The perfect storm? Dorico, high-end VSTs and NotePerformer.

Daydreaming a bit, I could easily imagine a scenario where the developers of Dorico, Spitfire (for example) or Cinematic Studios, and NotePerformer would join forces in order to create the industry’s ultimate music creation suite, with seamless integration of expression maps, etc. Dorico’s ease of use, Spitfire’s high quality samples, and NotePerformers smart playback could stir up the perfect storm. Of course the price of the end-product would go up considerably, but I’m sure many would be willing to pay for the ultimate. For those of smaller means, you could offer various “simplified” versions.

Wouldn’t it make perfect business sense also to offer something like the above?

Steinberg introduced its own high-end orchestral samples last year with Iconica and there have been suggestions (I believe from Daniel) that Dorico expression maps will eventually be available. I suspect this is the direction in which your daydreams are likely to bear fruit.

Nice to hear that. If Steinberg additionally can come up with something like NotePerformer, then we’re all set. :smiley:

Sample libraries are last century’s technology. Once you start to hear the flaws in a few samples (and the bigger the library, the less likely it is that everything is “perfect” the only way to get rid of the annoyance is buy another sample library.

There’s a reason why NotePerformer doesn’t use samples, but companies that have spent decades developing sample libraries aren’t going to stop, so long as people keep buying them in the hope that eventually they will find what they are looking for.

Rob, I agree that sample libraries are in a sense dinosaurs. They will eventually be replaced by something better; most probably a combination of NotePerformer-like AI and physically deep-modelled instruments (like Modartt’s Pianoteq and Organteq). But until that kind of technology is fully mature (which might take a while), one is stuck with Spitfire and the like, if one wants any level of realism in sound.

You should look at StaffPad’s playback nowadays. They reconfigured premium libraries for the playback engine…

I’ve heard samples. I wasn’t very impressed. Everything sounded like a really cheezy/low budget movie trailer. Everything was super exaggerated.

The solution is already available, but very expensive. It’s called “live players.” :sunglasses:

I’m a great admirer of Pianoteq – it works nicely in recordings and is easy to play. There is still a certain something missing in terms of range and pure individuality and realism of tone which the best sample libraries still possess and this is more so when comparing Note Performer with the top sample libraries as piano seems the easiest to get right. NP gives a wonderfully vivid rendering of symphonic music (I’ve used it myself for the bulk of mine) but the more focus there is on individual instruments, the more the shortcomings become obvious. I’d say we are some way off yet before sample libraries become obsolete. And we shouldn’t forget that sample libraries are becoming gradually also more sophisticated in automation and musical intelligence though I’d still put NP on top in this department.

There are different levels of samples in terms of quality. Something like Garritan is not going to come close to more expensive libraries like Spitfire in terms of quality, and will sound cheesy etc. You can get realistic enough results with high end libraries like that, especially if you have solid mixing and orchestration chops.

NotePerformer cannot match the results of good sample libraries in a DAW. I see lots of people commenting on NotePerformer demo videos “NotePerformer sounds so real, like a real orchestra!” but if you put it alongside a mockup done with good sample libraries, the mockup will win by far every time. What this means is if that if two people do demos for scoring say a short film and one does it with notation and NotePerformer and the other with high quality sample libraries, the one who does the demo with the high quality sample libraries will win hands down every time and get the job.

In my experience, the problem with the high quality sample libraries, why you can’t just plug them in to notation software and have them sound good, is twofold. First, there are balance issues which NotePerformer automatically adjusts for but the sample libraries do not. Second, NotePerformer is able to somewhat intelligently look at the material and decide how to alter its performance based on that, while sample libraries will just march on doing what they are told. The result from plugging even a high quality sample library into a notation program will often be sub-par, unless you do a ton of manual CC work and humanization of attack. Although the sound itself may be more realistic than what NotePerformer can do, balance issues plus unsuitability of selected articulation for certain passages means that the overall result sounds a bit odd and unrealistic. Mixing different libraries together from different vendors can also be more problematic in notation software in general because the mixers are frequently less capable than those included in DAWs.

I had tried a few times in Finale and Sibelius (back when I used those) to plug third party libraries in and was very disappointed with the results, paling in comparison to the results you get by default with NotePerformer. I haven’t tried in Dorico, and I expect the new Dorico playback features can address some of that but the balance issues can still be problematic. But, for me, if I want a polished product, I’ll do it in a DAW. If I’m working in notation, I’ve usually already done a DAW mockup up or I don’t need a mockup at all, and then NotePerformer is just fine even though it can be fake sounding, since it gives a very good general idea of how things sound and the balance.

“Every time?” :laughing:
I’ll admit that you may have a point in general, but even if the two musicians are equally adept, the one with better musical ideas may have an advantage even if he or she uses NotePerformer for the demo.

Very unlikely. In my experience, you would have to pit a very poor and unskilled composer with great samples against one who is much much better and only uses NotePerformer to see such a result. Directors are often worried more about the overall polish of the finished product rather than how brilliant the music is on its own. Any sort of melodic material is often shunned for worry it will distract from the film itself, and so there is often a major use of simple textural pads, ostinati, and other such things that can fade into the background and not distract, often with added synthesizers and loops etc. The mixing gurus who are OK composers who have a ton of great libraries and can make them sing pull off this stuff perfectly and deliver a polished product. Composers who work in notation and NotePerformer would have an exceedingly difficult time getting this sort of end result.

