Dreamtonics Vocoflex

The main Vocoflex screen.

Vocoflex imposes the timbre of one or more reference voices onto another — so just how convincing are the results?

We’ve had the ability to manipulate the pitch and formants of vocal parts for decades now, but while these processors have their applications I’ve always been a little disappointed when trying to obtain really natural‑sounding results. Supporting all the mainstream plug‑in formats for Mac and Windows hosts, Dreamtonics’ new Vocoflex, which also works standalone, tackles the same problem from a different angle, and I’ve been genuinely impressed by what it can do.

Basic Operation

Essentially, Vocoflex applies the characteristics of one or more separate voice recordings to another voice recording: it can load multiple ‘reference’ voice samples and then morph between their characteristics, before applying them to the input signal. So there’s scope for creating a wide range of vocal timbres.

Reference voice samples in WAV format, which need only be 10 to 20 seconds long (but should be solo voices, free from processing such as reverb or delay), can be dragged and dropped directly onto the Vocoflex GUI, at which point they are analysed and displayed as a series of points on a curve. Vocoflex breaks these segments down into shorter segments, and analyses the formant profile for each one (because vocal formants can change during a performance); the points relate to how the sample has been broken into segments. Higher‑pitched or female‑sounding voices are generally placed to the right of the screen and lower‑pitched, male‑sounding voices are arranged on the left.

For the process to produce a natural sound, the reference voice should, ideally, be in the same pitch range as the vocal you’re working on, but where there’s a discrepancy, clicking a tuning fork icon opens a small display next to the displayed curve and a slider below the tuning fork. Pitch‑shifting can then be used to match the input voice as closely as possible to the reference voice. This step is particularly important when changing a male voice to a female one with a naturally different register, or vice versa.

As you play the voice track, it’s possible to influence the result by moving a cursor on screen. This is joined by thin lines to the points on the reference voice curve and as you move it the contribution of the different points is changed. The differences can be quite subtle, as the reference formants may not change much from point to point, but things change completely when you drag in another reference voice. If you then place the cursor between the two reference curves, the two voice characters are morphed and the blend changes as you move the cursor. If you drag the cursor behind one of the curves, the display shows a subtle shadow being cast on the other curve and in that scenario, Vocoflex shapes the sound in the opposite direction to the shadowed curve — effectively the character of the shadowed curve is subtracted from the composite timbre. While that might not quite sound entirely intuitive, in practice it’s just a matter of moving the cursor around until you hear something you like.

Processing isn’t limited to two reference voices: you can drag in several of them, each with different characteristics. Once analysed, these are displayed as new curves, each assigned a different colour. Again the various voice types are arranged automatically across the screen, so that similar voice types are placed close together and in general terms male and female timbres are spread from left to right, respectively. Moving the cursor around the screen then creates a smooth voice morph, with the lines from the curves running to the cursor to show which references are influencing the source.

Discrete cursor positions can be stored as ‘waypoints’ and assigned for control using MIDI. For example, a keyboard or foot pedal can be useful when using this software live, but conventional DAW automation can also be used to control the cursor position. Waypoint mode is entered using the icon at the lower right of the window, after which a new waypoint is added at the current cursor position when you press the ‘+’ button. A Real‑time button allows for low‑latency ‘live’ use, with the processing time of Vocoflex going as low as 35ms. If this is overly taxing for your computer’s CPU, lower quality modes that ask less of your CPU are available.

No Voices? No Problem!

In a creative environment, where you’re looking for quick results and don’t have a huge library of reference voice samples, Vocoflex offers you another option. To the right of the window, around halfway down, are two icons, one for the ‘curves’ mode I’ve just discussed, and another resembling a dot within a circle. This brings up what Dreamtonics call the Voice Generator, which shows up as a coloured square in the centre of the window.

The Voice Generator can help when you don’t have suitable reference voices to hand.

