EastWest’s new Hollywood Choirs library combines cinematic gloss with sophisticated word-building tools.
Sample libraries geared towards the epic and cinematic currently seem to be coming at us from all angles. Given the predominantly orchestral bias of these, it’s no surprise that choirs are in big demand, particularly those that offer some means of verbalising lyrics, be they in Latin, Russian, English or just plain invented gobbledegook. In researching this review, I was surprised at just how many there are, each with their own particular twist on the subject (see the ‘Alternatives’ box).
Broadly speaking, they fall into two categories: those with ‘word builders’ that allow you to type in any words and (hopefully) have the choir sing them with some degree of intelligibility; and those that employ some sort of scripting to chain together pre-recorded syllables, whole words or tempo-sync’ed phrases, often in Latin, to give the impression of words that may or may not have any real meaning. Both approaches are valid and can, with a fair wind, give a real sense of animation and emotional intensity where static ‘ooh’ and ‘aah’ vowels on their own just don’t cut the mustard. Hollywood Choirs (HWC) falls into the former category, and features brand new recordings with, the makers say, better singers than their previous Symphonic Choirs library. Hopefully none of those singers who toiled faithfully to create Symphonic Choirs are reading this!
HWC is EastWest’s third foray into the word-building field; their first WordBuilder (WB) engine was introduced in 2005 with EWQL Symphonic Choirs, originally running as an add-on application alongside NI’s Kompakt engine. The library was subsequently reworked to run in the Play engine, with WB integrated directly into Play.
Whereas Symphonic Choirs featured separate SATB and Boys sections, HWC focuses on two sections: men’s and women’s, each loading as WB Multis into their own instance of Play. The range of each section covers basses and tenors for the men, and altos and sopranos for the women, so if you wish to write discrete SATB parts that take counterpoint and individual dynamics into consideration, it’s perfectly possible using four instances of Play — two for the women and two for the men. Also, HWC features version 2 of the WB engine, which has undergone a number of improvements since Symphonic Choirs (see the ‘WordBuilder 2 Enhancements’ box).
Before we get into the tricky (and fun) business of word-building, we should look at what else HWC offers. The Play Browser presents two root folders: Men’s Choir and Women’s Choir. Inside each of those are three folders: Consonants, Vowels and WB Multi. Checking out the vowels first, we find patches for sustained ‘aa’, ‘ah’, ‘ee’, ‘eh’, ‘eu’, ‘ih’, ‘oh’, ‘oo’ and ‘uh’. With a couple of exceptions, these are all sung without vibrato, and have a lovely clear sound; the modwheel crossfades through three dynamic layers ranging from a whisper quiet mp through mf to a rip-roaring ff. Staccato patches are included for ‘ah’, ‘ee’, ‘eh’, ‘oh’ and ‘oo’ — these utilise two round robins. Additionally, ‘ah’ sustains have Vib ff and Epic variations; the difference seems to be that the Epic patch uses CC1 to crossfade between Vib ff and mf non-vib layers.
There are also ‘ah’ and ‘oh’ patches designated as Legato; we need to be clear about what is meant here, as developers don’t yet seem to agree on the definition of ‘true legato’ which, when applicable to the human voice, implies forced monophonic behaviour with smooth, portamento pitch transitions from note to note. HWC’s legato patches don’t behave this way — they are fully polyphonic, and do not employ portamento transitions. I asked EastWest to clarify their implementation of legato and the main points in their response were these: “When playing connected (slightly overlapping) notes up to an octave in either direction a legato slur layer will be triggered. When playing disconnected notes it plays the sustain layer... The legato is a somewhat subtle effect with a large choir... They sung it as they would sing connected in a real performance situation.” So this is essentially intended as a slurred legato — the sound is continuous, and the pitch change is immediate. Selecting any one of the Portamento, Legato or Other buttons on Play’s GUI does force monophonic behaviour; however, the transitions are manifested as unnatural-sounding crossfades between notes, which presumably is why those scripts are left ‘off’ by default.
