I’m afraid to say that this is a parody rather than a test. If the assumption is that absolutely no time whatsoever should be spent in getting to learn the libraries and how to do the initial setup then NotePerformer is probably indeed a good idea though this sound is generally rather crude and unpleasant imo. If people know that their orchestral works have a ready professional orchestra just waiting to play them then they are lucky indeed – not many have this option. On this assumption, your test has proved its point but I simply can’t agree with the premise.
None of the others bear any resemblance to what I would expect from these libraries or results I can easily obtain from VSL or the BBC – and there are far better programmers out there than me. If you want to use them, you need to know the fundamentals of programming them. That doesn’t mean a huge amount of time has to be spent while actually composing but it does mean putting in the groundwork for understanding how all the dynamic controllers are used and appropriate choices made in the Expression maps for instance. It is absolutely essential to disable all Dorico dynamic humanisation (other than accents) in Playback options as this tends to play havoc with any half-decent VST which has its own humanisation built in.
Perhaps you could supply the actual Dorico file, then one or two might be tempted to use one of the other libraries to show it’s really not hard to improve on NP for this particular example. As others have mentioned in the past, there doesn’t seem to be much sign of further development to NotePerformer which is indeed a shame as it has undoubted strengths which are to some extent, especially with the strings, rather undermined by the low level of sampling.
While I agree with you, that better results can certainly be achieved quickly with the other libraries, I don’t think @MarcLarcher 's premise was wrong:
He was thinking about buying Iconica, and seeing that every “heavier” Sample Library than NP will require at least some work. It’s absolutely reasonable for the libraries to require this work, but it’s also absolutely reasonable for writers to not want to input ANY work into their sounds, because of various reasons.
On the respective websites there are of course only good sounding examples, as well as there are many excellent examples here on the forum.
But a comparison of “total defaults” has its value, and it helped me still relying on good ol’ NP with all its problems, which does good enough for me, as I want to download a library, install it, use it and never think about it again.
I would maybe have gone to BBC or Spitfire, listen to their demos, be so impressed to buy them, use them and be disappointed as Marc was. It saved me valuable time and money, so thanks again to him!
if the premise is that no work at all should be put into the libraries the you will also get suboptimal results from NP but I’m entirely in agreement that the difference is far greater with most other libraries than NP. I tried to make the distinction between on the one hand being willing to invest a few hours in getting to know another library well enough to use if efficiently and on the other the time required while actually writing to make numerous adjustments in the PLAY windows or whatever.
In other words, libraries can be quite quick to work with – VSL often comes into that category – but take time to learn. If neither the initial learning curve, nor the actual time spent while writing or later polishing are at all acceptable, then yes, go with NotePerformer and I can then accept the premise of the test quite happily.
I’m sorry you feel that way.
I have no problem in providing you the files, btw. I tried to present the thing honestly. I forgot to add that, for BBCSO, I turned down dynamic humanization to 10% (I might have tried 0%, ok).
I know better results can be achieved, but that was not my point. It was only a kind of real-life investigation that give some of us some insight about how new libraries might sound without curating either expression maps, or the score itself — actually, both.
This Mozart score is very “simple” as far as notation symbols are concerned. Not many staccato markings, no marcatos, etc. Because the readers are supposed to know But the machines do not, unless they are taught to (as NotePerformer probably is).
What my “test” shows here is that there’s one instrument that is suited to people writing scores, who do not want to export beautiful files but who need to audio-proof their writing. I hope that the refinements the Dorico team brings to the Play mode will allow us to use different libraries with a higher efficiency.
My experience is that percussion instruments are quite well rendered, because of their very nature. I really love how Pianoteq behaves. And I spent many hours building up the Spitfire Percussion percussion map offered in this forum. Superior Drummer sounds very good and fulfills my needs in that drum field.
The hard work is on strings and winds (woods and brass), because of the way the sound can evolve through a single note, the various articulations that can happen at the start, during the note(s), at the end… and the transitions between different articulations. I understand that most libraries out there are not designed to be used the way we want, through a score. They are designed to be used in DAWs. But is it a parody to try and use them, just for the sake of it? Maybe. But it could be a useful one, at least today.
In any case, I thank you for the advice about the humanisation, and I’m always happy to get more information about how to make these libraries work better, so please, don’t refrain from sharing your intel with us!
