Deservedly, Vivaldi's oratorio Juditha triumphans – the only survivor among the four works Vivaldi is known to have contributed to the genre – has led a privileged existence since its discovery. It belongs to the volumes of the Foà donation (the first half of the Turin manuscripts to be uncovered), and in an article of 1927 drawing attention to the find, the discoverer, Alberto Gentili, was already singling it out for special praise. It was an early candidate for revival among Vivaldi's vocal works, receiving its modern premiere at Siena in 1941. The Accademia Musicale Chigiana produced a facsimile edition in 1948, and since this time there have been two editions published by Ricordi: the first, edited by Alberto Zedda, in 1971, the second, edited by myself, in 2008. Its numerous recordings have in recent years included versions by Robert King (1998, Hyperion CDA67281/2), Federico Maria Sardelli (2000, Amadeus AM133-2), Claudio Scimone (2001, Warner Fonit 8573 85747-2) and Alessandro De Marchi (2001, Opus 111/Naïve OP 30314).

Far fewer oratorios than operas have survived from Vivaldi's time. An oratorio was usually performed privately and rarely reached the commercial copyists through whom operatic music primarily circulated. For the Ospedale della Pietà, the Venetian foundling home where Juditha triumphans was first performed in 1716, no other oratorios are extant between 1683, the year of the earliest appearance there of the genre, and 1768, when Bonaventura Furlanetto's Joseph pro-rex Aegypti was given. To find a surviving comparator for Juditha triumphans in terms of the richness of vocal and, more especially, instrumental resources deployed, we have to go to a serenata, Il ritratto dell'eroe, composed for the Pietà in 1726 by Giovanni Porta, the newly appointed choirmaster of the institution. Porta does not have a fraction of Vivaldi's musical inventiveness, but his work at least shows that the parade of rare obbligato instruments and use of a large number of solo singers in addition to a choir belonged to a ‘house style’ rather than being a peculiarity of Vivaldi's.

Even more than most oratorios, Juditha triumphans, for all its vivid dramatic qualities, is a work to listen to and imagine rather than to watch. Its original audience at the Pietà had no choice but to adopt the first mode of reception, since the performers, being all female, were screened from it, for reasons of decorum, by grilles and black gauze. But even if we concede that the staging of Juditha triumphans was already in its day a valid, albeit remote, possibility, there are strong pragmatic objections if historical ones do not already suffice. Since the new recording reviewed here has been compiled from four staged performances by Pinchgut Opera at Sydney in December 2007, it illustrates in vivid fashion the pitfalls that beset this form of presentation.

First, there is the problem of the recitative. Because Vivaldi envisaged that the recitative for this oratorio would be read from the page, it is slightly less formulaic and more varied than ordinary operatic recitative – therefore less easy to commit to memory. Unexpected and inappropriate hesitations on the part of singers bear witness to this difficulty in the recording. Second, there is the problem of dramatic plausibility regarding the chorus, which has to represent on some occasions Assyrian soldiers, on others Assyrian servants, on others the maidens of the besieged city of Bethulia, and, in the final chorus, the population of Bethulia. The Pietà's choir, whose tenors sang at pitch and whose basses sang either at pitch or an octave higher, could cope successfully with all these roles on account of its invisibility – and a modern mixed choir reading from the music does almost as well. An opera chorus, however, risks provoking involuntary mirth if dressed now in all-male, now in all-female, garb. There are disconcerting imbalances in the choral sound of this recording, which suggest that the male singers of the chorus for Bethulian maidens at the end of Part 1 were more distant from the audience (and hidden, too?), while in the choruses for the Assyrians it was the sopranos and altos who were more distant. In the choral interjections of the aria ‘O servi volate’ for Vagaus, the steward of the Assyrian general Holofernes, the production evidently removed the women altogether, reassigning the top two parts to male voices in the lower octave and thereby producing a barbershop-like effect. Staging has also modified – I would say adulterated – the style of vocal delivery in at least two places. The ‘barbershop’ ensemble close with a drop of pitch on their last chord as if to suggest drunkenness (although there is no reason why the servants, as they rush hither and thither to prepare a banquet for Holofernes and Judith, should already be in this state). Worse: when Judith performs her historic task of beheading the sleeping Holofernes, Sally-Anne Russell's vocal line mutates for a few bars into a hysterical Sprechstimme. More generally, the recitative often seems a shade too conversational for its oratorio context. Unsuitably operatic, too, are the unmeasured tremolandos occasionally introduced into the continuo bass. The instance at the opening of the recitative during which Judith murders Holofernes is especially unfortunate, since it negates Vivaldi's carefully planned cumulation of excitement.

Like nearly all the oratorios given at the Venetian ospedali, Juditha triumphans has a Latin text. The preference for Latin over Italian was not fortuitous: it highlighted the good education offered to the institutions' wards, conferred a more emphatic religious stamp on the product (very important for the public image of the ospedali) and offered local dramatists who fancied their skill at writing drama and poetry in Latin an opportunity to try their hand. One might in all innocence believe that for modern professional singers and choirs Latin is no more difficult to pronounce and memorize than Italian, but this recording suggests otherwise. There is a chaotic jumble of soft and hard ‘c’s and ‘g’s before the vowels ‘e’ (or ‘æ’) and ‘i’ that gives the impression that no one has bothered to attempt to impose uniformity. Even individual singers veer abruptly from one pronunciation to the other.

