TC‑Helicon claim that their flagship vocal processor can make you sound like a totally different person. But how well does it work?
The Voiceprism is the second product of a collaboration between TC Electronic of Denmark and IVL of Canada; IVL being the vocal‑processing specialists who made the original Vocalist range of auto‑harmony products for Digitech. It draws together TC's expertise in effects processing and IVL's pitch‑shifting wizardry to provide the live or studio vocalist with a one‑box solution combining a voice channel (preamp, EQ, dynamics), intelligent four‑voice harmony generation, and vocal effects.
The TC‑Helicon Voiceprism Plus is the newest version of the Voiceprism, which sees the debut of the Voicecraft card (already installed in the Voiceprism Plus). This provides digital I/O in both AES‑EBU and S/PDIF formats and introduces voice‑modelling. The Holy Grail of voice modelling in the long term is to be able to make any singer sound like any other singer (I'm not saying this is necessarily a good thing!), but voice modelling is a new science and currently the aims are rather more down to earth.
The Voicecraft board can modify basic vocal characteristics by changing the formants produced by the human vocal tract, by remodelling some of the non‑pitched glottal elements of the voice, and by modifying the pitch‑scoops when singers move to a new note. It's also possible to add context‑sensitive vibrato that mimics the modulation characteristics of real voices — a far cry from the clumsy sine‑wave LFO. The modified lead voice can then be balanced with the original lead voice to produce a thickening effect, or it may be used on its own. The harmony generation section of the device is also able to use somewhat less sophisticated, but still quite effective, reshaping techniques to change the character of the different added vocal parts for a more realistic ensemble.
Voiceprism Plus is a 2U rackmount processor with a slightly raked front panel for ease of access. There is a large LCD display, with four 'soft' knobs below it relating directly to on‑screen parameters. A pair of Menu Tab buttons navigate between menu pages.
Separate Edit buttons are used to access the individual sections of the machine designated as Vocals, Effects, Comp/EQ, Mix and Step (a function that enables the user to create a sequential list of patches that can be stepped through in live performance using the front panel controls or an optional footswitch). A large data wheel is used primarily to scroll through the presets, while, to the right, five buttons go under the general heading of Browser. The Browser is a fast way of viewing only the preset types you're interested in, the five categories available being Shift, Scale, Manual, Chord and Effects.
At the right of the front panel are four gain knobs that control the Lead, Harmony, Effects and Input levels, and above them are eight further function buttons including 48V (phantom power), Mic On (mic/line input selection), Utility, Store and Bypass. You should be aware, however, that turning the harmony level right down here won't kill any harmony reverb if it's used in the current patch, though deactivating the Harmony button will. Preview initiates a demo mode using internal audio samples, and Help provides context‑sensitive on‑screen assistance. A separate Phones output is located on the front panel, along with its own level control, alongside the Mic input.
Other than the Mic and Phones sockets, all other connections are on the rear panel, including the mains inlet and a full trio of MIDI sockets. The Line Input is on a balanced TRS jack and the front‑panel Mic On switch selects between the Line jack and Mic XLR. Another balanced jack provides an Aux Input that allows external signals to pass through the Voiceprism's effects section but not its harmony generator. This has only a gain setting switch for +4dBu or ‑10dBV, so its level needs to be adjustable at source.
Stereo audio outs are furnished on balanced jacks as well as on a coaxial S/PDIF digital output — without the Voicecraft card fitted, there is no digital input. A further jack enables the unit to utilise a one‑button or three‑button footswitch, the functions of which are configurable, and can include harmony muting, various bypass modes, and song/preset selection.
The metering for the input is located to the left of the LCD screen, where there's also a clear, numerical readout of the patch number. Processing within the machine is all digital, following the analogue mic/line preamp, and, in addition to harmony generation, the signal path of the unexpanded Voiceprism includes compression/gating, two one‑band equalisers and two effects.
When the unit is powered up, the first screen to appear is the Preset screen. This is the top level of display and provides key information regarding the preset, such as the harmony mode in use, the key and scale type and the number and type of harmony parts. Up to four harmony parts can be added to the lead voice and these include formant‑shifting functions to change the 'gender' of the individual parts. Little face icons show the gender type of the individual harmony voices and include men, women, men with dodgy beards (and some with even dodgier moustaches), aliens and what looks like a caveman wearing shades!
