How to get the most out of mediocre instrument samples?

[Der_Winter]

I really want to invest in Symphobia or Embertone plugins at some point but can't afford it right now, I have Ableton suite though which includes basic orchestral samples (the solo instruments sound very fake though.) For piano I have the Addictive Keys Grand (came free with my soundcard) which is okay, and the Ableton grand which is about the same quality.

I've made a little demo here, any tips on how I can make the instruments sound juicier and more natural (besides playing with the velocities)? I'm struggling to get used to mixing/producing this sort of soundworld!

 

[Meteo Xavier]

You get around it by focusing on composition, arranging and mixing. Video game composer professionals have had to work with mediocre samples or worse for 20 years and that's still the most beloved era of videogame music, so sample quality is almost sort of a given that it's not what makes the quality of a song most of the time.

[JohnStacy]

Really, I do this intuitively as imagining how to make a band of bad players sound good.  Have to make the samples blend together and balance well, and also match style.  You'll be surprised how far that goes.  Really just work on production techniques and write in a way to take advantage of what you have.

I can elaborate more, but won't be able to until tomorrow.

[Garpocalypse]

Velocity and timing go a long way towards making an authentic sounding performance so don't overlook that.  The best sample libraries take the attack of each instrument into account whereas the cheaper libraries often have 0ms of attack across all of their samples.  i.e. you hit a note and get the result immediately without any startup.  Take your bass pizzicato's for example.  They are RIGHT ON THE BEAT which does not work well with an emotive piece of music like this.  A bassist has to anticipate the moment that the string has to be released in order to come in at the right time which means he has to take the physics of his own instrument into account as well as playing off of the emotion present at the moment in the music in order to sound appropriate. There's anticipation and some hesitation there with every pluck and your sequencing needs to reflect this. If you think you can put static midi into a better sample library and have it come out the way you want you are going to be sorely disappointed.  It's more along the lines of the better the sample library the more time you will need to spend detailing each instrument which WILL take hours and hours to get as authentic as possible. 

I would stay far away from symphobia if you want any sort of flexibility.  IMO it's overpriced for what it is and you would be far better off with the older East West QL Symphonic Orchestra which you can pick up on the (relatively)cheap these days if you time it right. 

http://www.soundsonline.com/symphonic-orchestra

[AngelCityOutlaw]

Depends on how you're defining "mediocre" 

[Yoozer]

I would've given something precious to have that piano sound 30 years ago.

For ensembles, consider layering real strings with synth strings. It's an art to achieve that, however - not everything will work.

As for the rest - do consider you're setting very high standards for yourself. For fun, check out http://www.synthmania.com/sr-jv80-02.htm - then you'll hear how much better you have it nowadays. That doesn't mean you shouldn't upgrade - putting the bar higher for yourself is a good thing - but it's easy to get lost in buying things. Buying things won't make you a better musician or composer.

Even with more expensive libraries, adding (and considering!) the kind of realism @Garpocalypse mentions is necessary.

[AngelCityOutlaw]

On 9/14/2018 at 2:31 PM, Garpocalypse said:

It's more along the lines of the better the sample library the more time you will need to spend detailing each instrument

I disagree with this.

The better the sample library, the more intuitive it should be. I define a "mediocre" sample library as one that is incapable of playing a melody authentic to the rhythmic vitality of that instrument within a single patch.

Allow me to offer a somewhat different opinion here:

You cannot "get around" mediocre samples if realism is your goal for the simple fact that if you don't have a sample of something, your virtual instrument cannot play it. You can try to "fake" something by manipulating the sounds, but that almost always sounds obvious. So it's best to just write to the samples' limited ability.

A good sample library will feature all standard articulations for the instrument, most importantly notes of differing lengths (Cinesamples is great at this), and you can do it all within one instance of Kontakt or whatever. East West Symphonic Orchestra fails on this account. It doesn't have any meaningful variation in short notes and it's a nightmare to use. You'll spend hours making it play the line you want, likely across several patches for one instrument, only to realize that a better library could've done it in a matter of minutes. So much "orchestra" music now is just "chord with left hand, melody with right" and banking on the fact that the lines have "true legato" as being "realistic" even though no orchestra plays exclusively legato through every phrase. Symphobia is a textbook case of such a "mediocre" library.