Another popular feature found in many choir libraries is some means of real-time morphing (crossfading) from one vowel to another, using a MIDI controller. This is not inherently possible in HWC, since each sustained vowel is contained within its own unique patch. The Play engine doesn’t allow any kind of user configuration of MIDI controllers, so the only way to achieve manual morphing (and retain dynamic control) is to load multiple vowels into Play on separate MIDI channels, and control them with multiple CC7 (volume) and CC1 (dynamics) controls — assuming your controller keyboard is equipped to do this. EastWest point out that vowel crossfading is possible in the WordBuilder interface, and this is indeed true. However, as will be explained later, this has to be pre-meditated and programmed, and therefore embedded within a WB text sequence as a precisely timed event — it cannot be performed spontaneously in real time with a knob or fader.
The Consonants folder contains separate patches for every pitched consonant (eg. ‘d’, ‘v’, ‘z’, etc); one patch of breath noises; plus one patch of non-pitched consonants (‘s’, ‘ch’, ‘sh’, ‘h’, etc). The non-pitched variety are assigned one per key, and occur twice on the keyboard — one octave of lower pitched versions, and one octave of the same consonants at a higher (ie. brighter) pitch. The pitched consonants are interesting in that they continue to sustain after the initial consonant has sounded, so ‘g’, for example, sounds as ‘gerrrrr’, ‘d’ as ‘derrr’. These are useful textures in their own right; the ‘m’, for example, serves very nicely as a pad of humming choristers.
One school of thought holds that lyrics sung by choirs are rarely intelligible, especially if they’re not in your native tongue, or when the choir is accompanied by an orchestra going full tilt (less so if the choir is singing a capella). In some cases intelligibility is not so much of an issue, it’s the mood, animation and intensity that counts; the language and subject matter are neither here nor there. This explains the predominance of Latin, almost always liturgical in nature, in other phrase-based libraries. It just sounds good, and conveys a sense of gravitas. To that end, HWC’s WordBuilder includes a library of pre-programmed phrases to get you going — not just Latin, but English and German too. They’re not only useful as they stand, but also serve as tutorials for those ‘how do they do that?’ moments when a particular sound seems elusive.
Happily, your own hand-crafted apocalyptic prophesies can be saved to the phrase library. I’d recommend re-saving your phrase (and DAW project — belt and braces) every time you reach a point at which you’re happy. Even though there are undo and redo buttons, sometimes you’ve made such a pig’s ear of things it’s easier to reinstate the last known good version and resume from there.
Once a Men’s or Women’s WB patch has loaded, clicking the WordBuilder button on the GUI opens the Text Editor. The phrase ‘A nu bi gEn Eng’ (a new beginning) is displayed by default in the text area; each new note played on the keyboard advances the choir through the syllables one by one. My first impression was ‘eh?’ — so plunging into the deep end, re-working the phrase to be ‘A nyu! bi gin ing’ produced a subjectively more intelligible rendition to my ears. Which only goes to demonstrate that this is an inexact science — what sounds right to one person sounds wrong to another (see the ‘Babel Tower’ box).
Text can be input in one of three ways: English, Phonetics or Votox, the latter having been specifically developed for WordBuilder. English would seem the obvious choice to start with, and WB makes a decent crack at interpreting it. What it doesn’t deal with well is multi-syllable words — ‘butter’, for example, plays a long ‘baaah’ followed by a brief ‘tr’ when you release the keys. Simply separating the syllables doesn’t quite elicit the desired effect, so it’s time to get to grips with Votox and Phonetics, rules for which are listed in detail in the manual. Quick reference lists are also available as drop-down menus on the GUI. In the case of ‘butter’ it turns out that typing ‘ba tAr’ in Votox puts a better bit of butter on your knife.
Many more factors influence the success and realism of WB, not least of which is timing. A few milliseconds’ difference in note length and placement can make all the difference between lumpy and smooth, so expect to spend time not only with pronunciation but also shifting things around in your DAW’s piano roll editor — there’s a very fine line between rejecting something and suddenly hitting paydirt after what seems an insignificant note tweak.
