I\'m planning on starting an acoustic guitar library for fun today, and perhaps could use some tips before plunging in. It\'s my first time creating any instrument. My plan is to start recording into soundforge to create wavs for each note, then creating the gig from there. I\'m probably only going to look at the manual if there are problems, but if anyone advises that I\'d be saving myself grief let me know. Here\'s the plan- all single notes, no chords, from notes A (4 steps below normal low E string) to perhaps A an octave and half above the high E. I plan to also record all octave harmonics for all these notes, excluding probably a few of the highest notes. Recording will be in the basement with shure sm81s in XY setup -> dat for a/d -> soundcard. It\'s probably going to be a pain to add complexity, but I\'m very curious about doing release samples as well and this is not intuitive to me. Is this appropriate in my case, and how would I record the release on a note?
Any suggestions that might help or save time for a newbie are plenty welcome. Thanks-
Be very careful when saving the WAVs with Soundforge to specify the unity note.
If u fail to do so and leave them all to default C4, Gigasampler will get \"confused\" when reading the WAVs.
Also, be careful with the instrument editor, it tends to crash when doing heavy editing, so save your work often.
Normalizing with SoundForge ... I don\'t know if all the samples should be normalized, maybe only the low-velocity ones. Can someone tell me what they\'ve experienced? I know the debate is pretty harsh about normalizing.
That\'s it. Good luck!!!
1. As far as I know, Soundforge still only saves at a max. 16-bit resolution. You\'re editing at 24-bits, then either truncating (loosing the detail) or adding dither to try to capture the higher resolution. This is not a big deal if you\'re careful. Just try to do all of your editing, normalizing, etc. then save once. Multiple saves will cause deterioration of the wav. GS will only accept 16-bit .wavs, so I recommend recording the file, trimming, EQing (if necessary), normalizing, add dither, then save to 16-bit.
2. You are recording the instrument at multiple velocities with special effects. I would recommend creating folders for each velocity (e.g. v1, v2, etc.), harmonic, and release. When you save each sample, develop a naming scheme like \"a#1 va.wav\", \"b2 vb.wav\" where va and vb are your different velocities. I had problems when I named the files \"a#1 v1.wav\". I think the GS Editor Wizard has problems with a letter followed by a number, interpretting both as notes rather than a note and velocity. Nemesys says it\'s not a problem, but like I said I experienced problems with it. Save all the same velocities in the same folder, as well as the harmonics and release noises. If you do it correctly, it\'s going to be easy to run the wizard and mapping will be a breeze.
3. Normalizing. I recommend normalizing all wavs. Think of it this way. If you recorded your instrument at only one velocity you would normalize it and GS would play it back at different velocities correctly. For example, if you played it at MIDI velocity 127, the file would play with 0db attenuation. If you played it at MIDI velocity 64, it would take the same fully normalized wave and play it back at -Xdb attenuation. But it would sound right because GS is attenuating the sample according to its velocity scale. Now if you have multiple samples, I believe they each should be normalized. Let\'s say you recorded a quieter note and didn\'t normalize it so it\'s down by 10db. You map it to trigger at MIDI velocity 64. GS will play it back at using its velocity curve so it will playback at -Xdb + -10db, which will probably sound too quiet relative to louder notes. Now GS gives you a lot of options for velocity scaling, plus you can easily grab a velocity range and raise/lower it by Xdb, but normalizing all of your samples does work.
4. To capture your release velocity, you\'re just trying to capture the noise that occurs when the note is released. These can be subtle, but add to the realism. I don\'t know if you should try to capture the noise when you play a loud note, then quickly release it or let the note fade away and pull up on the string. You\'re going to want to trim that noise away from the note and make it its own wav file which will get triggered when you let up on a note in GS. Save the release notes in their own folder.
5. Experiment with one note. Capture it at all the velocities, plus the harmonic and release noise. Edit the wavs and bring them into the GS editor. Get a feel for mapping and make sure you\'re getting the sound you\'re after.
It might also help to record the different velocities on separate passes. Do all the pp, then all the mf and so on. That way, you can turn up the gain for the soft notes and not have to normalize them as much. This will result in quiter recordings. You may find yourself adding noise reduction to the softer velocities also. They may sound quiet on thier own but as soon as you start holding down the pedal and strumming chords, the noise may build up. For the release trigger, you can try a real loud staccato mute of each note. Pluck real hard and quickly mute and let the resonance of the guitar ring out. Then you can cut off the attack and put the tail as a release trigger. This will requier adjusting the release and attack envelopes of the sustain and release trigger dimensions in the Editor. Good luck and have fun.
Thanks for the tips everyone. This is definitely helpful. I finally got to start this project today (I\'d lent my good mics earlier). I made a trial run with all heavy velocities and learnt the basics. I actually screwed up building the instrument first time round, with interesting results. I had unity note preference to middle C(without knowing) and when I imported the notes and saved the gig, it worked. I think it only used one, or maybe up to 3 of the samples and created the full range anyway. The pitch shift is MIGHTY good if this is the case. I didn\'t even notice anything was wrong at first, until listening more carefully- chords seemed slightly phasy and single notes had a bit of min6 overtone. I figured out what was wrong and changed import preferences. The samples are not so consistant in quality now, but the above problems are gone.
A few folks discussed normalization... It makes sense to me. What do people think of peak level normalization vs. RMS(root mean square)? I hate to think of subjecting my samples to compression(rms), but I think rms would give more consistant results.
One other thing.. fretted and open notes become a consistancy issue when sampling. The open notes by nature are brighter and have a longer decay. Likely on my next try I\'ll use capoes to avoid ever fretting - the constant fine tuning won\'t be much fun though.
sone little tips from a professional library producer:
record the sounds raw 2 wav, make selection of the best samples of every keyrange, put anything in a non destructive HD system on the timeline, adjust all samples to the same LOUDNESS with a good limiter/compressor (forget simply normalizing, because the one sample may habe a peak and the averange lavel might be -12 db while another sample is 12 db louder),make necassery EQ-ing to make it sound homogenious and re-record this 2 a new file. For GS, make the samples as loud as you can, because the output of GS is very low. Again, it is important that all samples have the same loudness, so take the weakest sample, make it as loud you can, and adjust all other samples due to this reference.
I did an acoustic guitar sample once (an old guitar so the final sample is not that good). What I did whit the open strings is the following:
-except for the low E I sampled all notes on the frets (so no open notes). Ofcourse I sampled every note individually since GS has no memory limits.
-I then samples the open E A D G B E and put them on a sustain pedal layer. So what happens is as soon as I press the sustain pedal I get the open notes (when playing one of them), otherwise I get the fretted onces.
I\'ll try to make this available to you or anyone for close to free. Downloads seem out of the question though since I expect it will be ~300 Megs uncompressed. Maybe a small check for me to mail a cd would do, once it\'s complete.
Unfortunately... I\'ve been working on other things so this project has been on the back burner. With the reminder and interest, it\'s something I\'ll try to do soon.
I\'ll post back here when it\'s done.
The approach I plan to take is to record all notes as (simulated)open stings by using a capo extensively. I\'ll do it with only the picked string ringing, all others damped with my fingers. As far as having the sustain pedal down, that\'s the really intesting part to me..
I want to create the resonance effect when all the strings are allowed to ring along with the note played. I\'ve only heard about doing it by making another set of samples for the undamped notes. Instead of doing this I\'m going to try to use a magnetic pickup(Sunrise). I\'ll play the non-resonance samples for each note through one monitor with only the originally recorded string damped, and record the output from the pickup. Then I\'ll shift the start points of the pickup samples a few milliseconds as required to compensate for the distance of the monitors to guitar strings. The end result is that when playing, the damper pedal would trigger both the original sample and the correspoding resonance sample at the same time. This way there\'s not much for me to record (and a smaller disk footprint), and alternate tunings are easy by making another resonance set. The only problem is phase accuracy(I think when I did calculations the speaker placement needed to be accurate to within a half inch for the resonance samples to have accuracy to 10-15kHz) and how easy it\'s going to be to layer the sounds accurately in gigasampler. Anyway, looking forward to it.