[...] Your last letter made me very glad, although I was horrified at first to discover that you had not before very precisely grasped the difficulties of my Tannhäuser. [...] Difficult as the conductor's task is, I would remind you that that of the stage-manager is no less so. Here, too, after all, the music ultimately counts for most—but its effect will not be the right one if it does not accord very closely with the action and the whole presentation, that is to say, if the action and presentation are not so arranged that the music appears to be fully justified and explained by them. If you will take the trouble, for example, to follow my orchestration closely, you will [not fail to bear] me witness that it contains no striking feature whatsoever which is not definitely connected with what is presented on the stage to the eye of the audience, whether this be the scene itself, an action, or even a mere gesture. The stage-manager must therefore get to know the score very thoroughly, and it is really best that the conductor should also be the stage-manager and from the first acquaint the singers most exactly with all the meanings indicated in the score.
Speaking of the singers in particular, during rehearsals for my latest opera I found that if from the first I allowed them freedom in the singing of recitative passages they slipped back involuntarily into the old recitative manner, in which they dwell on certain notes, gallop, drawl or hurry others at will, and in that way the verisimilitude of my declamation is obliterated. On the other hand I made sure that—carrying the actual production in my mind's ear—I have indicated even the rhythm of the quasi-recitative exactly as most good singers would have worked it out for themselves from the old-fashioned recitative, and I now assert with conviction that my recitatives (if they are to be called so) can best be performed if they are sung exactly according to the value of the notes and in tempo. [...] So let the singers learn them accurately, according to the value of the notes and in tempo, and when they have thereby made the right accentuation completely their own, then (and not till then) they can be left somewhat freer; and then I willingly agree that the conductor must go with the singer.
I would not wish you to make any further cuts, and this is really not from vanity but because I should be obliged to deduce therefrom that the performers were thought radically unequal to their task. For instance, if those who take the parts of Tannhäuser and Venus cannot present the psychological development in Act I. Scene i. with such truth that the scene holds and thrills the audience throughout, and if from the first there is so little hope of their being able to do so that it is felt necessary to shorten the scene, I tell you plainly that it would be better not to produce the opera at all, until these doubts are overcome and reliable performers found. Only in case of extreme need, and then most unwillingly, would I consent that Tannhäuser should sing the first verse of his song in D major (instead of D flat major) and then (after the second verse) continue in D minor. [...]
You will soon discover one thing [...] that this opera particularly can only be effective and successful when it is effective and successful in quite another sense than is the case with the ordinary opera. Even little places (like Weimar and Schwerin) have, however, taught me that what I wish is possible. [...]
Source: Letters of Richard Wagner, Vol. 1, selected and ed. Wilhelm Altmann, trans. M. M. Bozman (J. M. Dent & Sons, 1927), No. 220