[LAU] there's a new salamander piano version in town.

[Leigh Dyer]

On 08/17/2011 07:45 PM, Philipp Überbacher wrote:
> Excerpts from Leigh Dyer's message of 2011-08-17 01:08:42 +0200:
>> On 17/08/11 08:55, Philipp Überbacher wrote:
>>> Excerpts from Leigh Dyer's message of 2011-08-17 00:29:28 +0200:
>>>> On 17/08/11 07:05, alexander wrote:
>>>>> http://rytmenpinne.posterous.com/65930001 Follow my salamander drumkit
>>>>> progress ;)
>>>>
>>>> Awesome -- I'll be following your progress closely! I definitely like
>>>> the idea of having a few separate SFZ files, too, to help with mixing;
>>>> getting separate drum outputs out of LinuxSampler is something that I've
>>>> had to hack my way to in the past with other kits.
>>>>
>>>> Thanks
>>>> Leigh
>>>
>>> Now I know what you mean by multisource, but I wonder: is
>>> sfz/linuxsampler not flexible enough to achieve this with a single sfz?
>>
>> You can -- in my last project I had a LinuxSampler session where I'd
>> simply loaded the same SFZ five times, with each instance going to
>> separate JACK outputs. That meant that I had to split my MIDI data
>> across five separate tracks so that I could send the kick drum to one
>> set of outputs, the snare to another set, etc.
>>
>> The other thing to consider is multiple mics. My scenario above is
>> simplified a bit, because I actually had *ten* LinuxSampler instances --
>> the kit I was using came with close-mic and room-mic recordings, so I
>> set up an instance of each for each drum so that I could balance the
>> levels between the two versions independently for each drum.
>>
>> Having multiple SFZ files available doesn't instantly solve these sorts
>> of issues, but it does give you some options around how to handle them.
>>
>> Thanks
>> Leigh
>
> Ugh, that sounds like it would be tough on the memory. Loading a sfz
> just once and routing different parts to different outputs would be way
> more ideal (even without level control because you could use an external
> mixer then).

The memory usage didn't seem to increase at all with multiple instances 
of the same sample set -- I'm guessing LinuxSampler's streaming engine 
is smart enough to share cached sample data amongst those instances. The 
kit I was using is 900MB in size, and LinuxSampler's memory usage never 
seemed to go much over 300MB.

I'd definitely prefer for that case if LinuxSampler could direct 
specific notes to specific outputs -- that's exactly what I do under 
Hydrogen, for example, thanks to its option per-instrument JACK ports -- 
but I don't think it's capable of that.

Thanks
Leigh