Virtual Playing Orchestra - pre announcement

[Michael Willis]

(10-18-2016, 03:26 PM)Paul Battersby Wrote:

(10-18-2016, 02:11 PM)Michael Willis Wrote: I noticed that some of the keyswitch SFZs will play a default sound without me having to trigger a keyswitch, I'm not sure what's different about them than the ones that didn't, but I'll look more into it.

That's good news provided we can figure out why. Can you give me the name of an instrument where the default sound plays properly and another where it doesn't and I'll see if I can spot an important difference.

Hmm, I must have been mistaken, because going over them again, I can't find any keyswitch files that give me a default sound. Oh well, it's really not that much trouble to trigger an initial keyswitch.

(10-18-2016, 03:26 PM)Paul Battersby Wrote:

(10-18-2016, 02:11 PM)Michael Willis Wrote: Something I forgot to mention that I discovered earlier is that sometimes if I press and release a key quickly, the note drones until I press and release the key again.

If you can provide the name of the instrument and the articulation, I'll look into it. I'd like to know if it also happens with Sforzando and I'd like to examine the waveforms to see if your hypotheses as to the cause, could be right.

I haven't narrowed it down as much as I would like, but I can get it to happen pretty consistently with Strings-1st-violin-SEC-sustain, Strings-viola-SEC-sustain, or either of the corresponding keyswitch files with the sustain sound selected, by sliding my hand across the piano doing a glissando.  Incidentally, I haven't gotten the same results out of the marcato articulations. After reading a bit I see that I misunderstood what the "offset" does, I thought it was a delay, but I see that offset is just a way to pick a different starting point within the sample. I tried seeing if I could reproduce the bug with marcato articulations, because many of them have a longer delay before the sustain sound picks up, but it didn't happen, which leads me to wonder if the delay actually prevents the bug?  I'm still not sure, and frankly I'm not even sure if this will ever happen during normal DAW use; as I said before I'm just piping midi events directly from my digital piano to the sampler.

I've been taking a better look at the SFZ files and the samples.  I'm assuming that the samples with the PB suffix are modified by you, given that those are your initials I'm curious what kind of sound processing or other modifications you did to the samples?  Was it mostly an effort to get samples from disparate sources to sound like they go together?

[Paul Battersby]

(10-19-2016, 03:24 PM)Michael Willis Wrote: Hmm, I must have been mistaken, because going over them again, I can't find any keyswitch files that give me a default sound. Oh well, it's really not that much trouble to trigger an initial keyswitch.

I wonder if a second patch might have had a default sound as a result of a keyswitch being pressed for a previous patch.

(10-18-2016, 02:11 PM)Michael Willis Wrote: I can get it to happen pretty consistently with Strings-1st-violin-SEC-sustain, Strings-viola-SEC-sustain, or either of the corresponding keyswitch files with the sustain sound selected, by sliding my hand across the piano doing a glissando.

That gives me something to try but at least it's only happening during an unusual circumstance.

(10-18-2016, 02:11 PM)Michael Willis Wrote: I'm assuming that the samples with the PB suffix are modified by you, given that those are your initials

Yes. I added my initials as a suffix to keep track when I made a change and so the original authors don't get blamed for anything I did to their samples.

(10-18-2016, 02:11 PM)Michael Willis Wrote: I'm curious what kind of sound processing or other modifications you did to the samples?

Depending on the particular sample, I may have done any combination of the following:
- converted from stereo to mono
- normalized
- applied a dynamic auto tune
- reduced reverb
- truncated silence at the beginning
- shortened the attack

At the sfz file level, I've also done any combination of the following:

- adjusted the attack, sustain, decay, release and/or offset
- applied a static pitch adjustment
- adjusted the volume
- adjusted the panning
- removed "bad" samples, stretched the mapping of others

[Samulis]

Hey Paul, just saw this thread. This sounds like a fantastic project and is EXACTLY the reason why I made the CE sample set public. If you'd like an extra pair of eyes/ears on the sounds, let me know.

When it comes time to share it with the world, I would also be happy to hyperlink to it from the Community Edition page (as I have done for Bigcat's conversions), and alert the individuals following the specific Community Edition mailing list (it is only a few, but a few is better than none). Let me know if you would like that.