If it takes you hours and hours to craft a believable phrase, and even then it ends with a mixed result, it's a sure sign the sample library is garbage. And take it from me, a guy who spent a lot of money and time on this in recent years: A lot of "budget" libraries really suck.

It's not the advice anyone wants to hear, but the reality is, if you're serious about making realistic mockups, there are no "tricks" anyone is going to be able to give you — outside of learning more about orchestration of course — but it is inevitable that you need more capable samples eventually.

 

[Meteo Xavier]

I refuse to believe professionally made "budget" libraries suck, especially since musicians are more than well known for being zealous about sound quality, since the chaotic nature of composing, arranging and producing eventually mandates that songs will require "that" kind of sound that could be anywhere in anything. I routinely have to deny top-quality samples and instruments in the tracks I'm working on simply because they don't work or flow inside the track I'm working on - this is why I still keep almost ALL the extremely outdated soundfonts and VSTs I had before I got good at this. Sometimes I still need a shitty flute or violin or choir AH sample because that's what turned out to fit the spot I needed filled in.

Funny Yoozer mentioned the ancient Roland orchestral card. A few months ago I bought the Roland SRX-06 card so I could have access to the Final Fantasy Tactics and Zelda 64 sounds as needed for my ambitions.

[AngelCityOutlaw]

1 hour ago, Meteo Xavier said:

I refuse to believe professionally made "budget" libraries suck, especially since musicians are more than well known for being zealous about sound quality,

I'm not referring to "sound quality". Obviously, the recordings of these companies is top notch.

The content they actually sampled, and the interface with which to use it, are another story.

Here's the example I always use when someone says something to the effect of "samples don't matter": Make me a mockup of John Williams' Flight to Neverland with something like Orchestral Essentials and then get back to me.

[Meteo Xavier]

Sure thing bro. Orchestral tracks are $275.00 per audio minute plus mastering costs and Paypal's fees. 

Do you prefer a paper invoice or a virtual one?

[Jorito]

10 hours ago, AngelCityOutlaw said:

A good sample library will feature all standard articulations for the instrument, most importantly notes of differing lengths (Cinesamples is great at this), and you can do it all within one instance of Kontakt or whatever. East West Symphonic Orchestra fails on this account. It doesn't have any meaningful variation in short notes and it's a nightmare to use. You'll spend hours making it play the line you want, likely across several patches for one instrument, only to realize that a better library could've done it in a matter of minutes.

Just pointing out that "all within one instance" isn't necessarily a criteria for "better". Some people (myself included) actually _like_ to have articulations on separate tracks rather than messing around with keyswitches, both because of the overview it provides and because it's easier to manage. I am aware that my writing here is as subjective as your comment, but whether or not you can/should do all from within a single instance doesn't make a sample library "better". At the end of the day, workflow is a personal preference and it ultimately is all about that choice + the quality of the samples (which only to an extend is objective but also subjective taste).

Anyway, sample quality is only part of it, as other people said. Better libs do make things easier, but at the end of the day it's all about the composition, the orchestration and performances. All things (or at least the first 2) you can even do with "mediocre" samples.

If you're looking into Embertone samples anyway, I can recommend the cheaper libs they have on offer. They sound great in their own right, so make sure you give Chapman Trumpet, Shire Whistle, Jubal Flute and Crystal Flute a listen at least. I think they are like $20-$30 and I use them regularly.

[Neifion]

Pick a song and mock it up. Import the original into your DAW, map the tempo, and reconstruct it one instrument at a time. Match the dynamics, phrasing, and energy. A/B the original and your track constantly to hear if you’re getting the same feelings. It’s great ear training and mix training as well, since you have a mix reference right there. If your transcribing skills aren’t the greatest yet, consider buying the score so you can check your accuracy.

I’ve mocked up several John Williams cues and every time, I find I’ve not only learned so much about orchestration, but also performance and getting samples to sound like an actual expression of emotion rather than a Frankenstein patchwork of static recorded snapshots.

Lastly, emotion always trumps realism. Yes, you need a certain level of realism and control in your samples to be able to express that emotion, but if you can get your listeners to believe that there is feeling and intent in your music, (which you can achieve even with mediocre samples as long as they have dynamic control), that’s paramount.

 

[AngelCityOutlaw]

8 hours ago, Jorito said:

Just pointing out that "all within one instance" isn't necessarily a criteria for "better". Some people (myself included) actually _like_ to have articulations on separate tracks rather than messing around with keyswitches, both because of the overview it provides and because it's easier to manage. I am aware that my writing here is as subjective as your comment, but whether or not you can/should do all from within a single instance doesn't make a sample library "better".

Libraries like Cinesamples give you the choice, though. By default, all articulations can be activated via velocity-switch. IMO, there is no reason why any modern sample library should exclusively be limited to one articulation per track.

[Nabeel Ansari]

Most big-time composers work using separate articulations per patch instead of keyswitching.

Yes, it's dumb UX to not offer any switching (and VIenna knew this a long time ago and got on a great mapping software), but to say that the sample library like HO is incapable of getting a good result even after spending "hours" compared to "minutes" (lot of hyperbole here) is just plain wrong. A lot of people use the several-track-articulation workflow to great effect. It gives better mix control because you can control articulation volumes independently.

Keep in mind also, just because the VST doesn't offer switching doesn't mean that's the end of the read. Logic composers get to map things however they want externally. There's also a handful of free ways to create articulation maps intercepting MIDI, and this use-case is actually superior to in-built articulation switching systems inside libraries because your output is just going to separate MIDI channels, and so you can thus create 1 MIDI map that works for several different orchestral libraries from different companies (by just loading the same MIDI channels with the same artics like "1 leg 2 stacc 3 spicc 4 pizz")  instead of having to go and configure all of them individually inside their tiny GUI's.

[JohnStacy]

My skills leading ensembles transfer to using samples, in the times where I use samples. More times than not, I write in a very realistic way, then try to make the samples conform to this. For example, the piano sound I use in virtually everything I do is very mediocre. It's not great, but it's small and sounds consistent. I record all my piano stuff in manually, even if I have to do it at slower tempo and speed it up. Because of the recording everything manually, the velocity data is recorded from the keyboard, so it adopts my performance style. The result? With a little bit of EQ it actually sounds like a pretty good performance. I also have to apply humanization in Logic to it because my midi keyboard sends velocity somewhat uneven. A gradual crescendo sometimes turns into something that suddenly gets much bigger.

Same for drums. I record all my own drums now on a free drum sound. Sennheiser released a drum library called Drum Mic A (Drumica) that was designed as an advertisement for their microphones. Their microphones are great, and the library sounds great. Record those sounds on a midi kit and you can have some REALLY good drum tracks.

That said, if you do this for all instruments, it helps. Mediocre sounds, if you just punch in the notes as all the same velocity, perfectly identical note lengths, etc, will sound bad not only because the sample quality is low, but also because the actual music data the samples are applied to is dull, inhuman, and static. Even the worst samples sound better if you apply musicality to them, and include the little inconsistencies and imperfections that come with it.

Going further than that, quantization is a pretty bad thing to do when you have lower quality samples. Many samples, when played exactly on top of each other will sound very unusual, which can be a neat effect but otherwise is not what you want. Most of the time, depending on the tempo, I will quantize to the 32nd, or for really fast tempos, the 16th note. This makes everything line up, but allows for some variety that comes across as performance rather than sloppy rhythm.

Doing a lot to balance the samples you use in context also works wonders. Applying reverb, predelay, and other basic audio techniques can create a space and a field that can be believable. You mention orchestral samples, so thinking about writing orchestral music is a really solid idea. In an orchestra, you probably aren't going to have a flute that sounds louder than the entire orchestra. In an orchestra, especially one of good players, sounds will blend and come out when needed, back off when not. Filter automation can assist with this, lowering the cutoff for softer dynamics, and raising it for louder dynamics. Having a trumpet section playing with a soft tone (darker, less brassiness, lower filter cutoff) but being heard loudest in the mix just sounds off, in a very extreme way. Having a trumpet section playing with a loud tone (brighter, more brassiness, higher filter cutoff) but buried in the mix also sounds off. Balancing the sounds with your writing is a big way to continue.