The screen also shows the compressor/EQ blocks and the two effects blocks. Some patches have some of the soft knobs assigned to effect types, in which case they access the onboard preset effect libraries. Pushing a soft knob causes a drop‑down menu of the optional assignments to appear, making it easy to change what's being controlled. For the main screen, the soft options are Key, Scale, Porta(mento), Gender, Detune, Vibrato, Scoop and Timing.
Pressing any of the Edit buttons to the right of the display accesses a series of 'tabbed' windows relating to the selected function. For example, selecting Vocals and then navigating to the Harmony tab brings up four faces illustrating the gender of the backing parts. Using the soft knobs, it's possible to change the level, pan, gender and voicing of each part, and, as the gender value is changed, the little face icon morphs through its repertoire of facial types.
The harmony generation section is the main feature of the Voiceprism, and can be used either to fatten voices or to add automatic harmonies. Voice thickening uses relatively simple detuning to create a fatter sound, but the harmony generation is quite sophisticated. The pitch of the incoming voice is tracked quickly and accurately using some very clever software, then the information is used to establish what degree of pitch‑shifting is necessary to create the right harmony.
In addition to creating extra harmony parts, and specifying the intervals to be used, there are options that can be engaged to make these artificially generated parts sound more human, including the introduction of small timing and/or tuning discrepancies, changing the way the virtual singers scoop up to notes, and adding vibrato. Also, when the Voicecraft card is fitted, the vocal thickening section expands to offer a whole range of brand new vocal reshaping treatments that can significantly alter the fundamental character of the voice in a relatively natural‑sounding way (see the 'Voicecraft Or Witchcraft?' box). Not every song demands a Beach Boys vocal arrangement, so there are several harmony modes available, including having no harmony at all or only using some of the available harmony parts — some are better for studio use and others primarily for live performance.
Smooth Shift mode is a non‑intelligent, parallel pitch‑shift option where the most useful settings are based either on octaves or subtle detuning. The term 'smooth' denotes the fact that the harmony notes slide smoothly from one semitone to the next, unlike the alternative Stepped Shift mode, where the harmony pitches are always quantised to the nearest semitone value. Chordal is a useful studio or live performance mode, as it creates diatonic harmonies that conform to the musical scale suggested by whatever chord is being input from a MIDI keyboard or sequencer. This is a true intelligent harmony mode, but be aware that the harmony notes jump from one pitch to the other in a rather unnatural‑sounding, hard‑quantised kind of way.
Smooth Scale mode is probably the most commonly used for live performance, where the key and scale type of the song must be entered along with various harmony customisation options, enabling the machine to automatically add a musically appropriate harmony — as long as you sing the main part reasonably accurately. There's a section at the end of the manual that offers some useful background and advice on using scalic harmonies, as well as tables showing how the harmonies are generated. Smooth Scale mode allows harmonies to follow bends that pass through non‑scale notes, as is common in blues styles of performance, and if you sing a little flat, the harmony will follow you. Step Scale creates harmonies in the same way as the Smooth Scale mode, except that the harmony pitches are scale quantised to the nearest allowed harmony note, which again can sound unnatural, albeit very precise.
Notes mode allows for more off‑the‑wall effects to be created, as it forces the harmony notes to track the pitch of a MIDI source — even just reading the vocal part will result in musically correct (though admittedly not very natural‑sounding) harmony notes that correspond to the notes being played on the keyboard. You can wail away all you like but the pitch of the harmony remains rock solid, which makes it sound somewhat vocoder‑ish! Taking this idea one step further, there's also a Notes 4 Cmode, where four monophonic lines are fed in on four consecutive MIDI channels so that up to four harmony voices may be controlled independently. This could be quite practical in the studio when using a sequencer, as you can dictate precisely what harmony note happens and where, but again it doesn't have the simulated humanity of the Smooth Scale mode.
