Real Orchestra vs Virtual Instrument Mockups

orchestra-vs-synth-mockupsAre virtual instruments getting good enough to replace a traditional orchestra?

That’s one of the questions raised by an interesting series of posts at sound design site AudioCookbook.

UK composer Tom Player has been sharing a series of examples, looking at synth mockups of his compositions and comparing them to the final recordings with real orchestras.

“I hear lots of sample mockups, and undoubtedly they are getting better – but there really is something irreplaceable about a live orchestral performance, which to me will always be preferable,” notes Player. “For one, there is the sound – but it stretches further than this, with the psychology of writing better music once you know it’s going to be played live, and so many other factors.”

His first example, below, compares two sections from his composition Beckoning. With each section, the example features first the virtual instrument arrangement and then the orchestral version.

Here’s what Player has to say about his first example:

It’s a simple line, recorded in isolation. This is probably one of the hardest tasks for a convincing midi demo… simplicity in isolation. In the real recording, the orchestra breathes as one. Literally – you can hear the breaths as the leaders of each section signal to the rest of the players to come in.

Interestingly, there’s some debate at the site, suggesting that Players comparisons aren’t valid, because his virtual orchestrations are primarily intended to be functional representations of the music, while the orchestral version are intended to be expert-level performances.

Reader Pedro Almeida shared an example of a more expressive virtual orchestration of the same theme:


The series does more than just compare live orchestra vs virtual instruments. It highlights the state of virtual orchestration and makes the case that an expressive performance, whether with ‘real’ or ‘synthetic’ instruments, transforms notes into music.

This example is just the first of a planned series of six orchestra vs virtual instrument posts. See the AudioCookbook site for parts 2, 3, 4 & 5.

Check it out and let us know what you think of it!

via John Keston

24 thoughts on “Real Orchestra vs Virtual Instrument Mockups

  1. It depends on what you mean by “replace.” The major expense of hiring an orchestra has a lot to do with the proliferation of orchestral plug-ins. That includes fairly substantial practice time, as even the best players need what they need. A real orchestra will always have the edge in sound quality, but should any of us give up a massive part of the instrumental field out of a sense of misplaced purity? Its either a plug-in or nothing. If you can afford to hire an orchestra, your areas of real-world concern are different from Synthtopia’s anyway. There is also the matter of being well-versed enough to fully APPLY Vienna Symphony. If you are that capable, this debate barely includes you! Its more about those of us who will never write a full symphony, but who understand the instruments well enough to apply them within their normal ranges. Its the organic version of not turning a decent sample into a chunk of BLORT by pushing it outside its range until it honks. I’m not about to pull a full Tomita, but I’m a big fan of Jerry Goldsmith’s blending of orchestra and synths. If you’re not aiming for the highest purely orchestral realism you can manage, the next level is understanding that a French horn sample from a dedicated orchestral set won’t be as full as the real thing, but it’ll generally beat the pants off a workstation ROM sample. For me, its never about replacing the orchestra. Its about getting the Real Orch World and Synth World to highlight one another’s pluses. I love me some cello trio with a juicy sawtooth foundation or a tubular bell with a touch of Prophet-VS single-cycle bell wave to give it a bit more presence. Check it out, yo.

  2. It’s already hard enough for musicians to hear the difference, let alone for the average movie goer, who has more interest in the actors. They have absolutely no clue.

  3. Interesting comparison..but highlights what I was already thinking when I hear about vst orchestra vs real one. A real one will always sound much more full, given every single small modulation in the pitch between EACH instrument. It’s like when you combine similar waveforms on a synth, but with each its own little detuning.

    Also, I still have yet to find a convincing sample library/modelling plugin for wind instruments, especially brass sections. Even the ones who are several hundreds or thousand dollars worth sound awfully fake and dull to me, which has to do, most of the time, with the attack of each note in relationship with the breath and air pressure/support. It’s interesting how very “complicated” instruments plugins like the piano have achieved a very high level of realism, but simpler, almost archaic instruments like the horn and the clarinet don’t have convincing counterparts in the digital world.

  4. Wow, differences are completely there, even with a digitally compressed recording. The virtual arrangement sounds completely linear, the real one has wider low frequencies that add depth, and you can also hear the strings natural vibrato and the brass “blow” in the attack on the real orchestra. Emulation has come a long way though but listen to the rest of the examples linked in the article. I’d say there is a huge difference in articulation rather than timbre itself, the real orchestra sounds a lot more alive.