Only if you actually have samples for what you need. If you like to utilize extended techniques (Or even not very extended techniques, like mutes) NP blows away any sample library by actually having the sounds. Even the massive full fat Vienna Symphonic, with over a TB of samples, doen’t have for instance, muted Trombone glisses. Even the non-muted glisses are of a fixed duration.

Oh of course, but when people are writing for the samples (ex. film composers), they make do with what they have. If they don’t have samples for an effect, they don’t write that effect. There are sometimes ways around it, ex. using pitchbend-related stuff to create the gliss. Obviously if you are writing something that an actual ensemble will perform, NotePerformer gives you options other things do not. I use Sample Modeling for brass writing because it can do all articulations with all mutes - I don’t like to be limited. But it would be hard to include some NotePerformer sound mixed in with good samples in the same track - unless the NotePerformer sample was particularly strong for NotePerformer or it was an instrument that didn’t really stand out.

There are many problems with sample libraries vs NotePerformer (which does use samples by the way, but combines them with some modelling / processing). Here are a few.

  1. Dynamics. Your Celli may be ff but they don’t match with the violins playing ff. Or the celli arco ff doesn’t match the celli pizz ff. Mix and match with any dynamic and instrument. The sample dynamics are not normalized, you have to do this yourself. NotePerformer goes to great lengths to make this work, among all sections (even brass vs strings vs woods vs percussion).

  2. Divisi. Forget decent divisi with sample libraries. I have several that advertise support for divisi. Divisi doesn’t just mean you can split a string section in 2. You may need to split it in 3. Or 4. NotePerformer does this. With sample libraries you have to fudge it.

  3. Legato and articulations. Forget it. Legato in sample libraries either means using special legato processing, which always sounds weird to me, doesn’t handle multiple notes often (solo line only). With NotePerformer you just put a staccato dot, or a slur, or a tenuto mark or an accent and it plays it. With sample libraries it’s often a key switch to another patch, with wrong dynamic which you need to tweak.

  4. Missing patches. So many sample library vendors don’t seem to think it’s possible to play pizz while muted. Or any other articulation while muted.

  5. Baked in artefacts. Many sample libraries I have have a baked in swell in the long patches. Also baked in dynamic eq.

So yes you can get a good result with sample libraries, but you

a) need to have a heap of them, and they cost a packet
b) need to spend an enormous amount of time battling the 5 things I listed above

NotePerformer allows a composer to just focus on the music. You want to change something from pizz to arco? No problem. Staccato to normal? No problem. No key switch or dynamic changes or fudging expression lanes required.

This is one of the ways in which sample libraries influence the composition of composers. People end up compromising their composition to reduce the pain associated with dealing with the sample library.

In the end, I think there needs to be some other way than VST interface to do this. the VST interface is an encapsulation of MIDI. MIDI is note on note off. There’s no way in MIDI to tell the synthesizer what is coming NEXT AFTER the note that is playing. No way to tell it there is a break or a gap which is why NotePerformer need 1s delay. The end of the note affects what end sample or modelling is used. To get realistic automatic synthesis from notation I think it will require playing ahead. This is incompatible with MIDI-keyboard driven synthesis.

I’m not criticizing NotePerformer. I use it extensively and recommend it to everybody as an add-on that greatly improves playback in notation software. But it has a very different use case - giving you a good idea of what it sounds like and the balance, rather than providing a finished polished product. I don’t bother loading up samples into Dorico or Sibelius before that, I just use NotePerformer. If I am writing a concert piece and no mockup is needed I just use NotePerformer.

Obviously working with samples is a huge amount of work and you have to add all the shaping yourself manually and manually try to work around sample weirdness where the recording has artifacts or there is a sudden jump in dynamic, but for media composers there is really no way around that.

Don’t forget that NotePerformer is pretty much the first of its kind. Give the technology another 5 years, and see how far might improve. The first release of Pianoteq was pretty weedy compared with the current version 6 (which has an instrument model endorsed by Steinway)

The speed of improvements to NP would probably increase if there were some competitor products, of course!

I have PianoTeq 5 and 6, and can’t use either in any film projects. It sounds too fake/weird. I continue to use older piano sample libraries like Art Vista Malmsjo when I have to write exposed piano parts. I’ve tried to use PianoTeq 5 and 6 and the results just were not anywhere near as good, and the directors of the short films that I composed for heard that as well.

I would love for products like NotePerformer to eventually be able to pull off realistic sounding playback at the same level as high quality samples. When I’m writing for media at the moment I have to be both a composer and every single individual performer, and shape every line like the performer would manually for every instrument. It is a lot of extra work vs. just writing, and I would rather not have to do it manually.

Part of the problem is getting the computer to shape the performance in an intelligent, musical way, since the notation does not necessarily convey everything. Even some things that you can write into notation with text, NotePerformer can’t necessarily interpret it the same way as a human performer, ex. markings like “with great sorrow q=80” it would just see the q=80 and ignore the “with great sorrow”. At least if it gets in the ballpark and you can make manual edits to controllers to get the desired effect, that is better than what we have now.

NotePerformer is remarkable for what it is, but myself I could never use most of it for a final recording, its just not good enough to be that…nor is it intended to be. Its the best I have ever heard though in terms of fairly automatic rendering from a notation program, good enough to get an idea about the composition. Also some attention to level balancing means you can somewhat trust that the orchestration will be close to what note performer is representing, compared to using sample libraries where you get so deep into the weeds that you can have no trust whatsoever that the level balance between the sample libraries is accurate enough to verify that your orchestration is pretty sound, or close to it anyway.

Yes, I agree 100%