• Register
  • Help
Results 1 to 8 of 8

Topic: Is this a plausible-sounding solo violin?

Share/Bookmark
  1. #1

    Is this a plausible-sounding solo violin?

    Becker\'s recent reports on his efforts to get a good solo violin inspired this little experiment in slicing and dicing DDSV to get a little bit more timbral variety:

    http://www.stanford.edu/~gmsmith/mp3/Ave_Maria3_short.mp3

    Never mind the guitar part; I didn\'t give it the same attention I gave the violin part.

    At this point I\'ve done so much tweaking and listening I can\'t tell whether I\'ve made DDSV sound better or worse, or both. I\'m especially interested in what y\'all violinists would have to say.

    Please do know that I started with someone else\'s midi file, taken from one of the big midi archives -- but I chose it because it was the flattest rendering I could find, with no CC\'s, and I flattened tempo and velocity before I began. I did leave some of the staccato articulations in the guitar part, though.

    If this is so bad that it\'s embarrassing to comment on it, I guess I\'d like to know that, too.

    Oksi

  2. #2

    Re: Is this a plausible-sounding solo violin?

    Sharmy,

    Thanks for the reply. For me this was an interesting experiment that involved using an audio editor to bandpass-filter each DDSV sample into two samples -- one for the fundamental, one for the upper partials -- and then layering them with independent modulations in the hope of getting slightly more timbral variety than in the original sample. Sort of a poor man\'s (or lazy man\'s) \"group additive synthesis\" approach. Before I split the samples, I also applied a transfer function to the original slow attacks so that they became pseudo-fast attacks that I could either attenuate or not.

    Maybe it was a nutty idea in the first place, and maybe after mixing and remixing I wound up with not much more timbral variety than I would have gotten just using standard dynamic filtering. By \"lack of natural expression\" I wonder if maybe you mean, for example, the double-stops, which I didn\'t spend any time trying to do right. Or do you mean the too-pregnant swells? Or both? I notice spots where the remixing of the split samples sounds almost accordion-ish. Not sure what you mean about the incorrect attacks. Because of my tweaking and remixing of the samples, the rendering on the first pass started out with way too much bow crunch at the beginnings of notes, and maybe I went overboard using note tapers and overlaps to cover that up. But I don\'t think you mean the rendering is too legato; probably you mean just the opposite --that there\'s too much \"bowing\" going on?

    From poking my nose into way too many academic papers, I know violin and sax are the holy grail of imitative synthesis experiments, so I wasn\'t expecting anyone to fall down in awe. But it sounds as if you\'re saying it\'s not even in the ballpark.

    Although this particular rendering wasn\'t done with my wind controller, I\'m holding out for the ideal of having one sample patch respond as flexibly as possible to wind control, with maybe some minimal after-the-fact tweaking, as opposed to stringing together bazillions of pre-recorded \"articulations\" like letters in a ransom note. That\'s what drives my nutty experiments.

    Just curious, but do you like this cello snippet any better?

    http://www.stanford.edu/~gmsmith/mp3/bachcello.mp3

    This one I *did* do with my wind controller, and with less fooling around with the samples (Xsample).

    As for the guitar part, it doesn\'t cut it for me, maybe because that\'s my instrument.

    Thanks again.

    Oksi

  3. #3

    Re: Is this a plausible-sounding solo violin?

    Oksi-

    I\'ll leave the technical critique to others. I thought the cello particularly is good. Couldn\'t help thinking how great it would sound in a duet
    with another instrument-tends to hide any rawness.
    Of course, unless you\'re Yo Yo Ma even a live solo cello sounds a bit naked!

    Wonder why you use a wind controller instead of keyboard? More expressive, or just happen to have one?

    May the experiments continue...
    Carl

  4. #4
    Senior Member
    Join Date
    Jan 2002
    Location
    Dorset, UK
    Posts
    470

    Re: Is this a plausible-sounding solo violin?