WB provides plenty of tools to fine tune details, notably the Time Editor, in the lower half of the GUI. Selecting a syllable in the Text Editor displays each phoneme used to generate it on individual lanes. Activate Solo in the Word box and you can repeatedly play that syllable while editing, without stepping through the whole phrase — very handy. The phoneme lanes can be moved around, lengthened or shortened — very useful for increasing or reducing the length of consonants. Each lane also has its own adjustable volume envelope; this enables the creation of customised diphthongs (connected vowels as in ‘boil’ and ‘naïve’) that crossfade from one vowel to the other over an exact length of time. Votox includes special characters for diphthongs, but sometimes you need the timing to be quite specific, and this manual volume sculpting is the way to do it. HWC also has a handy speed learning function to assist in such cases: highlight the syllable in question, click Learn and play the word or phrase. HWC adjusts the syllable’s length accordingly and the time editor now shows its actual time scale, making it easier to gauge the shape of each vowel’s volume envelope to get the transitioning effect you want.
Because phrases step through their syllables in a cyclical loop with each key press, some means of starting a phrase at the beginning is necessary. The Reset Position button on the GUI accomplishes this; the same action can be applied in your DAW sequence by inserting a MIDI CC20 command (value 127) at the beginning of the phrase. But what happens when you’re working with long phrases — even entire verses’ worth of lyrics? Having to start from the beginning of such a phrase just to hear a bit at the end could frankly drive stress levels through the roof!
Fortunately, WB2 includes DAW sync’ing to get round the problem. As with the speed learning described earlier, play through the entire sequenced part in Sync/Draw Only Learn mode, and HWC will now follow the lyrics from any point. There is one caveat; if your DAW is configured to chase notes (ie. notes sound wherever in their duration you hit ‘play’) you must start playback during a gap between syllables. If you start playing in the middle of a note, WB will simply continue its cycle from wherever it last left off (usually in the wrong place), and will never realign itself correctly. Nevertheless, bearing that caveat in mind it’s good to know you don’t have to be driven nuts trying to examine the final few words of a long phrase.
HWC Diamond edition (on which this review is based) was recorded with a total of 13 microphones, presented as five fader groups: the Main mix, a pre-mix of all mics that sounds great for most applications, and is the only selection available in the HWC Gold Edition; Spot, four close mics; Wide, two outriggers capturing the most stereo width; Stage, a L/C/R Decca Tree of three omnidirectional mics; Mid, a Neumann Binaural Dummy Head and a Josephson placed in the audience; and Surround, two mics placed left and right at the rear of the auditorium (also provided in the Gold X ComposerCloud edition). Having control over the individual mic groups has obvious benefits such as emphasising the close mics to improve intelligibility, or when working in surround. The hall graphic in the centre of the Player page GUI helpfully illuminates the currently active mics.
The Mixer page shows eight channel strips (14 if a WB Multi is loaded), each one showing MIDI channel for that part, pan and volume, mute, solo and FX buttons, and an output router for directing specific mic pairs to separate inputs on your DAW. A sub-mixer button reveals a further five strips that reflect the mic level settings on the Player page. All mixer and sub-mixer pan and volume faders are linked when a WB Multi is loaded, so any adjustments will automatically be carried across to all the WB’s parts and mic settings.
The FX buttons reveal insert effects for each channel strip. These comprise an amp simulator, Ohmicide multiband distortion, Play’s convolution reverb, an SSL-licensed plug-in comprising HP/LP filter, EQ, compressor, gate/expander and transient shaper, and an SSL stereo bus compressor. One curiosity here is that although the ‘Master On’ button ensures that changes to the master volume and reverb send level affect all parts of a WB Multi equally, it doesn’t apply to the effects settings. This is problematic if you want to apply any of HWC’s effects globally to the entire WordBuilder output, as there is no master out channel strip with its own effects insert. The only method I could find was to set up the effects for one part, save it as a preset, and load that into the other 13 parts — a rather tedious approach, especially if you make any adjustments, which means you have to repeat that process all over again. Also, the transpose/tuning controls in the Master section only affect the currently selected part, but not all — this is certainly something worth correcting.