  5. Is an artificial human better than a real one? I don’t think it’s a question of better, because the answer to that is obvious. The purpose behind it depends, and it’s subjective for everyone. Now, who really has that budget?

  6. There are MANY factors that go into the performance of even chorale passages like that. It is clear that the live performances sound richer. However, with a great degree of painstaking work, great results can be gotten from VI’s, especially with use of CC’s (like breath controllers and sliders) that continually affect dynamics, timbre and vibrato. As others have suggested the gap between real and virtual has narrowed considerably.

    I would disagree with one point that Player makes: “It’s a simple line, recorded in isolation. This is probably one of the hardest tasks for a convincing midi demo… simplicity in isolation.”

    I get his point, but for me, orchestral VI’s (apart from percussion) sound most compromised when playing faster more complicated melodic lines.

    As an aside, I remember back when Garageband sold “Jam-Pack” CD-Rs, and I got the orchestral one. I was disappointed to hear that MOST of the sampled phrases/loops were obviously made with a sampler. Pretty crappy.

    The concept of a MIDI mock-up is a great way for a composer and/or orchestrator to do their work. If the final product is never played by humans, that is often a schedule & budget consideration. It doesn’t change the fact that this work is about creating good content. That can even include pushing edges of playability (when it might be called for).

  7. Virtual Orchestras have come a long way, and I agree, for low budget films, they are pretty much passable now, and might fool most of the audience, those not fooled, may even give a positive pass if it’s handled well. But listening to these recordings, there is a certain detail to the actual orchestra that the virtual ones lack, one can really heard those horse hairs creating friction across the strings, while the first virtual orchestra sounds almost like a synth, not quite, but almost. However, as more instruments are added, it becomes slightly more murky. The cellos at the end of the second virtual orchestra were impressive to my ears. But there is that “life” quality that a real orchestra has, it feels alive and even a convincing virtual orchestra seems to lack that.

  8. I have heard some amazing choral and string synthesis performances, but they always depend on the synthesist recording dozens of individual monophonic performances, rather than ‘playing chords’.

    It seems that most people just don’t want to put in the work it takes to make great synth orchestrations – they want the virtual instrument to do it for them.

  9. Sample based digital constructions of orchestras are at a level where they sound really good, and I’m always happy to hear them be part of a trance or ballad tune, or wherever they find themselves. However, if anyone has actually heard a orchestra live, or even listening closely to the example there are things that are just not possible to do with samples, at least not easily. Particularly things like unison vibrato and tremolo sound much more realistic and the vibrato can have a more exponential oscillation than a more likely linear programmed one. Also string players (and other fretless instruments), like singers are very likely to play closer to Just Intonation. They seek those most perfect of harmonies fairly naturally because it’s where the oscillation of the pitch differences (beating) in a harmony are the smoothest and consequently best sounding. This is sort of a form of hermode tuning because it realys on a dynamic interpretation of what scale they are currently playing in. This is simply not possible given current technology to emulate these characteristics properly.

    Perhaps someday when physical modelling is remarkably complex and sounding amazing but it’s along way off. And that human dynamic element we hear as feeling will always sound better with people playing it.

    In short, samples are great but you can’t beat the real thing.

  10. Synthesized orchestras are like any other machine. They’re either a benefit or a hazard. If they’re a benefit, it’s not my problem.

  11. the more you listen to what is lacking in a strings patch, you can add in a lot of the details depending on your synth. i have a korg z1 for example that, combined with a sampled attack, can add a lot of the textures and variability through mod sources. the real trick is taking the time to set up your mod sources on your controllers (velocity aftertouch foot controllers mod wheel etc) to the right destinations in the synth’s pitch and tone, to get what you personally want. not photorealistic strings still but a lot of the human character can be re-inserted this way. takes a lot of time to learn exactly what it is you personally want to insert and why though.