    Originally posted by Oksi Moron:
    Sharmy,

    I\'m holding out for the ideal of having one sample patch respond as flexibly as possible to wind control, with maybe some minimal after-the-fact tweaking, as opposed to stringing together bazillions of pre-recorded \"articulations\" like letters in a ransom note. That\'s what drives my nutty experiments.

    <font size=\"2\" face=\"Verdana, Arial\">Absolutely.

    I can hear the timbral variation produced by your filtered layer. The biggest problem, as Sharmy sais, is the attack, which is crucial in our recognition of violin sound, especially solo. DDSV\'s strength is the slow natural attack, which gives it a lot of its realism. If you bypass the natural attacks you\'ll find some of the DDSV samples sound more like the mating call of the Siberian Crane. That is why some of your notes sound accordion-like. Relative to the sustain part, attacks contain more bow noise and \"altered\" harmonic response.

    The slow attack is also DDSV\'s weakness, as it makes it difficult to play smooth legato phrases or faster detache, without the attacks producing a wah-wah effect.

    When playing slower violin phrases the notes are linked, even with detache playing, as the bow is not released from the string, (unless required for expression or rest) and is may be necessary to overlap notes within a phrase a little to produce this; (-very difficult to do with a wind controller!) I\'m not even talking about full-on legato.

    I just recorded a short violin phrase to demonstrate but just can\'t get it onto my webspace (%&*~#@!). Can I refer you to my previous demo for now?, which also shows what I mean: That Bach thing

    The timbral change produced purely with changing bow velocity is subtle, but easy to pick out when it is wrong. (A lose - lose situation: nobody notices when it is right.) It is impossibly to reproduce this accurately with the filters availably in GSeditor. I think I nailed it?, using other methods. The dynamic variation in the demo is not great enough to show this off, but listen at 0:51 and 1:13. What do you think?

    Sharmy -what about you?

  5. #5

    Re: Is this a plausible-sounding solo violin?

    Originally posted by Beckers:
    I can hear the timbral variation produced by your filtered layer. The biggest problem, as Sharmy sais, is the attack, which is crucial in our recognition of violin sound, especially solo. DDSV\'s strength is the slow natural attack, which gives it a lot of its realism. If you bypass the natural attacks you\'ll find some of the DDSV samples sound more like the mating call of the Siberian Crane.
    <font size=\"2\" face=\"Verdana, Arial\">I agree, but I think what you\'re talking about is lopping the attacks completely off and then trying to envelope what\'s left. As in the DD \"fast attack\" articulations that became available a while ago, where the sample offset was set in as far as Giga will allow.

    That\'s not what I did -- although maybe the net effect is so much the same as to make no difference. Before I split each DD slow-attack sample into fundamental/upper-partial samples, I used Cool Edit to re-envelope the attacks, so that all but the first 30 milliseconds had the same amplitude as the peak amplitude for the whole sample. So in other words, I started out with way *more* bow crunch in the attacks than in the original sample, and in the sequencing phase, while trying to modulate the fundamental/upper-partial samples in a separated but a coordinated way -- trying, for example, to bring out a bit more brightness in swells than was in the original sample -- I wound up with even *more* bow noise in the attacks. So I tried to fix it with a combination of tapers and overlaps. Believe me, it would be easy enough to get the attack noise back in there -- more than you, I or Sharmy would ever want.

    That is why some of your notes sound accordion-like.
    <font size=\"2\" face=\"Verdana, Arial\">I know what you mean, but I think it\'s because I get the proportion of fundamental to upper partials wrong in places. At the bottom end of the violin\'s range there\'s almost no fundamental, but at the top end the fundamental is about as strong as the upper partials. One of the things I was aiming for was to introduce a differential between the degree of \"wiggliness\" of the fundamental\'s vibrato vs. the vibrato of the upper partials. It seems to me, looking at spectral maps of violin vibrato, that the vibrato depth gets deeper and deeper the higher up the overtone. I was going for a little bit of that.