The effects section operates slightly differently depending on whether the Voicecraft card is fitted or not, so I'll look at the facilities the unexpanded Voiceprism offers first. The Comp/EQ edit button gets directly into the compressor/gate/equaliser menus, where the parameters and input assignments can be set up. In the Voiceprism, there are actually two EQ blocks, so that separate EQ can be added to the lead and harmony voices if need be, and there are also choices to be made as to how the dynamics processing is accessed by the lead and harmony voices. The physical routing is always shown by lines linking the processing blocks in the Preset window, and the dynamics section can either be used by the lead voice, the harmony voice, or neither. EQ1 and EQ2 are identical one‑band processors, offering a choice of five filter types. These include high‑pass and low‑pass filtering variable from 80Hz to 16.3kHz; a variable‑frequency low‑shelf section; a parametric mid‑band; and a variable‑frequency high‑shelf section. All offer ±12dB of gain.
The compressor page has Threshold, Ratio, Attack and Release controls that follow the soft knobs, but there's no make‑up gain. In practice, the compressor brings the signal level up, so it may be necessary to reduce the level setting in the Mix screen to avoid clipping. A separate gate has Threshold, Attack and Release controls, it's main purpose being to attenuate noise during pauses between words or phrases.
The Mix button brings up level controls for the Lead and Harmony voices, as well as for the FX1 and FX2 returns. A further page sets the levels of each of the harmony parts.
FX1 can be configured as a chorus, flanger or delay, with numerous mono and stereo delay options that are similar to those offered by dedicated multi‑effects boxes, including a tape delay simulation. Both the lead and harmony voices can share the effects, or they can be assigned to one and not the other. FX2 presents exactly the same choices as FX1, but with the addition of high‑quality reverb algorithms. Here you can choose from Studio, Chamber, Club, Hall and Arena, with control over pre‑delay, decay time, high‑frequency damping and low‑frequency roll‑off.
The Voiceprism Plus (or a Voiceprism with a Voicecraft card fitted) adds to the above flexibility by doubling up on the EQ and compressor/gate capability, enabling two sets of EQ and separate compressor/gates to be used in both the lead and harmony voice signal paths. The parameters are the same as before, but now there's a greater degree of independent control over the lead and harmony voices. In addition to the extra facilities, the dynamics, equalisation, and effects algorithms within the Voiceprism Plus also, according to the manufacturer, sound superior to those of the standard Voiceprism, though I had no opportunity to verify this for myself.
Both versions of the Voiceprism are easy to navigate, and the accuracy with which incoming vocals are tracked is almost uncanny, providing the voice level is high enough to keep the Lock LED lit. A Lock indicator in the display window lights when the input signal pitch is being followed successfully, which is pretty much all the time when the input is loud enough.
The harmony generation also seems pretty foolproof, even in the scalic modes, though you do have to enter the right musical key for the song and sing in tune. Some songs include key changes, in which case you need to create two presets for the two sections of the song or manually change the key on the front panel. You may also find some songs with 'out of scale' sections that don't respond well to auto harmonisation, in which case you have the choice of muting the harmonies at that point or using one of the MIDI control modes to ensure the correct harmonies are generated. For example, a live performer working from sequenced backing could use the sequencer to control the harmonies, even though the stepped nature of such harmonies is not altogether natural. If you sing slightly off key (as I invariably do!), the 'in‑tuneness' of the harmonies makes you sound worse, as the designers decided against building in lead vocal pitch‑correction facilities.
In most respects, the operation of the harmony modes isn't that different to earlier Digitech Vocalist products, but the quality and flexibility of both the lead and harmony sounds have evolved quite a lot. Taking the harmony sounds first, these include formant shifting, enabling individual voices to be changed in apparent gender. The process is, however, fairly crude compared to the voice‑modelled lead voice processing and, when you listen to the harmonies in isolation, they do sound unnatural when shifted too far. Very deep voices have a robotic growl to them, whereas the higher female voices take on a hint of cartoon‑character/helium‑breath timbre, but if you moderate your adjustments, the sounds are plausible enough in the context of a mix. You probably wouldn't use these harmonies when recording a ballad, as there's always a slight hint of electronic artificiality present, but in today's musical climate the sound would probably slot in quite nicely with many pop styles. Adding effects such as reverb to the voices helps disguise the artifacts, as does setting slightly different delay, pitch and scoop parameters for each of the harmony voices, but they never manage to sound completely natural.
Of course, the real area of interest is the voice modelling, which boldly goes where the tonsils of man have never gone before. Working through the presets provides a very fair idea of how the process works when somebody has taken the time to optimise the various parameters, after which it's very easy to go into edit mode and juggle the available user parameters.
As expected, the most realistic treatments are the ones where only subtle changes have been made to the original voice — a bit of breath here, a bit of throaty growl there — and it helps if the formant restructuring doesn't stray too far either. However, I wouldn't say any of the modelled voices sounded quite 'real', especially when heard in isolation. Changing male voices to female, and vice versa, definitely sounds artificial to my ears, though it is still perfectly usable in a musical context where you're going for an effected rather than natural sound. The more drastic treatments always seem to have a noticeable metallic, digital edge to them that effects can't always disguise, and adding too much growl makes you sound somewhat like Darth Vader gargling with Swarfega — far too robotic to pass for human. Providing your expectations are realistic, and you use VM to fine‑tune your voice rather than to completely rebuild it, some of the results can be surprisingly good, if not entirely natural sounding. It's also interesting to see how little you have to change some elements to make the voice sound as though it belongs to a completely different person. I found the inflection and vibrato parameters worked a lot more convincingly than those that attempted to change the essential character of the voice.
Judged as an all‑in‑one voice channel and voice processor, the Voiceprism Plus comes out pretty well, with a clean mic preamp, friendly and musically kind compression/EQ, and effects that manage to both sound good and be simple to control. The harmony features are as easy to use as earlier Vocalist units (if not easier), but the way in which the character of the individual harmony voices can be modified and made to sound less synthetic has advanced considerably.
The voice modelling aspect of the machine is to my ears the least successful, and clearly there is some way to go before voice modelling can completely rebuild a voice and still make it sound completely natural, but it is remarkable what has been achieved already. You can't yet dial in the name of your favourite singer and expect to sound like them, but you can add throatiness for soul and blues singing, or you can change your voice character to make you sound like a larger or smaller person. The human vibrato modelling is also a real step forward from simple electronic vibrato. A competent male singer can expect to go from blues singer to boy band without too much trouble, but going from Louis Armstrong to Britney Spears might just show up the limitations of the current technology. Being realistic, I'd say you might get away with some of the modelling effects live, where the voice also has other effects added to it, but in the studio, it's still only at the stage where you could use it as a special effect rather than as a convincing substitute for the real thing.
OK, it's all cheating if you choose to look at it that way, but then, with any other instrument, you can buy new hardware or learn new playing techniques to change the sound to imitate that of other players if you want. Using Voiceprism Plus, you can at least modify your voice to suit the song style, and, though these are early days for this new technology, who knows what level of sophistication the future will bring?
When the Voicecraft card is fitted, or when you buy a Voiceprism Plus which has the card already installed as standard, the signal path of the unit is made to work slightly differently. Instead of a Voice Thickening (VT) block, you now get a Voice Modelling (VM) block. Voice modelling is an immensely complicated subject, so presenting the user with all possible adjustable parameters would be about as sensible as asking plankton to learn how to program a video recorder! Instead, the designers have produced a library of treatments which can be applied to the lead voice for a range of different effects.
The Lead menu is used mainly to create thickening effects which can be achieved by layering the remodelled voice with the original in any desired proportion. All the user needs do is pick a Style and an Amount, though be warned that some of the Styles towards the end of the list are a little 'abstract'. Any of these Styles may be stored as part of a user preset, along with all the other Voiceprism settings.
To fully understand what this level of voice modelling is able to achieve, it is necessary to examine those VM effect parameters that have been made accessible to user control. VM Spectral is a set of special preset equalisation curves that are tailored for use with human voices and that may be used in addition to the various resonance parameters available in the VM Warp section, or used simply to provide additional tonal control. Like each of the VM effect menus, there is a separate Amount control.
VM Warp is a way of modelling vocal tract resonances so that they can be superimposed on the source vocals. This involves moving formants around and can make the resulting voice sound very different to the original, even to the extent of changing the apparent gender of the performer. Once again, the VM Warp options are presented as a range of Styles along with a warp Amount control.
VM Glottal deals with unpitched sounds and is divided into Breath, Rasp and Growl. Breath adds breathiness or a whisper‑like characteristic whereas Rasp adds a little earthy grind to the voice — it should be calibrated in bottles of Jack Daniels, but you only get an Amount setting! This is the one to play with if you want to sound more like Rod Stewart or Bruce Springsteen, but without straining your voice.
Growl is intended to add the raunchiness of a soul or blues singer to a voice — probably the parameter Charlotte Church would go for if asked to cover a Tina Turner song. Some of the available Styles are level sensitive so that you can add more by singing harder, and again it's up to you how much you dial in once you've found a preset Style you like.
Voice modelling isn't just about imitating the spectral response of another singer. It's also about controlling the way you change pitch and add vibrato. VM Inflect adjusts the way the singer scoops from one note to another, and can handle both scooping up and scooping down at a choice of fast, medium or slow speeds.
The 'onset' of pitch inflection is variable, depending on the processing Style chosen, which in this context relates to how long you need to pause before 'scoop mode' is enabled for the start of the next sung phrase. Again the degree of inflection can be varied in the same way as the other VM effect components and some of the more extreme special effect examples are definitely on the weird side.
Vibrato is taken care of by the VM Vibrato effect, where the vibrato characteristics are based on those of real singers. Some intelligence is used by the program to determine when the vibrato should come in, so it's not like applying a blanket synth modulation treatment. Also, the actual modulation waveform of a human vibrato is quite different to a simple sine wave modulation and may not even be symmetrical.
Another enhancement provided by the Voicecraft expansion card is the addition of digital I/O in both S/PDIF (coaxial) and AES‑EBU formats, with a choice of 44.1 or 48kHz clock rates, external sync and dithering to 8, 16 or 20 bits. Digital routing is accessed via a new menu where you can select the digital source (AES‑EBU or S/PDIF), the analogue or digital source for the lead voice input, and the analogue or digital source for the auxiliary Input, with the option to choose either the left or right digital channel as a source. The built‑in digital output changes function once the Voicecraft card is installed, so that it outputs the unprocessed mic/line input on the left channel and the auxiliary input on the right channel.
All digital processors take some time to do their thing, and the Voiceprism is no exception. In this case, it takes around 20mS for optimum operation, but it is possible to choose a shorter delay time for greater comfort in live performance, with correspondingly less accurate analysis.
If you're considering the Voicecraft's physical modelling for critical lead‑vocal work in the studio, it's important to bear in mind that the modelling algorithm changes the vocal sound quite noticeably even when set to a 'flat' setting (with the Amount controls on all the modelling pages set to zero percent). The manufacturers have informed me that this is because a new generic formant pattern is imposed on the vocal when the sound is stripped down and remodelled by the algorithm. While this might happen to accurately match the particular voice you are processing, I found that it tended to dull the vocals I put through it, making them sound rather woolly. What's more, the tonal change was sufficiently complex that I could not easily compensate for it using either the spectral‑shaping options within the modelling pages, or the two available bands of EQ — I found that high boost with the latter brought up low‑level processing artefacts unpleasantly.
The fact that the voice‑modelling algorithm changes the vocal tone means that if you try simply to add a bit of extra growl to an otherwise finished vocal sound, you'll probably lose more than you gain. However, if you're intent on drastically changing the sound anyway, then this might not bother you very much. Mike Senior
- All the elements of a voice channel, plus harmonies and voice modelling in one box.
- Easy to use.
- Best ever tools for making harmonies sound more realistic.
- Voice modelling can change the character of the lead voice in quite realistic ways, providing it is used with discretion.
- The harmony voices and modelled lead voice can sound quite synthetic if you try to introduce dramatic changes.
This is the first voice‑modelling product that attempts to modify a lead vocal in a natural way, and while it is not always successful when used to extremes, it can be a very creative tool when used with care.