[LAU] Convolution in LinuxSampler

Q lists at quirq.ukfsn.org
Thu Mar 4 17:25:18 EST 2010

Julien Claassen wrote:
> Hello!
>    sorry for crossposting, but I didn't know, where to best reach the 
> LS-wizards.
>    I wondered about convolution in LinuxSampler. The GS3 format certainly 
> supports in-sampler convolving (having their own files for that). I seem to 
> rememb3er, that working out the format was not the real issue, but the 
> implementation itself. But there I think Fons offered his convolution classes 
> - as used in jconv - to easily implement it.
>    As I now have more and more nice sounds, including even more handsome IRs 
> (either rooms or body IRs) I was wondering, if the convolution issue is still 
> persued. I certainly would very much appreciate it!
>    So could someone please give me an update of the way LS is going?
>    Warm regards
>               Julien


I'd like to turn the lone voice into er, well, at least a duet, if not 
actually a chorus :-)

Ever since I managed to get LinuxSampler working a while ago, I've been 
steadily buying more and more Giga instruments, pianos mostly, and find 
it a very useful tool.

I bought a few pianos in December in Gigasampler3 format for the better 
bit-depth samples, not realising that not all the features of GS3 are 
currently supported in LS.

They sound pretty damn good to my untutored ears, but I can't help 
feeling that I'm possibly missing out on their full potential and they 
could sound better. The manual certainly says that the .gsi file should 
be used for it to work properly -- LS doesn't recognise those and will 
only open the .gig files as far as I can tell. When listening very 
closely on headphones playing solo you can sort of hear that it's a 
sampled piano and not the real deal and I can't help feeling this would 
be lessened to some degree if the impulse response of the piano was 
being used.

Ultimately, it would be nice if all the features were supported, but 
from what Julien says, it sounds like implementing convolution could be 
a quick win, if it's just a question of implementing an existing 
convolution engine into LS. I'm not a programmer, so I've no idea 
whether or not it really is that simple, but if it is that easy, it 
would be nice to see the feature added to what is already great software.



More information about the Linux-audio-user mailing list