On that note, a thing I stick by is that if your orchestration/arrangement sounds badly balanced, and muddy in general midi sounds, there is going to be an extreme amount of mixing gymnastics you will have to do to fix those balance and transparency problems. If the general midi sounds applied to the arrangement sound balanced, just with bad sound, then applying any better sounds will improve the sound of the mix. This is only if the orchestration is good to begin with.

TL;DR - Focus on audio production and orchestration, then apply the samples to what you have done.

[AngelCityOutlaw]

7 hours ago, PRYZM said:

Most big-time composers work using separate articulations per patch instead of keyswitching.

I have my doubts that "most" do from what I've seen, especially given time constraints. Also, the libraries I were particularly praising use velocity-switching by default. Individual preferences may very, but I'm not buying that the two methods are equally efficient when the chips are down.

7 hours ago, PRYZM said:

but to say that the sample library like HO is incapable of getting a good result

Didn't say anything about Hollywood Orchestra, but

7 hours ago, PRYZM said:

even after spending "hours" compared to "minutes" (lot of hyperbole here) is just plain wrong.

Uh huh. Again, make me mockup of like, the first couple minutes of Beethoven's Op. 59 with Symphonic Orchestra, OE or one of the garden-variety NI string libraries and then do it with something like CSS and compare the result. Be sure to time yourself.

7 hours ago, PRYZM said:

Yes, it's dumb UX to not offer any switching

 

7 hours ago, PRYZM said:

Keep in mind also, just because the VST doesn't offer switching doesn't mean that's the end of the read. Logic composers get to map things however they want externally. There's also a handful of free ways to create articulation maps intercepting MIDI

Spending money on sample libraries is, frankly, a decision far too many people in computer music take too lightly.

The fact that third-party customization options are available to you, provided that you're willing to put up with an additional, unnecessary learning curve is not incentive to buy something when options are available to you that will work as you require right "out of the box".

Trust me OP, if you're still out there: buying into the romantic notion that tools don't matter that much will wind up costing you and you will wind up sacrificing musical ideas to the mercy of either incapable samples, tedious workflow, or both. Over the last two years, I spent more money than I would've liked, but in doing so I got myself a setup I'm perfectly happy with, and don't see myself upgrading anytime soon simply because it can play whatever I throw at it with no hassle or convoluted routing and scripting. It's what I should've done from the start — it would've saved me so much money and time. Now, I'm left with a number of sample libraries I struggle to find a use for, and cannot resell.  

[Meteo Xavier]

No one said tools don't matter, they just don't matter the way you're arguing here. It's common sense discovered via experience, not romance.

Buying a Roland sound canvas sure didn't get me laid at any point in time, that's for damn sure.

[AngelCityOutlaw]

2 hours ago, Meteo Xavier said:

No one said tools don't matter, they just don't matter the way you're arguing here.

My challenge to prove it still stands

[Jorito]

I’d rather have you prove that you’re right in stead of us proving you’re wrong. Put your money where your mouth is, and all.

[AngelCityOutlaw]

2 hours ago, Jorito said:

I’d rather have you prove that you’re right in stead of us proving you’re wrong. Put your money where your mouth is, and all.

You doubt that sample libraries that utilize all articulations within one patch, velocity or keyswitch control, combined with the fact that it samples a wider range of note lengths, is both more intuitive and likely to produce a realistic result, in a lot less time, than using a sample library that does not have such features, and is spread out across multiple tracks? That's a very odd thing to doubt.

I know for myself from using the libraries. A good example for brass, but it applies across the board, would be the Indiana Jones theme. CineBrass VS East West.

I could demonstrate this for you if you want, but I honestly don't want to summon up the willpower to attempt it with Symphonic Orchestra and I know it's not doable with Orchestral Essentials. But maybe this video will shed light:

Excluding his talking, he did that in under ten minutes. Now, he could make it more realistic yet by expanding those first four bars out to separate trombone patches (takes minimal effort) and adding some tempo automation, etc. And he used the keyswitch map too, which isn't as fast as the velocity map for obvious reasons.

