Indeed!
Greg
GSnap struggles sometimes...
-
- KVRAF
- 12977 posts since 29 Sep, 2003 from Ottawa, Canada
-
- KVRist
- 188 posts since 13 Jun, 2004 from SW England
Looks like I won't have any time this evening, but I'll definitely have a look this week. I've been working on some other things lately so it'll be good to go back to GSnap with fresh eyes.
-
- KVRist
- 188 posts since 13 Jun, 2004 from SW England
Well, I've done a bit of tweaking. I'll skip the long-winded, technical explanation and just ask you to give it a go Mr Boogooloo, or anybody else for that matter... (handy link).
-
- KVRist
- 188 posts since 13 Jun, 2004 from SW England
Sorry it's taken so long to reply. Here's the long-winded, technical explanation:
Right, the problem was in the pitch-detection stage. To detect pitch, GSnap basically "looks" at the waveshape and tries to spot the repeating section, i.e. the fundamental period. (Obviously GSnap doesn't have eyes, so it uses maths instead).
To avoid too long-winded an explanation, I've made a couple of pictures. Below is a picture of Simon singing an F. I've highlighted the fundamental period.
Now, the problem arose because of the strong second harmonic. The second harmonic is at twice the fundamental frequency. You can probably see what I'm talking about, but I've highlighted it below just to make sure.
So, in short, GSnap was getting confused between two frequencies an octave apart. Given the indecisive output of the pitch-detection stage, the pitch-correction stage just sort of chewed on the signal a bit.
Bringing down the "Max Freq" solved the problem because it removed the incorrect, higher frequency from consideration.
I've improved the pitch-detection algorithm now, so GSnap should cope better with more inputs. Thanks to Simon for the encouragement
.
(Incidentally, if you like the look of Simon's strong second harmonic, then I'd recommend you follow the link in his signature and have a listen.)
Right, the problem was in the pitch-detection stage. To detect pitch, GSnap basically "looks" at the waveshape and tries to spot the repeating section, i.e. the fundamental period. (Obviously GSnap doesn't have eyes, so it uses maths instead).
To avoid too long-winded an explanation, I've made a couple of pictures. Below is a picture of Simon singing an F. I've highlighted the fundamental period.
Now, the problem arose because of the strong second harmonic. The second harmonic is at twice the fundamental frequency. You can probably see what I'm talking about, but I've highlighted it below just to make sure.
So, in short, GSnap was getting confused between two frequencies an octave apart. Given the indecisive output of the pitch-detection stage, the pitch-correction stage just sort of chewed on the signal a bit.
Bringing down the "Max Freq" solved the problem because it removed the incorrect, higher frequency from consideration.
I've improved the pitch-detection algorithm now, so GSnap should cope better with more inputs. Thanks to Simon for the encouragement
.(Incidentally, if you like the look of Simon's strong second harmonic, then I'd recommend you follow the link in his signature and have a listen.)
-
- KVRAF
- 12977 posts since 29 Sep, 2003 from Ottawa, Canada
Cool, thanks for the explanation. 
-
- KVRist
- 188 posts since 13 Jun, 2004 from SW England
Incidentally, GTune and GSinthG both use the same pitch-detection code so they're both improved too.
-
- KVRAF
- Topic Starter
- 3971 posts since 19 Apr, 2005 from Brissie
Hi gy/guys,
Using the latest version, I've hit another snag - here's www.simonwatson.com.au/songs/voxnogsnap.mp3 and www.simonwatson.com.au/songs/voxwithgsnap.mp3
As you can hear, I'm singing a note which wavers a bit around C#, and GSnap doesn't like it one bit!
Again, no need to tell me to sing in tune - in context, it's only one note, and I really like the performance, so I wouldn't want to sing it again, if I could help it.
Can some other kind folk give GSnap a go on the dry file, and see if the same problem occurs? The reason I ask is that I've JUST started having my first glitches on my new computer, and part of me wonders if it's not GSnap's fault...
Much appreciated!
Watto
Using the latest version, I've hit another snag - here's www.simonwatson.com.au/songs/voxnogsnap.mp3 and www.simonwatson.com.au/songs/voxwithgsnap.mp3
As you can hear, I'm singing a note which wavers a bit around C#, and GSnap doesn't like it one bit!
Again, no need to tell me to sing in tune - in context, it's only one note, and I really like the performance, so I wouldn't want to sing it again, if I could help it.
Can some other kind folk give GSnap a go on the dry file, and see if the same problem occurs? The reason I ask is that I've JUST started having my first glitches on my new computer, and part of me wonders if it's not GSnap's fault...
Much appreciated!
Watto
I've joined Lurkers Anonymous.
-
- KVRAF
- 12977 posts since 29 Sep, 2003 from Ottawa, Canada
Are you using automatic correction? Drop into MIDI correction for the one note and then back into auto, if so!
Greg
Greg