There’s a definite art to this word-building business. On the journey there will be laughter, there will be tears, then a point where you sit back and go ‘that’s feckin’ brilliant’. Most of all it will be absorbing and fun. Those expecting out-of-the-box instant gratification may be bemused at their first attempts; like any instrument of complexity and depth it requires time and patience to tame, but it can deliver rewards given the chance. Some may use it as a mock-up to be replaced by a real choir, or to add weight to additional real voices, or be more than happy to embrace it as it is. It will be fascinating to see where developments take WordBuilder in the future.
As for the lack of portamento-style legato and real-time MIDI control of vowel morphing, I suspect the reason most people will be drawn to HWC is the word-building, and they may well be owners of other libraries that cover those missing features. The question of whether owners of Symphonic Choirs should add HWC to their collection is very much down to the individual. Essentially you’re getting an alternative set of men’s and women’s choirs, new phrase presets, new vowels and consonants and various other WordBuilder improvements such as Sync to DAW. Perhaps an expensive upgrade for Symphonic Choir owners, but for those new to the WWW — the Wacky World of Word-building — it could be just what they’re looking for.
There’s no shortage of choir libraries out there, so these alternatives concentrate on those that offer some kind of ‘verbalising’ facility. One’s choice is ultimately guided by how specific (or not) you want the lyrical content to be. Word builders allow for completely freeform input of words, syllables and consonants; the rest work within the constraints of pre-defined words, phrases or syllables. The descriptions here are painted with a very broad brush; the term ‘syllable phrase builder’ simply signifies some sort of scripting or other means to chain the supplied words or syllables together in the required order.
EW Symphonic Choirs
Virharmonic Voices Of Prague
Virharmonic Czech Boys
Syllable phrase builder
Soundiron Olympus Symphonic Choir
Syllable phrase builder
Strezov Storm Choir 2
Syllable phrase builder
Syllable phrase builder + tempo-sync’ed Latin chants
Syllable phrase builder + multi-vowel sequencer
Syllable Phrase builder + tempo-sync’ed Latin chants
Tempo-sync’ed lyrical multi-syllable phrases
Best Service Cantus
Syllable phrase builder
Best Service Mystica
Syllable phrase builder
Fluffy Audio Dominus
Syllable phrase builder
VSL Vienna Choir
Performance Samples Oceania
The issue of pronunciation is an interesting one that very much depends on which continent you come from and which language you’re using. HWC’s documentation, tutorial videos and the Votox/Phonetics rules shown in the GUI’s guide menus approach this from the perspective of American pronunciation, which is frequently at odds with the pronunciation you’d expect to hear from a classical English choir.
For example, the manual suggests using a Votox ‘u’ for the ‘o’ in ‘money’. That actually sounds as ‘merney’ — the US pronunciation. To sound English, you actually need a Votox ‘a’ to get an ‘uh’ sound. Similarly, Votox ‘a’ is suggested for the word ‘copper’ — again, this gives the US-sounding ‘cupper’, so a Votox ‘o’ should be used instead. The ‘per’ of ‘copper’ is also potentially awkward; the manual lists Votox ‘ur’ as in ‘purple’, but that actually sounds like ‘ooor’. Typing ‘Ko pAr’ finally does the trick, producing something closer to the flat-vowel English-sounding ‘per’ word ending.
I also found (as in this case) that going to the Options and selecting Latin input mode as opposed to English mode sometimes produced more predictable results. Alternative language versions of the Votox and Phonetic vowel pronunciations would be a great time-saver and make a useful addendum in the manual.
The new version of WordBuilder includes a number of new features, starting with over 100 new phrase presets (and the promise of more to come), with some new vowels and consonants added to the sample pool. In a move to increase intelligibility, the relative balance between vowels, pitched and non-pitched consonants can now be adjusted with three volume sliders; these volume settings are applied per instance of Play and all new instances default to 100 percent for all three volumes.
In a similar vein, the volume scaling of consonants adjusts automatically according to how many notes are played in a chord. Alternate takes of non-pitched consonants can now be specified using a numbering system, intended to avoid the inconsistency of random round robins. Volume, pan and mic mix settings are applied across all parts of a WB Multi, avoiding the need to adjust each part individually. Most significantly, WB now has Sync to DAW functionality (described in the body text) potentially saving huge amounts of time and frustration when working with especially long phrases.