Good luck getting that kind of result with a library that requires you to juggle multiple tracks for all those articulations, articulations which SO or Symphobia straight up don't have, in that amount of time. Last I checked, both libraries cost easily as much as CineBrass. Keep in mind, these are pretty simple phrases for these instruments to play as well in real life! Sure, with Symphobia or SO you get a "whole orchestra", but forgive me if I think it's worth it to save your money.

They even did the Raiders March

   I could still do my own mockup, with more detail, if you wanted, but I think I've made my point.

[Meteo Xavier]

5 hours ago, AngelCityOutlaw said:

My challenge to prove it still stands

 

5 hours ago, Jorito said:

I’d rather have you prove that you’re right in stead of us proving you’re wrong. Put your money where your mouth is, and all.

 

On 9/16/2018 at 4:08 AM, Meteo Xavier said:

$275.00 per audio minute plus mastering costs and Paypal's fees.

 

[AngelCityOutlaw]

Yes, that's very clever.

[bustatunez]

To the OP, Der_Winter:

https://neobustatunez.livejournal.com/2990.html <-- If you're using Kontakt, this is how I get away with using samples from 2004, some even from the 90s. It's all about programming in dynamic expression, and using certain 3rd party scripts. Warning that the post is 11 years old and so my UI will look a bit different from modern versions of Kontakt.

https://www.dropbox.com/s/n4qn3hjaqhv1rxo/GDC16_AAA_Virtual_Orchestration_on_an_Indie_Budget.pptx?dl=0   (audio examples here) <-- My GDC talk on more broad concepts of how I sequence, use reverb, live overdubs, and so on. Lots of tips and tricks and such for each section of the orchestra. FYI, normally this talk can be found in video format at the GDC Vault ( https://www.gdcvault.com/play/1023727/AAA-Virtual-Orchestration-on-an ) but it's not loading today for some reason.

 

To AngelCityOutlaw:

For some reason Pryzm tells me to tell you that I use articulation patches and almost never use keyswitching except for choir syllables (Oceania) and 1/2 vs. wholestep trills (Hollywood Strings). Keyswitching makes it harder to tell what's going on at any given point, tougher to edit individual articulations, and it can make stemming a pain in the ass if you haven't specifically set up separate tracks for shorts vs. longs beforehand. It also makes it obnoxiously difficult to orchestrate for a live recording, especially if you're working with external orchestrators. 

One tip that might help though - whenever I record tutti horn melodies (EastWest H.Brass), I always arm three tracks at once: Legato, Marcato, and Staccato. I play everything in realtime on keyboard, then do some manual timing correction on all three at once. Then as a final step, I'll mute the notes I don't need from each of the three. So for example, obviously for actual staccatos I'll only need the Stacc patch, for quick motion maybe I'll have Stacc+Legato, for bold notes with oompf I might keep the Legato and the Marcato, and so on.

(You can see more of my brass sequencing technique in the GDC talk)

Cheers,

- Will

[AngelCityOutlaw]

2 hours ago, bustatunez said:

For some reason Pryzm tells me to tell you that I use articulation patches and almost never use keyswitching

I shall then tell you to tell Pryzm that he's too hung up on key-switching, as I'm a velocity-switching and footpedal supremacist. =P

Which also solves the orchestrating/notating problem key-switches can present.

2 hours ago, bustatunez said:

Then as a final step, I'll mute the notes I don't need from each of the three. So for example, obviously for actual staccatos I'll only need the Stacc patch, for quick motion maybe I'll have Stacc+Legato, for bold notes with oompf I might keep the Legato and the Marcato, and so on.

The classic strategy, as I've always understood it. 

I still argue in favor of not layering and just having samples of the actual articulation style you want.