+1 to this. For me, Dorico’s playback is a composition aid and a proofreading tool, not an end product. Sometimes I’ll send out a mockup to musicians before a rehearsal, but let’s be honest, most musicians aren’t going to listen to it anyway, LOL. For some of my copying clients, I’ll provide a mockup to assist with their own proofing and make sure I’ve done everything correctly, but working on this mockup is typically not billable time as it’s not an end product for them either. Better, more realistic sounds are of course desirable, but for those of us for which this is just a composition or proofreading aid, or unbillable work, finding libraries that can produce acceptable output with a minimum of fuss is definitely desirable!
This is a bit unfair, since the premise was clearly stated: on one side there is NotePerformer, on the other traditional sample libraries. Both with no additional efforts at all to make them sound good.
I can identify myself in the user being very disappointed when listening to the raw playback from the score and a traditional sample library. It is not how it should be presented to the listeners, but the contrast with the already clean and intelligible result with NotePerformer is stunning.
I think for those users who do spend time tweaking for playback, it may seem inconceivable that other users would demand “realistic” playback (whatever that means) without any extra work in play mode. But that’s exactly what they (we) want.
Honest admission: when I’m using notation, I don’t want to spend an hour learning a sample library. I want playback out of the box. Maybe I’ll install an expression map that someone else made, but I don’t want to do anything other than that. Is that unreasonable? Yeah, maybe it is. But I think that perspective represents a large percentage of users of notation software. And NotePerformer has shown us that’s not a pipe dream.
Hi Marc,
thanks for your understanding. I did get a bit angry when I heard these renderings as there just didn’t seem any point in doing them. This is perhaps because even the out of the box rendering is so bad, I even began to suspect some sabotage or something . I never got anything like this, even when I was new to such libraries.
But of course I perhaps should lighten up and see the funny side. I have never denied that NP is the best for doing basic audio-proofing --in fact Halion would do that as well but there’s no doubt that NP sounds more like music. I think your point that NotePerformer is taught how to interpret (also has a built-in delay allowing it to read ahead) how to read the music is actually not very different from other modern libraries. The difference is that with NP, all the work has been done for you and that’s not necessarily the case with the others. So in one sense it’s not a fair comparison and in another it is --all depends on how you look at it.
The question several members (and also some of the leading virtual instrument providers) have raised is whether people using notation software actually care about the best possible playback. Of course many couldn’t care less about playback at all. Others want the best possible mockup within reason because that’s how people will realistically get to know their music. There is no reason whatsoever why Dorico shouldn’t be used for that.
I think for the purpose of the exercise Note Performer is outstanding.
I am truly shocked by how horrible the rest of the libraries sound.
I personally have a hate/hate relationship with string sound libraries…Everything sounds terrible and a massive amount of work is always needed for me to get the desired effect…
When recording with a real orchestra, the sound in my head sounds exactly with what is produced by the players, when I work with sample libraries, everything feels like a wrestling match…
This morning I may have put an end to my work on the template for the VSL complete VI collection. I started working on a template for this library at the end of 2016.
Here you are. Mozart – Eine kleine Nachtmusik NotePerformer.dorico (992.9 KB)
Honestly, using poco vibrato for everyone at the beginning and even no vibrato for the basses and violas the vast majority of the time in this score will get even better results with NotePerformer. Quite simple to do, I deleted those markings to be fair to the other libraries (that would probably have me record the whole vibrato CC lane to get a convincing result.)
I wanted to make a Halion version but… I forgot
I guess I totally stopped using it like four years ago and… well, it slipped my mind.
Be sure that I did not want to sabotage anything. Tell me which library you own, and I’ll send you the used file. You’ll be able to check on your system. I’d be glad to know which parameters you’d change to drastically improve their sound quality.
as the BBC SO was probably the worst, I’ve gone with that just based on your posted score and my own version of the Core template. Other than putting in a few marcatos at the beginning for instance where I think they ought to be in the score, I have made no corrections to the playback but I did make the essential changes to neutralising all humanize features in the Playback options. It’s possibly a bit better than I’d expected, considering virtually no work. This is the result of a section:
my only real point here is to show that some knowledge of the library and how to set it up can make quite a difference to the result without doing any work on the piece itself. And knowing that also makes the corrections quicker. But let’s face it, we’d all love something that just works perfectly out of the box – deep down I suspect we’d all rather just write music than programme libraries.
. I never got anything like this, even when I was new to such libraries.