Steven Curtin wrote about my realtime midi version of csound:

> Cool.  I'm starting work on a similar project that will include 
> DirectSound support.  

Well, I don't have the luck of having the version 4.2 of MS Visual C++,
so I don't know
the programming features of DirectSound. I read the following text about
DirectSound
in the "mellosoftron" shareware program help (mellosoftron is a software
sampler that 
works in a similar way as csound):  

"The low latency feature is only really there for static sounds, i.e.
sounds stored in their "entirety in memory buffers and blitted out on
demand. In order to use such buffers, you must "give up envelope shaping
and looping (except for the trivial case of repeated playback of the
"entire buffer). You must also give up decent resampling, since all that
DirectSound does is "change the output rate (no interpolation). The
resulting sound quality is horrible, totally "unsuitable for music.
"
" The last problem could be overcome (given enough RAM) by
pre-resampling the sounds and storing "them separately for each note,
resulting in something very similar to the old Mellotron, but "this is
hardly worthy of being called a "solution".
"
" If you want to do envelope shaping and decent resampling with
DirectSound, you must use "streaming buffers. A look at the
documentation shows that you must then write ahead by at least "1 second
in order to avoid breakups. Needless to say, a latency of 1 second is
quite "unacceptable for live applications."

With my Pentium 133 PC the mean latency value is about 1/5 of second
(still too much big
for my taste and it don't use DirectSound)

Do you know if is it possible to eliminate latency time in realtime
sound processing 
with the newer versions of DirectSound?

Bye

Gabriel