Scoring Central
Experimental samples - Printable Version

+- Scoring Central (http://scoringcentral.mattiaswestlund.net)
+-- Forum: Technology (http://scoringcentral.mattiaswestlund.net/forumdisplay.php?fid=5)
+--- Forum: Samples & Sample libraries (http://scoringcentral.mattiaswestlund.net/forumdisplay.php?fid=8)
+--- Thread: Experimental samples (/showthread.php?tid=381)

Pages: 1 2 3 4


RE: Experimental samples - bigcat1969 - 04-25-2018

My personal nfd mapping for kontakt, supply your own picture & verb, for anyone who can use it aka Sam... Someday another person hear is gonna get kontakt someday....
http://www.mediafire.com/file/4y69ley2t231x20/Westlund_Violins.nki

Loving the bottom 2/3rds. Agree that a touch less squeak on the high octave would be perfect, but all high violins drive my nuts really...

https://soundcloud.com/joe-stevens-16/matt-by-mattwest

Awesome stuff as usual, thanks for your work!


RE: Experimental samples - Mattias Westlund - 04-26-2018

Yet another new version. Spent some more time EQ'ing the samples, and I think I have gotten to a point where any further tweaks can be done at the sampler level, i.e. using the SFZ filters.

Also, since I'm going for a kind of oldschool vibe here, I fed the samples through my H/W Lexicon reverb (instead of using an ambience from ValhallaVintageVerb like in the previous versions) using the Recital Hall algo. Right now I think it sounds pretty great, though I will probably change my mind in a few days...

I have added mod wheel crossfade patches for the sustain and trem samples, plus velocity crossfading for everything else. Everything feels a little smoother now.

Known issues:

- The mapping is still off by an octave, I need to fix this.
- Some of the lowest notes sound kind of wonky. Since I used a hardware unit in the signal chain I had to render in realtime, and it seems that causes a slight glitch when REAPER starts rendering a new region. I'll have to re-render the offending samples.


RE: Experimental samples - Paul Battersby - 04-27-2018

First, these violins are sounding good. I don't want my feedback to suggest otherwise.

I like that there are now 2 instruments, taking up less disk space than your original 1st violins download (all because of a switch from .wav to .ogg?). There is a noticeable delay when loading up a patch, but I imagine that's normal for anyone using something with the quality of a professional library.

(04-26-2018, 06:36 PM)Mattias Westlund Wrote: Spent some more time EQ'ing the samples

I can still hear a hissing sound on most samples (sustain and tremolo). I had the same problem with my own samples and spend a lot of time EQing that out of each individual sample. I was able to do that at the .sfz scripting level. I'm beginning to wonder if perhaps that hissing sound that I hear, would still be present in a live performance when a group of violin players are holding the same note.

(04-26-2018, 06:36 PM)Mattias Westlund Wrote: I fed the samples through my H/W Lexicon reverb ...

These samples sound like they still have the flexibility for a composer to apply their own reverb, which is good. I think you've baked in less reverb that you did with your Sonatina Symphonic Library. If I recall, you said people used to complain about the amount of Sonatina Symphonic Orchestra baked in reverb.

(04-26-2018, 06:36 PM)Mattias Westlund Wrote: I have added mod wheel crossfade patches for the sustain and trem samples

It appears, and sounds, like the mod wheel only controls the percentage of each velocity layer being played. My personal preference, (which I did with my own library) is to have the mod wheel simultaneously control both crossfade and volume. Seems more natural to me that way. I've seen some composers use 2 faders, one for crossfade and another for volume. I've also seen a composer use a single controller (likely the mod wheel in that case) that does both.

Dare we hope for an entire orchestra with this new level of quality from you? If I recall, this is how Sonatina Symphonic Orchestra got started, with you asking people to evaluate some string patches you'd been experimenting with.  Big Grin


RE: Experimental samples - Mattias Westlund - 04-27-2018

(04-27-2018, 03:57 PM)Paul Battersby Wrote: I like that there are now 2 instruments, taking up less disk space than your original 1st violins download (all because of a switch from .wav to .ogg?). There is a noticeable delay when loading up a patch, but I imagine that's normal for anyone using something with the quality of a professional library.

Well yeah, ogg is lossy compressed format like mp3 so sound quality-wise it's not optimal, but for demo purposes like this it makes everyone's lives a little easier. I will likely go with wave or FLAC for the finished product though. The delay happens here as well, and it's probably sforzando decompressing the samples.

(04-27-2018, 03:57 PM)Paul Battersby Wrote: I can still hear a hissing sound on most samples (sustain and tremolo). I had the same problem with my own samples and spend a lot of time EQing that out of each individual sample. I was able to do that at the .sfz scripting level. I'm beginning to wonder if perhaps that hissing sound that I hear, would still be present in a live performance when a group of violin players are holding the same note.

This hissing sound that you describe is in every library I own, to varying degrees, and I think it's just the way strings sound in the high register -- some combination of bow noise and resonance of the space they're recorded in or something like that. Maybe Samulis can shed some light on this, as he has personal experience with recording live violin ensembles whereas I have not. At any rate, that sound is part of what makes violins sound like violins IMO and completely EQ'ing it out will remove a big chunk of the instrument's character. I have tried attenuating it down to more reasonable levels (since I'm layering samples here, noise also gets multiplied for each layer) but I definitely intend to keep it in there. As long as it's not harsh and grating, I don't mind it.