EDIT:

But I see what's happening here — Nabeel is on a quest to prove that I'm completely wrong — so I've gone ahead and googled Keyswitches Vs Single Articulations and taken from sources that deal with composers using specifically orchestral instruments.

https://vi-control.net/community/threads/key-switches-vs-separate-tracks.60844/

https://vi-control.net/community/threads/multi-articulation-patches-with-key-switching-vs-single-patches.46256/

https://www.vsl.co.at/community/posts/t48376-Key-switches-vs-Separate-Articulations#post269778

https://www.vsl.co.at/community/posts/t41818-Key-Switches-vs--Single-Articulations#post250918

https://www.gearslutz.com/board/music-for-picture/888790-orchestral-template-keyswitches-vs-single-articulations.html

Some stand-out quotes:

Quote

I have no idea why I would want articulations on separate tracks. Even if there really, really, really was some odd reason to separate one articulation (separate output because of some additional processing or something), I'll just load that one patch in a second instance - or just route it differently, depending on the sample player - and leave everything else as it is.

Quote

Using separate MIDI channels for each articulation is the old fashioned way of doing it. The fact that many libraries use the terms "longs and shorts" tells you that it is based in the past. These days it is far faster and more economical to put everything on one staff per instruments. You also have to realise that many people working in the profession write by layering, rather than by conceiving an orchestral sound, so the 1 track per articulation makes more sense in that situation

Quote

In my view, it depends on how many articulations you want to use per instrument. So if you use Special Edition Libraries, it may make sense to put a staccato-, a sutain-, a sforzato- and a legato- track per instrument. But as soon as you play with full libraries (more articulations), this "track-per-articulation system" would be a bit confusing. Especially if you change the articulations a lot, what would actually be necessary for a natural sound. Then a melody would be split up so often that everything would be quite confusing. Furthermore, one has to keep in mind that samples also often include control curves. In such a situation, I personally like the melody in one track and not splitted up into many many pieces ...

Quote

For me, I like one instrument per track and use keyswitches.  I customize each matrix in VIpro2 with black key/white key combinations for X&Y values.  Generally, I can cycle through the correct articulations and still ride the mod wheel, which I've assigned as my velocity x-fade for a live performance, then go back and ride any other CC, such as expression or slot x-fade for tweeking.

The purpose of this is for me to treat it like it is a real instrument... like an organ or something.  I've found that yields better musical results.

and my favourite, which is the point I've been trying to drive home:

Quote

You could easily end up with twenty tracks for the melody alone.

Among some of the posters, or information shared by the posters, there is no shortage of composers who've worked on everything from Zorro to Tomb Raider (Movies) who use keyswitches. Few seem to argue in favour of single articulations unless they like layering, processing individual articulations, or just because that's how they've always done it.

I remain unconvinced that single articulations are the norm (doesn't really matter anyway), and I remain unconvinced that it's inherently a better method when it comes to making mockups. Feel free to disagree, I know you will, but I'm steadfast on this opinion and will share it again in a heartbeat.

[bustatunez]

Hmm, well I'm late to the party and don't know why this is an argument. People do both - even in the threads you linked, you'll see various opinions, plus several of the posters acknowledged it has more to do with the libraries themselves.

I personally use separate arts for the reasons above and also, more importantly, for the sound - it's rare that I'll find one exact art that does exactly what I want it to 100% of the time, but when combining them together, you can get much closer. Having the freedom to do that and tweak at will has been crucial for me in gigs like CoD where the mockup needed to be as realistic as possible - not to mention the additional headache stemming would bring if your keyswitch patches aren't organized in the same way as you'll be stemming. Lastly, even beyond layering, like you mentioned it's much easier to do mix tweaks and other often-essential customizations on an art-by-art basis, which can be a pain when you have everything in just one patch.

That said some people do prefer keyswitches, hence why many libraries provide both these days - a grand KSW patch and then all the individual elements. Plus like I mentioned earlier, most folks (myself included) use a combination of both, based on the libraries themselves. It's just a matter of preference and doesn't really correlate with experience/notoriety - you might as well ask, "What is the best DAW?"

(the answer is reaper tho)

 

1 hour ago, AngelCityOutlaw said:

I'm a velocity-switching and footpedal supremacist. =P

Ooooh, do you have organ-style pedals you're using for keyswitches? That sounds pretty friggen awesome!