[Paul Battersby]

(10-21-2016, 06:49 AM)Samulis Wrote: If you'd like an extra pair of eyes/ears on the sounds, let me know.

Sure. I'll send you a private message with a link to the library.

(10-21-2016, 06:49 AM)Samulis Wrote: When it comes time to share it with the world, I would also be happy to hyperlink to it from the Community Edition page (as I have done for Bigcat's conversions), and alert the individuals following the specific Community Edition mailing list

That would be great.

[Samulis]

Oh, I forgot to mention- this isn't exactly explicit, but there are actually some different instruments in the 100 and 256 "packs" on the CE site that aren't in the 1,952 pack. They're very low detail, but might be mixable with other libraries.

Edit: If you're looking for a Marimba, I posted up a subset of the Marimba from Versilian Studios (which ended up in CO2) on freesound-
http://www.freesound.org/people/Samulis/packs/15684/

There's also an old gong hit- I'm not sure if I put this in CE or not, but here it is if you would like one-
http://www.freesound.org/people/Samulis/sounds/209398/

I found the NBO orchestral chimes set is indeed gathered from my Freesound as well, in case anyone asks where they originally came from-
http://www.freesound.org/people/Samulis/packs/12231/

(this all makes me realize how much I need to upload more things to freesound... if only their uploader was better designed for mass batch uploading...)

[Samulis]

So, my general first impressions:

• It seems to work fine in ARIA!
  
• Overall, this is very nice. It's clear a lot of work has gone into stitching these libraries together, and I agree with the artistic choices you've made.
  
• While I'm not much of a fan of mono audio (especially reducing stereo to mono- I hope you did that by deleting one track rather than merging the two tracks *wink wink nod nod*), I think it works out OK. It might be nice if you panned out each percussion instrument sample/group so that it is found in a spot it might be found.
  
• The modwheel for attack is a cool idea, but I'd prefer it more if it had some control over the dynamics, even just a filter. In fact, adding a lowpass filter to nearly every instrument, mapped to velocity, would greatly enhance the dynamics of most instruments- this is the approach used by GPO, as it lacks velocity layers on most instruments. I'd be happy to dig around and see if I can help set that up with you if you'd like.
  
• Re: Modwheel, I would also suggest reversing the current behavior, i.e. that at a higher modwheel level, a more direct attack should be used than at a lower level. This is closer to what one might expect on traditional modwheel crossfading instruments and would be more intuitive in my opinion.
  
• Vibrato is rare on Horn, and uncommon (but still possible/sometimes done) on Clarinet.
  
• It might be interesting to look into Kurt Landre's variation upon the Iowa trumpet (Van Houzen trumpet) to see if he made any changes worth incorporating or duplicating- I found the demos to that very convincing, and I know it can't all just be magical WIPS goodness.
  
• The timpani rolls are pretty clearly faked by repeating samples. There are live rolls in the CE set, albeit I was a moron and decided to not use a louder velocity layer, which I may resolve when VSCO 2.1 comes out (I could probably just send you some of the v5 samples of that now if you'd like with some special license thingy or something).
  
• The 2nd snare (CE) has mono hits but stereo rolls. Not sure if this is a mistake or intended as an effect (if it is, it's a pretty cool effect).
  
• There is not much of a difference between 1st and 2nd violin sections. Since 1st is bigger than 2nd typically, you could consider blending in the CE Violin Section with the firsts, or alternately, since the CE Violin Section is only 5 players, it could be the 2nds (if a larger sound is desired, blend the two velocity layers together). Another option is to EQ the 2nds differently, typically knocking down some of the highs slightly (this is what we did with the VSCO 2 Pro).
  
• I do think sampled trems are useful, and I'm not just saying that because CE has them.
  
• The marcato presets for some instruments sound more like a Fp < (forte-piano crescendo).
  
• It sounds like there is a bit of the sample being cut off in the 2nd (and presumably 1st) Violin Section Pizzicato, I guess from some random script. Consider adding a 5-20 ms attack on all instruments as a safeguard (it saved my bum on many instruments).
  
• You should be able to simulate harp harmonics if you want by EQ'ing out the fundamental of each note (I think).
  