(04-27-2018, 03:57 PM)Paul Battersby Wrote: These samples sound like they still have the flexibility for a composer to apply their own reverb, which is good. I think you've baked in less reverb that you did with your Sonatina Symphonic Library. If I recall, you said people used to complain about the amount of Sonatina Symphonic Orchestra baked in reverb.

I tried a few of the different algo's but the Recital Hall (I keep typing 'Rectal Hall'... I wonder what Freud would have had to say about that) was by far the most transparent one. As you say I'm trying to keep it at a level where you can add your own reverb on top of it. SSO had way too much really, but that was a compromise between flexibility and sound quality. Some of the string sections sounded like complete garbage when dry; using an IR to give them some room to breathe was really what made the samples work.

(04-27-2018, 03:57 PM)Paul Battersby Wrote: It appears, and sounds, like the mod wheel only controls the percentage of each velocity layer being played. My personal preference, (which I did with my own library) is to have the mod wheel simultaneously control both crossfade and volume. Seems more natural to me that way. I've seen some composers use 2 faders, one for crossfade and another for volume. I've also seen a composer use a single controller (likely the mod wheel in that case) that does both.

That is a good idea, and I will probably add that!

(04-27-2018, 03:57 PM)Paul Battersby Wrote: Dare we hope for an entire orchestra with this new level of quality from you? If I recall, this is how Sonatina Symphonic Orchestra got started, with you asking people to evaluate some string patches you'd been experimenting with.  Big Grin

To be perfectly honest, I don't know. If there's one thing I can say with utmost certainty though, it's that I will NOT be making an entire orchestral library. If things work out and I can get a level of quality that I'm happy with with these samples, I might release a strings library. I have a few ideas for a winds library as well but that's way down the line and right now I'm not sure whether I'll even be able to finish this one. But who knows Smile


RE: Experimental samples - Samulis - 04-27-2018

^ I believe your explanation with the strings hiss is right. It's also tied to microphone placement, I believe, with the closer mics being more susceptible. You will notice with larger ensemble recordings and farther mics that it tends to go down a bit. The noise of the bow on the string is quite noticeable up close in person, even on lower notes. VSCO is quite closely recorded, so it's particularly noticeable.


RE: Experimental samples - peastman - 04-28-2018

This is sounding really awesome!  I'm looking forward to seeing where you go with this.  A full strings library would be great.

The one thing that jumps out at me is that it needs volume normalization.  In the VS instruments there's an abrupt jump in volume when it switches between layers.  And in the staccato there are a lot of notes where the two round robin samples are noticeably different in volume.  In the 1st violin sustains, there's also a range of notes right around middle C that have a harsh thump/click sound right at the start.


Quote:I will likely go with wave or FLAC for the finished product though.


Are there any samplers that don't support FLAC?  If not, I think it would be great if all the sample libraries eventually moved to that.  Half the size with no loss in quality.  A GB isn't what it used to be, but they still add up quickly.


RE: Experimental samples - Samulis - 04-28-2018

(04-28-2018, 03:56 AM)peastman Wrote:
Quote:I will likely go with wave or FLAC for the finished product though.


Are there any samplers that don't support FLAC?  If not, I think it would be great if all the sample libraries eventually moved to that.  Half the size with no loss in quality.  A GB isn't what it used to be, but they still add up quickly.

The only problem with FLAC is that it uses CPU to decode (same with any other compressed format). On the other hand, HISE and Kontakt both use compression methods which are perhaps slightly less efficient, but are (I believe) much lighter on the CPU. The compression is still a very good idea for freebie delivery, though.

Not to mention many people get their knickers in a twist if they learn the phrase "file compression" was related to a commercial product, even lossless. I always see a post on some group or forum by someone suggesting that they think Kontakt compression hurts the sound. Rolleyes


RE: Experimental samples - Mattias Westlund - 04-28-2018

(04-28-2018, 03:56 AM)peastman Wrote: The one thing that jumps out at me is that it needs volume normalization.  In the VS instruments there's an abrupt jump in volume when it switches between layers.  And in the staccato there are a lot of notes where the two round robin samples are noticeably different in volume. 

Hmm, I'm not really noticing this. All samples are normalized, the lower layer to -6dB and the upper layer to -3dB. Maybe your midi controller and mine have very different velocity curvers?

(04-28-2018, 05:21 AM)Samulis Wrote: The only problem with FLAC is that it uses CPU to decode (same with any other compressed format).

But not in realtime, surely? I've always assumed that samplers decode the audio on load, as decoding on the fly would introduce latency. I could be wrong though.


RE: Experimental samples - bigcat1969 - 04-28-2018

Kontakt at least must decode in real time as the memory footprint is smaller.
Actually if I remember right twice in my life I remember not liking the sound of Kontakt compression and going back to the original wave files, but that might have just been in my imagination anyway.


RE: Experimental samples - Samulis - 04-28-2018

^^ Both typically use some sort of DFD/disk streaming as far as I know, which sort of implies that the decoding is done on the fly (unless the cache some decoded version somewhere, but wouldn't that kind of negate the whole point of DFD?). Computers decode compressed video and audio on the fly constantly, so it's not really that surprising once you think about it. I believe once they've decoded a file, they leave it in RAM decoded. ARIA nicely shows you how this works in the 'performance monitoring' part of the 'settings' tab.