I'll make those changes and send a new pull request to the sfz branch. What values would you like me to use for attack and release? These are just default values. Ultimately you'll want to customize them for each instrument, but right now I'm just trying to auto-generate files, and I don't know of any automated way to pick appropriate values for different instruments.
The main purpose of the volume correction is to make all the samples for each note consistent. Otherwise the volume jumps when it goes between layers and the round robins sound different from each other. I also do a little bit of normalization to smooth out differences between adjacent notes, but that's just a local correction. It doesn't change the balance between high and low registers. Unfortunately, the largest allowed value for "volume" is 6 dB, and some of the samples are already hitting that limit. If you want the recorder to be louder, I think you need to actually amplify the samples. Some of them are super quiet.
Is there any way I can tell which convention is being used for each instrument? I could try to figure it out by analyzing the sound, but that won't be completely reliable. The routine for calculating frequency sometimes picks up on a prominent harmonic rather than the fundamental.
I might try crossfading if I'm feeling really ambitious! But I suspect that without phase locking, that's not going to work too well.
Is it ok if I rearrange/rename the folders a little so that whenever there are release samples, they'll always be in a folder called "Releases" and the corresponding sustains will be in a folder called "Sustains", with subfolders for the articulations? That's how some instruments do it, but others arrange or name them a bit differently.
There's one case I'm not sure how to interpret. The Steinway grand piano has three folders called "NoSus", "Sus", and "Rel". What's the difference between "Sus" and "NoSus", and which ones are the releases meant to be used with?
The main purpose of the volume correction is to make all the samples for each note consistent. Otherwise the volume jumps when it goes between layers and the round robins sound different from each other. I also do a little bit of normalization to smooth out differences between adjacent notes, but that's just a local correction. It doesn't change the balance between high and low registers. Unfortunately, the largest allowed value for "volume" is 6 dB, and some of the samples are already hitting that limit. If you want the recorder to be louder, I think you need to actually amplify the samples. Some of them are super quiet.
Quote: - Piano is 1 octave too high (note that most instruments are C3=60=middle C, only a few, if any, are C4=60)
Is there any way I can tell which convention is being used for each instrument? I could try to figure it out by analyzing the sound, but that won't be completely reliable. The routine for calculating frequency sometimes picks up on a prominent harmonic rather than the fundamental.
I might try crossfading if I'm feeling really ambitious! But I suspect that without phase locking, that's not going to work too well.
Is it ok if I rearrange/rename the folders a little so that whenever there are release samples, they'll always be in a folder called "Releases" and the corresponding sustains will be in a folder called "Sustains", with subfolders for the articulations? That's how some instruments do it, but others arrange or name them a bit differently.
There's one case I'm not sure how to interpret. The Steinway grand piano has three folders called "NoSus", "Sus", and "Rel". What's the difference between "Sus" and "NoSus", and which ones are the releases meant to be used with?