Wagner's Tannhäuser (and some chatting about alternative tuning)

It’s not far off. Consider how well NotePerformer already does! Now imagine those excellent algorithms paired with much higher-quality samples…

Honestly, NP is so close in many ways. The slow strings are painful, and the balance isn’t the best (winds are too loud by default). But it’s a brilliant concept, and proves that such playback is already possible. There are plenty of scenarios where the NP default playback is so pleasing that, even if a listener might be able to tell it’s not real, it’s still so good that it’s completely undistracting. Fast passages in particular.

I’m not talking about realism that can fool the listener… yet. But maybe not far off.

Of course higher-level interpretations will always need human shaping. But perhaps Dorico will someday include some “macro” functions, like the OneKnob plugin, where a user can simply shape phrases within whole sections, or ride some sort of fader for slight tempo alterations. A big, simple control that performs a bunch of simultaneous functions, but hidden behind a simple UI. That’s pretty much the majority of “interpretation” at that point.

1 Like

What? The internal balance of parts is appalling!
David

to be fair, Noteperformer is one of the most useful around for letting you hear everything that’s in the score as it doesn’t really blend. It’s best for crisp, fast and strongly articulated music and is completely hopeless in things like this Wagner. I’m actually quite impressed by the latest VSL rendering – it probably goes beyond what I’d be able to do in Dorico or at least be willing to spend the time on.

I’m not talking about balance, as that’s fairly easy to adjust. I mean the interpretations of the phrases. NP brings life to the lines in a really compelling way. That’s what’s nearly impossible to duplicate in other, more expensive sample libraries.

I confess I didn’t listen to the samples posted above; I was speaking generally. As dko says, sometimes NP is great. Sometimes it’s painful. I’m just saying there’s real opportunity there.

Derrek,

My guess, also. AI will probably help us go a little nearer to the final result, but then there are so many variables, to make a computer sound as a real musician, that there will likely always be some finishing work to do.

Paolo

If I understand it correctly, Hermode Tuning tries to keep the bass in the currently selected tuning system, and then changes the intervals inside the chords. I don’t know what it does with polyphonic voices, or delayed chordal notes.

Paolo

I think you might agree with me if you were to listen to the examples. I, on the other hand, have never listened to realisations of other pieces by NP; but I think it has met its Waterloo in these Wagner extracts, which require lines that follow through and relate to each other.
David

That would certainly work for some styles of music, in that it would presumably result in beatless harmonies. But it would also make some of the melodic intervals of the upper parts very strange!
David

The example I saw mostly involved making the thirds pure based off of the bass note, but it used intelligent chord recognition to adjust other intervals as the chord became more complex, but it was always starting from a bass note and a tempered scale so that overall pitch wouldn’t shift. Also, if I recall correctly (I can’t find the vid and it’s been over a year since I saw the demo) the inner voices dynamically changed if certain voices were held and the other intervals shifted around it. Because a computer is doing it, it can happen so fast (and it’s only shifting pitches a few cents anyway) your ear can’t necessarily catch it.

Regarding the latter point, I’m sure his choir DID drift if no instrumental support was used. It’s human nature through and through.

I think the other thing it could do was make it sound like you were playing in a particular historical temperament but it would always change the chord to make it sound like it was tuned to that key… As in, you could play quarter comma meantone but when you hit a B major chord it wouldn’t sound out of tune and sour, but rather as if the meantone was based on B instead of C? Something like that.

From what I can infer from the original article on the HT, I would think that melody is kept as much as possible near to ET, by adjusting the other intervals of the chord.

So, it seems to be a balance between keeping the bass and the melody as near to the ET as possible, trying to find a mean tone between them…

https://sethares.engr.wisc.edu/paperspdf/hermode.pdf

Paolo

No, Romanos. It’s easier than that: all you have to do is make sure you have a d-sharp which is a pure third above the b, and not an e-flat, which would be sharper. That’s what I sometimes do with my harpsichords if I am playing in E minor. (E.g. the second of Pachelbel’S Hexachordum Pieces.)
David

Paolo, I might be wrong but, please check bar 42 1st. Violins. I think there is an a# missing. In your recording at around 2:28

Congratulation for your ear! Yes, there was a # missing! Thank you!

Paolo

@Paolo_T I don’t suppose you’d share the notes as XML?

I really enjoy trying to match your results.

Let me go on a bit with the piece, and I’ll see if I can overcome my natural reluctance to show my musical underwear :slight_smile:

Paolo

FWIW, here’s an mp3 of Orlando Gibbons’ Silver Swan using a brass ensemble in quarter comma mean-tone tuning. I ported the notation file over from Finale, where I used the Aria Player to set the tuning. I had to use the Aria Player in Dorico to achieve the same results but I’d be curious how this could be done using NotePerformer.

2 Likes

In Dorico you can fine-tune the scale. So, you can program any meantone tuning system you want. NotePerformer should be able to respond to microtuning from Dorico (it responds to quarter-tone accidentals).

If you like, I can try.

Paolo

David, do you mean that notes are too stiff, and don’t “inflate” to connect to the following note? If it’s this, there is a bit of this type of programming in my mockup, but apparently not enough. I was planning to make them stronger. If you meant this, I will for sure work on this.

Paolo

That’s a nice offer but I’d rather try to figure out how to do it myself. A google search took me to the Edit Tonality System dialog article, but this didn’t explain how to ‘detune’ specific notes, just how to set the number of subdivisions between each interval and to create custom accidentals, neither of which I’m interested in.