• For added realism to vibrato, you can reduce the volume change and instead pass some of the LFO onto a filter or EQ (like WIPS does, if I recall correctly). Vibrato consists of about 2-8 cents of pitch transform and approximately 2-8 dB of volume change, depending on instrument, and for winds is more accurately a transition between louder and quieter dynamics (as winds get louder, they get sharper slightly and also gain more middle and upper harmonics). I've attached some analysis notes I did back when Simon A. and I were working on trying to figure out if we could enhance the VSCO 2 crossfading to be more accurate using filters (turns out we couldn't with the technology and time we had). Theoretically if you were to apply an EQ around 3.5 kHz to a forte bass trombone sample with a medium-sized Q that brings that portion down a bit when a quieter velocity is played, you should be able to simulate a mp dynamic, for example. (the only problem is, the EQ frequency would need to be adjusted for each note, theoretically).
  

[Paul Battersby]

Samulis,

Thanks for you detailed comments. It's appreciated. Here's my reply.

(10-21-2016, 02:15 PM)Samulis Wrote: It seems to work fine in ARIA!

That's good. I'd expect it to work with ARIA given that the Garritan ARIA player and sforzando both use the ARIA engine.

(10-21-2016, 02:15 PM)Samulis Wrote: Overall, this is very nice. It's clear a lot of work has gone into stitching these libraries together, and I agree with the artistic choices you've made.

Thanks. Yes, it has been a lot of work. You know what that's like, I'm sure!

(10-21-2016, 02:15 PM)Samulis Wrote: While I'm not much of a fan of mono audio (especially reducing stereo to mono- I hope you did that by deleting one track rather than merging the two tracks *wink wink nod nod*), I think it works out OK. It might be nice if you panned out each percussion instrument sample/group so that it is found in a spot it might be found.

The reason I made solo samples mono, is because I just couldn't fix the changing stereo panning in various samples. During the same note the panning location would change and I just couldn't make them all sound like the same source within the same note and from one note to the next.

I chose not to pan any instruments because I felt that should be up to the composer. For example, in some orchestras, the 1st and 2nd violins are seated on opposite sides of the conductor instead of both being on the left. Without panning, it also makes it easier to use the individual instruments / sections for purposes other than full orchestra.

(10-21-2016, 02:15 PM)Samulis Wrote: The modwheel for attack is a cool idea [...] adding a lowpass filter to nearly every instrument, mapped to velocity, would greatly enhance the dynamics of most instruments [...] I'd be happy to dig around and see if I can help set that up with you if you'd like.

Yes, I'd appreciate that. If you can come up with the .sfz opcodes and settings that you think will improve the dynamics, I'll give it a try.

(10-21-2016, 02:15 PM)Samulis Wrote: Re: Modwheel, I would also suggest reversing the current behavior, i.e. that at a higher modwheel level, a more direct attack should be used than at a lower level.

I wanted the default mod wheel setting to result in a fast attack, so that without touching the mod wheel, I would have an articulation that is between that of sustain and staccato.

(10-21-2016, 02:15 PM)Samulis Wrote: Vibrato is rare on Horn, and uncommon (but still possible/sometimes done) on Clarinet.

I'll remove the .sfz added vibrato from the solo horn. I haven't done anything to the clarinet samples regarding vibrato. I'm using them as I found them.

(10-21-2016, 02:15 PM)Samulis Wrote: It might be interesting to look into Kurt Landre's variation upon the Iowa trumpet (Van Houzen trumpet)

That's for Kontact right? I don't have Kontact.

(10-21-2016, 02:15 PM)Samulis Wrote: The timpani rolls are pretty clearly faked by repeating samples. There are live rolls in the CE set, albeit I was a moron and decided to not use a louder velocity layer, which I may resolve when VSCO 2.1 comes out

Actually, I created left and right hand hits for the timpani, then recorded the midi of myself playing a roll on the keyboard. I fixed any gross timing errors in my performance (which maybe I shouldn't have), then I played back that midi for the various timpani notes and recorded the audio. I can hear now that I need to mess around with the midi notes to more greatly vary the timing and velocity. I think I'll go back and take another shot at the timpani roll.

(10-21-2016, 02:15 PM)Samulis Wrote: (I could probably just send you some of the v5 samples of that now if you'd like with some special license thingy or something).

Sure. What sort of licensing do you have in mind? Everything else is Creative Commons.

(10-21-2016, 02:15 PM)Samulis Wrote: The 2nd snare (CE) has mono hits but stereo rolls. Not sure if this is a mistake or intended as an effect (if it is, it's a pretty cool effect).