The left/right axis effects a male/female transformation while the vertical axis adjusts tonality. Below the coloured square are buttons for Random and Apply, as well as a code for the current colour. The actual vocal timbre is mapped onto the colour of the square. Clicking Random brings up a new colour and a new voice character, though you can also enter your own colour codes. Clicking Apply adds the new timbre to the main curves view, where it appears as a single point. Add one or more reference voice samples and this new timbre can be morphed with them to create further variations. The rabbit hole of timbral tweaking can get pretty deep, of course, but the process is simple and intuitive from the user’s standpoint.

Vocoflex can also be used in conjunction with Dreamtonics’ Synthesizer V Studio virtual instrument plug‑in, with which you can synthesize surprisingly realistic‑sounding pitched lyrics. Synthesizer V Studio itself can only be used with voice databases licensed from Dreamtonics (so as to avoid copyright issues), but interestingly its vocal output can be transformed using Vocoflex. A free light version of Synthesizer V Studio with a couple included voice types is available from the Dreamtonics website. A database of 40 voices is included with Vocoflex, and additional ones can be purchased if desired.

Putting It Into Practice

There are all sorts of reasons why you might want to turn one voice into another, but the big question, really, is how effective Vocoflex is at doing that job. I have to say that the transformations can be immensely useful in giving you a more commercial‑sounding voice that leans in a specific direction — assuming of course that you start off with a proficient vocal track in the first place.

If the voice you want to transform has a similar pitch range to your reference samples, then you can expect to achieve natural‑sounding results fairly easily. Inevitably there are some complex voices, especially those that include a lot of throat or lips noise, that create occasional audible artefacts when processed, but most cleanly recorded vocals work fine. Even transforming male voices to female ones or vice versa worked rather better than I expected — you just have to remember to use the onboard pitch‑shifter to match the source voice to the reference one.

This software also raises some interesting possibilities and ethical/legal questions. For example, if copyright weren’t an issue, and you used an isolated snippet of a commercial recording as your reference, could you use Vocoflex to make you sound like Elvis, or make Bob Dylan sound like Kate Bush? The answer would have to be no. Why? Simply because there’s a lot more to a singer’s sonic identity than their vocal timbre: phrasing, regional accent, the way a singer shapes their note transitions and other aspects of delivery... All these things remain unique to individuals. The way that the source vocal timbre changes during performance also has a bearing on this.

As for going the other way, Dreamtonics tell us they apply additional security, such as ID verification and file watermarking to avoid the misuse of copyright vocal recordings. While I’m unsure as to the legality of extracting somebody else’s formants without permission, I suspect there could well be serious legal problems if you were to take somebody else’s vocal track and try to disguise it using Vocoflex!

Vocoflex is undoubtedly a powerful tool that can produce natural‑sounding vocal transformations that are genuinely useful in a musical context.

When used as intended though, with reference voices you’ve recorded or to which you own the copyright, or with Dreamtonics’ voice databases, Vocoflex is undoubtedly a powerful tool that can produce natural‑sounding vocal transformations that are genuinely useful in a musical context.

Out of interest, I also experimented with processing some sources other than voices and, most notably, I found that I could give an eBow guitar recording an interesting vocal character, even if there were a few octave pitch jumps in places where Vocoflex struggled with this new ‘voice’. That is perhaps not surprising, given that the eBow can generate very strong harmonics in a very non‑vocalist kind of way. But my point is that experimentation of this kind can occasionally produce very worthwhile ‘sound‑design’ results. As a further experiment, I took some dialogue recording from a podcast I’d done with SOS Technical Editor Hugh Robjohns, and processed his voice using a sample of mine as the reference. The end result didn’t sound remotely like me — Hugh has a different accent and a different style of delivery — but importantly, what came out still sounded extremely natural and, to my ear, quite unlike either of us.

Verdict

The more I worked with Vocoflex, the more impressed I became. Simple jobs like giving the same singer different characters for creating layered backing vocals worked beautifully, but the process is clean enough that in most instances it can be applied to a main vocal without ever giving the game away, and the results really can sound exceptional. If you work with vocals a lot, or perhaps if you need alternative voices for your songwriting demos, then you should find plenty of uses for Vocoflex.