keep reading, shop and buy Dreamtonics Synthesizer V Studio Pro here.