The roll should already be mono. I used the width=0 opcode and I'm looking at the meters and it shows both channels are equal at all times. Given that the Garritan Aria player you're using and Sforzando that I'm using both use the ARIA engine, I'd expect the width=0 opcode to work the same in both.

(10-21-2016, 02:15 PM)Samulis Wrote: There is not much of a difference between 1st and 2nd violin sections.

Isn't that to be expected since the 1st and 2nd violin are actually the same instrument?

The 1st and 2nd violin section sustain patches come from Sonatina Symphonic Orchestra. I'm using the samples as I found them but I could try a high end EQ reduction to the 2nd violins as you suggested. For the 1st and 2nd solo violins, I tried to differentiate them a little, I remapped the samples for 2nd violin solo and applied EQ changes as well.

(10-21-2016, 02:15 PM)Samulis Wrote: I do think sampled trems are useful, and I'm not just saying that because CE has them.

I have tremolo for the string sections ready for the next update.

(10-21-2016, 02:15 PM)Samulis Wrote: The marcato presets for some instruments sound more like a Fp < (forte-piano crescendo).

Yea, I've struggled with that articulation trying to get it to sound right.

(10-21-2016, 02:15 PM)Samulis Wrote: It sounds like there is a bit of the sample being cut off in the 2nd (and presumably 1st) Violin Section Pizzicato, I guess from some random script. Consider adding a 5-20 ms attack on all instruments as a safeguard (it saved my bum on many instruments).

Some pizzicato already have an ampeg_attack of 0.03. I missed that on others. I'll make sure all pizzicato have an attack setting to prevent clicks.

(10-21-2016, 02:15 PM)Samulis Wrote: You should be able to simulate harp harmonics if you want by EQ'ing out the fundamental of each note (I think).

I'm using the VSCO2 harp samples as I found them. If you can suggest some .sfz EQ setting to make it sound even better, I'm interested.

(10-21-2016, 02:15 PM)Samulis Wrote: For added realism to vibrato, you can reduce the volume change and instead pass some of the LFO onto a filter or EQ (like WIPS does, if I recall correctly).

If you can give me .sfz settings that you think will work best, I'll try it.

Samulis Wrote: If you're looking for a Marimba

After I release version 1.0, I'm already planning a 1.1 with a few more instruments and perhaps enhancements of others. I'll look into your Marimba samples then. Is it just me or does the freesound web site only have individual notes that have to be downloaded separately one at a time?

[Samulis]

Well Paul, I may have gone a bit above and beyond on releasing the samples, and expanded the CE sample set by about 35%... but these things happen at 5 AM. XD
http://scoringcentral.mattiaswestlund.ne...838#pid838

All the percussion from VSCO 1 and Miscellania I and II is now in there (it's primitive, but there's some diamonds in the rough in there), and a few very very light (but still quite nice) mallet percussion reductions. I also added another tenor trombone I had lying around, a piccolo, and tinkered around with a few things.

I won't have time to do any deep digging in sfz opcodes until next weekend at the earliest, but I'll see what I can do. I know what it is in theory, I just don't know the exact codes to achieve the result.

[Paul Battersby]

Release 0.4 is now ready for download. Same link as in your private emails, just download the 0.4 patch file over top of the existing files if you're interested.

Some highlights:

- tremolo articulation for all section strings
- added English Horn (Cor Anglais)
- replaced the timpani rolls with something ironically more humanized
- fixed upper/lower case issues for LinuxSampler
- adjusted instrument ranges

I'm hoping this is the last pre-release and that the next release will be first official release containing any final changes if there are any.

For the first official release, unfortunately, I think I should rename all the .sfz files to remove the instrument class from the names. The names are too long. For example:

Strings-1st-violins-SEC-staccato.sfz

... would become

1st-violins-SEC-staccato.sfz

The classes of instruments will remain grouped together in their separate directories so the instrument class in the file name is redundant.

Samulis: it's great that you've released more samples to roam free in the wild but for now, I need to release what I have or I'll be tinkering with it forever. When I'm ready to update the first official release, that will be the time for me to consider what I can use from your latest free samples.

[Otto Halmén]

Everyone, send him more caffeinated beverages! Every time he stays up to 5 AM is a chance.

Jokes aside. Thanks, Samulis!