  12. I feel like the virtual orchestra lacks the soul and liveliness that the real orchestra has. The virtual orchestra seems more on the grid and has a rigid tonality that almost seems calculated, whereas the players in the real orchestra have a groove and feeling are capable of creating slower attacks and movement with their instruments.
    Seeing as hiring a real orchestra or finding people in your area that can play orchestral instruments well is costly and not time effective, I would use virtual instruments in my compositions, but I would also use modulation and automation to create a faux liveliness. You aren’t cheating anyone by using virtual instruments; you’re just doing what is more cost effective to achieve the sound you want.

  13. I think the actual point here isn’t that “real” instruments sound better than sampled, but that orchestra instruments in general, while absolutely fabulous, have had their day in the sun and need to get in the toolbox as equal members with every other sound. Nothing brings me out of a good film these days faster than hearing the same swells and stabs and bits that I’ve heard for decades. Seriously, we have a huge sonic pallet now (more than just glitch and noise too, thank you) and it’s time more film composers started embracing them fully, rather than relying on “the way things have always been”.

    Debating the merits of real vs sampled is just a resonant peak… the last dying gasp of a way of thinking that is about to go away.

  14. Each individual live musician interpreted the piece so slightly differently, which created a more interesting performance. If he built a sampled string section from individual instruments (rather than one section) and varied each instrument’s expression slightly, it would be a more equal comparison.

  15. Comparing a live orchestra to a recording of a digital orchestra is an absurd and meaningless comparison. A useful and fair comparison is that between a recording of a live orchestra and a recording of a virtual one. Why? Because any life performance is a different art than a recorded art. It’s like comparing a photograph to a painting, or a live play to a film–again, these are different mediums and a comparison is not very useful.

    Whether a MIDI recording sounds expressive, nuanced, detailed, subtle, with tempo variation, attack-release variation, articulation variation, etc. depends upon the talent and skill of the composer/orchestrator/producer. Every DAW with a notation editor is capable of a very high level of musical expression and detail, assuming that one has access to a high-quality orchestral library (I use VSL Orchestral Cube).

    Here’s an article I wrote published in Keyboard magazine about this:

    And here’s an example of what a virtual orchestra recording can sound like when one is not thinking of it as merely a “mock-up”. That word immediately implies: fake, cheap, second hand, rushed, etc. Which is why I think the term does an injustice to the medium. (1st two movements only)

    Jerry Gerber

  16. One thing that may be missed is with a live performance there is imperfections, slight out of tune, slight off time, varying dynamics, and the echo characteristics of the hall they are in.
    Now with the virtual you have one person doing it, varying is no longer there.
    One thing to do is to have a bunch of keyboard players play in to speakers next to them on a stage and record that to see what you got.
    Another to look at is the piano, the sampler on them are huge, one reason is that a single note is a chord, playing a one huge spring reverb.
    Personally , I think talent has more to do with it and John Lennon said “get me a freakin tuba and get you something out of it”

    1. Actually, that’s not true. With an advanced library like VSL, you can program in imperfections of pitch (random fluctuations of cents on attacks or longer) and there are many MIDI sequencing techniques to introduce tempo fluctuations, variations in dynamics, attack/release fluctuations, timing fluctuations such as having a note or passage begin late or early, etc. etc.

      But even with such capacity, the comparison between a live performance and a virtual orchestra recording is a meaningless comparison. Comparing a RECORDING of a live ensemble with a RECORDING of a virtual one–now that is a worthy comparison.


  17. This is a truly interesting discussion!
    I’ve read all the comments and the article. For sure orchestral libraries are getting better and in some situations it is impossible to recognize,either it is a real orchestra or the samples. On the other hand, it is obvious that the real orchestra recording is somehow on a different level of music. The problem is that, such recording is quite expensive.
    My solution for that problem is idea, to mix sample libraries, with real recordings. It is much cheaper, and the result is better, than samples alone. It works pretty well when you cover the violin, or brass section with three live players. Moreover For solo lines regular recording is irreplaceable ( in my opinion)
    Interesting thing about our technic is that we record each player separately. It gives a lot of mixing space then, and allow you to ” animate” each musician ( which was one of the issues in the comments)

    check out or

    greetings to all of you!

    Wojtek Slawinski

  18. A better test would have been to do these as AB blind testing; by indicating ahead of time which is which, confirmation bias can not be eliminated. I also note this article is now 7 years old, and clearly technology has continued to progress and simulation is much more sophisticated.

Leave a Reply to skunkCancel reply