The performance is not without its good points. Some of the obbligato instruments sound wonderful – the soprano chalumeau in Judith's aria ‘Veni, veni me sequere fide’ cannot ever have been lovelier in tone – and the tempos of the arias and choruses (there are no duets or ensembles) are well chosen. All the soloists are capable: I particularly liked Sara Macliver in the role of Abra, Judith's maid, since she, alone of the soloists, preserves the measured delivery proper to oratorio singing. The countertenor David Walker, as Holofernes, does not sound out of place as the sole male singer in the quintet, even if his diction is occasionally spoiled by distorted vowels or weak final consonants.

But we must return, sadly, to negatives. One very unfortunate feature of the arias throughout is the manner in which ornamentation has been applied. For a start, it is limited to da capo repeats instead of appearing, in increasing doses, in all three vocal sections. More important, what we hear is not so much embellishment as recomposition. Students of harmony may recall the type of exercise where a texture is complete except for one prominent line, which has to be reinvented to fit the context. This is the style of the da capo variants, which are so consistent in nature as to appear to be the carefully prepared work of a single person. They are not always more complex or faster-moving than the original line and frequently use few, if any, of the original notes as scaffolding, contrary to the advice of theorists and the surviving notated examples of the period. Admittedly, these variants are often highly attractive in their own right – but that is hardly the point, for the impression they give of the practice of ornamentation in Italian music of Vivaldi's period is fundamentally false. On one occasion – the aria with chalumeau mentioned earlier – this recomposition results in a truly disastrous change of poetic allusion: from birdsong effects mimicking the faithful turtle dove (its characteristic motives are familiar from Vivaldi's ‘Summer’ Concerto, Op. 8 No. 2) to an evocation of the treacherous cuckoo.

Then there is the choice of sinfonia. Vivaldi's autograph score lacks its overture, which was written down on a separate gathering that is today lost (the sinfonia was perhaps borrowed for use in another work and never returned). In purely pragmatic terms, it is very effective to dispense altogether with a sinfonia and begin in medias res with the blood-curdling D major chorus of Assyrian soldiers, as do most modern performances (and also Sardelli's recording). The alternative is to find a suitable sinfonia elsewhere in Vivaldi's oeuvre. But one inflexible Vivaldian custom needs to be borne in mind. This is the composer's insistence that the last chord of one number be different from the opening chord of the next so that a genuine chord progression, with its affective potential, links the two movements. Counterintuitive as this may seem, it means that D major is excluded as the key of the sinfonia unless the chorus takes the place of its finale after a slow movement placed in a different key. The present recording chooses, as De Marchi did earlier, the opening movement of Vivaldi's later (but stylistically compatible) concerto rv562a, which, like the chorus into which it is made to lead, is in a D major resplendent with trumpets, drums and oboes – but this very similarity results in a surfeit of fanfares and lessens the impact of the chorus when it arrives. It would have been far better to seek out a sinfonia for strings alone in A major or G major.

Then there are the transpositions. Some recitatives for Ozias, high priest of Jerusalem, and Abra are ‘punctuated’: that is, selected pitches are altered to bring the vocal line into a more comfortable register. No quarrel with that, since this practice is as ‘authentic’ as it gets: several Vivaldi scores (for example, that of his opera L'Olimpiade) give evidence of customization of this kind for individual singers. The transposition of certain arias for Ozias and Abra becomes more problematic, however, since it upsets the delicate tonal balance between these arias and the adjacent recitatives. All would have been well, had the beginnings and closes of recitatives been modified following Vivaldi's own example, either to preserve or to remodel in an acceptable way the tonal interface between arias and recitatives; but nothing has been done to mend the bad joins.

I have left the worst till last. This is the presence of several unexplained and unjustified cuts of arias and surrounding passages of recitative. In Part 1 Vagaus loses the aria ‘Quamvis ferro et ense gravis’ (its place is taken by an otiose improvisatory passage for harpsichord, perhaps recorded afterwards in a studio) and Judith loses the da capo of the A section of her famous aria ‘Agitato infido flatu’ (so that the movement ends unexpectedly in the dominant); in Part 2 Abra loses ‘Fulgeat sol frontis decorae’, Holofernes ‘Nox obscura tenebrosa’ (its place usurped by a quasi-improvised cello solo in G minor after the manner of the Bach unaccompanied suites), Judith ‘Vivat in pax’ (amazingly, for this is a favourite movement of the oratorio, comparable with ‘Nulla in mundo pax’ opening Vivaldi's motet of that name, rv630) and Abra all but the opening A section of ‘Si fulgida per te’. The reason for these damaging, seemingly purposeless cuts, to which the booklet makes no reference (even the libretto printed there tacitly leaves out the lines concerned), cannot have been the capacity of the two discs, each of which contains less than an hour of music. I can only guess – and may everyone connected with the recording forgive me if my surmise is incorrect – that these movements were in fact included in the live performance, but none of the takes of the cut movements and passages passed the quality threshold.

For that reason alone, this recording would not be recommendable. On a superficial hearing, it sounds good enough, thanks to Vivaldi's exciting music and the basic competence of the performers, and perhaps it will do something for the composer's reputation in the Antipodes. But when one listens to it carefully, score in hand, it falls to pieces. It is remarkable how many of its shortcomings are due, directly or indirectly, to the decision to stage Juditha triumphans as an opera and then record it live. With a more historically aware approach to vocal embellishment, the same singers, players and musical direction could certainly have achieved much more satisfactory results in the concert hall or recording studio. The moral is clear: let operas be operas, let oratorios be oratorios.