    When playing slower violin phrases the notes are linked, even with detache playing, as the bow is not released from the string, (unless required for expression or rest) and is may be necessary to overlap notes within a phrase a little to produce this.
    <font size=\"2\" face=\"Verdana, Arial\">Well, that\'s what I thought I was going for by tapering the attacks and using so much note overlap (admittedly, while trying to cover up for the increased bow noise at the beginnings of notes). I didn\'t overlap both the fundamental and upper-partial samples to the same degree. Mostly I overlapped the upper-partial samples, because I\'ve noticed, while looking at spectral images of violin runs played all on one bow, that on some notes the fundamendal seems to drop out entirely; in other words, our sense of legato-ness may come from the continuity of the upper partials.

    I think I nailed it?, using other methods. The dynamic variation in the demo is not great enough to show this off, but listen at 0:51 and 1:13. What do you think?
    <font size=\"2\" face=\"Verdana, Arial\">I admire your efforts, and I\'m fascinated with the notion of layering a sample with a physical model. I have Reality\'s bowed-string model, and I\'ve futzed with programming it for my wind controller and then convolving the result with several violin-body IRs that I got from an Italian site. But I\'m a long way from getting acceptable results that way.

    The first time I heard your G-string rendering, several days ago, it sounded to me about on par, in terms of expressive variety, with the results I was getting programming for wind control in GS Edit. In particular I could hear discernable overlaps in some of the \"slurred\" transitions.

    Still, I defer to your judgment in all of this. You\'re a bowed-string player; I\'m not. I would love to keep in touch with you about this. Unfortunately, you need to maintain secrecy because you have in mind to market something. Me, I\'m strictly an obsessed hobbyist.

    Sharmy, thanks for all your observations. Everything you say is true, true, true. I knew about the too-uniform attacks in the Bach snippet (there was no modulation of attack time in my cello patch at the time), but I was just trying to get an idea whether you preferred examples that were rather more raw and up-front than the violin example. I haven\'t yet figured out how to put decent vibrato into the XSample samples, but I firmly believe that it can and will be done, and then the XSample folks will look like visionaries.

    Nicolash, I\'m a great admirer of Antonio Genovino\'s and I have an MP3 CD of his stuff, but I thought his cello piece was a good demonstration of just what we all (including Antonio) are up against.

    Carlsmith, thanks for your kind words about the cello piece. I have a keyboard controller, and I\'m not the least bit married to any particular technique. The violin experiment was inspired by a chapter in my CSound book, Andrew Horner\'s chapter on faking a French Horn sound using \"contiguous-group additive synthesis.\" His totally-synthesized horn sounded pretty reasonable to me

    http://www.stanford.edu/~gmsmith/mp3/horn.wav

    and that got me interested in adapting his enveloping techniques to split-up samples, rather than using purely synthetic wavetables, the way he did. Anyhow, I\'m open to anything that works. (And I\'m spread way too thin, experimenting with this and that).

    I understand that it\'s a bit uncool to post half-baked experiments in public and then turn around and say, \"Just kidding; I could easily have done better,\" but on the other hand I wish there were a forum for freshly hatched experiments of this kind.

    I can\'t promise not give you another rendering later. I\'ve got all the time in the world. When I was a teenager I tried to trisect the angle, too ;-)

    Oksi

  6. #6

    Re: Is this a plausible-sounding solo violin?

    Hi Sharmy,

    The links to Antonio\'s music work for me. However, you may like to try using Google with \"Antonio Genovino\" and \"the digital style\" in the search field to get to his web site.

    Best regards
    Nicholas

  7. #7
    Senior Member
    Join Date
    Jan 2002
    Location
    Dorset, UK
    Posts
    470

    Re: Is this a plausible-sounding solo violin?

    Thanks Sharmy. Can you elucidate on the sameness? You mean timbre, articulation, expression?. I agree about the vibrato which is a problem to manage.

    With your friend\'s Vitous thing, are you talking about the solo violin or string section?

  8. #8

    Re: Is this a plausible-sounding solo violin?


Go Back to forum

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •