Hello everyone, for those who are users of note performer 4 and dorico 5, now with the improvements of dorico playback do you think it can replace note performer?
No
Jesper
5 Likes
Agree with @jesele in a sense but I think whatās going to happen here is this will be much like the key editor last year. The first release was a toe in the door, then the updates substantially filled out the capabilities. Daniel said as much in the blog.
So by the end of the year weāll have more algorithmic humanization, for example you can imagine phrase recognition which puts pauses (shortens the last note) between phrases. OK, how close is this to NP? Weāll find out but Iāll opine that it will not be too far off, and the key is that you can edit what itās doing, unlike NP. Which ultimately is much more powerful than NP.
But starting out on a project I donāt want to deal with humanizing at all but do want good temps, so NP is there for me. Later I will switch it out and rely on Dorico getting me say 90% of the way. And there are others who wonāt want or need to mess with it at all for which NP is ideal.
So really we have a suite of tools here that appear to address all use cases. For those of us that need more sophisticated mixing, we will simply export the full MIDI out of Dorico to use in a DAW.
3 Likes
Not even close for now. I tried it with vsl syzd special edition plus und Halion Symphonic Orchestra. Interestingly I liked the halion much better with the stuff I tried it on.
Still I think, the potential is huge. In the VI forum Daniel S. told me, that Dorico already looks 4 - 5 Seconds ahead to make phrasing decisions (instead of one second in NP) and could even look at the whole score. But for now it doesnāt touch the envelope of longer notes. And the team isnāt convinced yet to implement it. But if they do and maybe make a custom version of iconica, that could be mind blowing. Also I See sample Modeling work great because of the perfect consistency of the sounds. And even an advantage over np could be, that you could manipulate every cc to your liking.
But using the playback of Dorico today makes me appreciate NP and NPPE even more. Arne Wallander is a genius
1 Like
I like your optimism about what weāll be possible by the end of the year with playbackā¦
1 Like
Obviously with sunglasses onā¦
1 Like
I wouldnāt call it optimism but just a most likely projection. Daniel has said theyāll be doing a lot more in this with updates, and just basing it off what we saw before itās not a great leap.
tl/dr
Note Performer has this āAI techā so letās dissect that (disclaimer; I work professionally in AI and have been in that game for 30 years, which has been a lot of interesting history, but know nothing about NP internals and this could be seriously off base). First as famously noted decades ago what is AI today is bread and butter tomorrow. Itās kind of a meaningless term, because all it means is a new algorithm that quickly becomes an old, standard algorithm (the A* algorithm is so old now itās moldy, but at the time it heralded a revolution in AI. If you play video games your bad guys are moving towards you using A*).
Meanwhile Daniel used the words āacademicā and āalgorithmicā to describe what theyāre doing which sounds accurate to me. So back to NP - are they using modern neural net learning as to how to play with feeling? It could be done, you can take scores, and take recordings and train a DNN on it. Is there enough training data? Thatās always the big question ā¦ Iād say maybe, but at any rate NP has been around longer than modern NN so I think itās doubtful itās using anything like that. And weād probably see evidence of spectacular failures if it did, and since AFAIK thereās no reports of that then probably not (a feature of NNās and that kind of AI is when they fail they often fail hard).
So NP probably has various classical programming algorithms to do itās job, with copious amounts of hand tweaking which is traditionally a failure for say, agent based AIās (e.g. ALICE) , but here would probably work well.
Long story short it seems a reasonable engineering problem for Dorico to do a lot more with humanization.
2 Likes
I apologize for hijacking the thread, but what would be the best investment: Upgrading from Dorico 4 to Dorico 5 or purchasing Noteperformer 4? I will be using the BBC Orchestra for playback (or Spitfire SO/EW Opus) and realistic playback is a priority.
Of course upgrading to NP 4 is free if one owns NP 3. I would do that at any rate. If the decision to add one or more VST engines to that is present, I would probably add the engine(s) first unless one sees a feature in Dorico 5 that would significantly improve workflow. One can always save up and upgrade to Dorico 5 later in the year.
That said, nothing is better than upgrading both, but I realize that is not something everyone feels he or she can afford to do.
NP, even if you donāt already own NP3. Playback in D5 is a step in the right direction, but not even close to what NP4 with NPPE offers right out of the box.
1 Like
Agreed. I find D5 to not be worth the upgrade for me at present. NPPE does better. That plus I am an arranger and do not use drum machines.
The biggest thing with Note Performer is that it just works. No futzing around with expression maps, or loading in dozens (or hundreds) of GB of samples.
Issues:
1.) Live Stage dose not appear to work with Note Performer - no instruments appear and neither instruments or ensemble can be added
2.) Groove Agent us great for producing great sounding drum parts, but not for producing written drum parts. Especially in swing charts when dragging a square patter icon to Dorico, you get what was actually played neaning loads of erroneous triplets. Nobody write swing that way. Therefore using GA in swing charts is not viable unless you only care about playback and not whether itās actually readable at all. So Two things with great potential, but apparently not ready for prime time.
Iām not sure what the actual comparison is supposed to be here. If you compare just about anything current to a library like Halion (or the stock libraries in Sibelius or Finale), then itās an improvement ā a 4-5 second lookahead cannot mask the rather dated samples and programming in the first place. And NotePerfomer itself, as Arne himself admitted, reached the end of the road in terms of possible tonal improvements he could do with it which is why he focussed on using NP intelligence in more detailed sampled libraries.
The real question is whether the new D5 playback features can work together with NPPE to give even better results or whether you must just follow NPās recommended defaults which were obviously created for Dorico 4? There are discussions on this going on elsewhere.