Re: Re: Ninjam techbabble (was: Chinapainting article)

[rs@moinlabs.de]

>I don't want to come off arrogant but when I listen to the examples of 
>online jam sessions (Ninjam)  they sound horribly out of >sink?

You're not coming off arrogant, merely ignorant.

The short answer: if the playing on Ninjam's public servers sounded 
completely untight or out of sync, this may either have to do with the 
reason why most amateur punk bands sound out of sync (btw, really liked 
your misspelling there), namely because of lack of qualification, or 
because a bad tech setup, or because they follow an approach for playing 
that makes the experience for a third party listening in less enjoyable 
(see explanation below).

The rest of the message is rather long-winded...

Ninjam has this concept that it "rounds up" transmission delays over the 
net in a way that make musical sense iff you play in time and have the 
interval set to a meaningful value.
So for this to work (for beat-based music), the musicians need to know 
what Ninjam's tempo and beat offset is. There are several possibilities, 
all of which are slightly less than optimal:

1. Use the metronome. Downside: the metronome is output on the same audio 
output as the music, so you can't (or at last most probably wouldn't want 
to) use this approach when playing to an audience.
Solution: Os has developed this VST version of the ninjam client which 
features separate outputs. Unfortunately, no public version exists. It 
should be sufficient if one beat-oriented player (e.g. drummer) listens to 
the metronome, the others can just play along with him.
I never did that myself.

2. Use MIDI clock. One of the clients (the wasabi client) does output a 
MIDI clock based on Ninjam's time. Downside: the wasabi client won't 
handle ASIO interface. Workaround: use two computers, one of them running 
the normal client which handles audio and one of them running the wasabi 
client which outputs MIDI clock. Problem here: you have a timing offset, 
which will hurt you iff any of the other participants use either the same 
approach or the metronome. Workaround: have your MIDI-synced equipment 
setup so it compensates for that offset (all sequencers and drum machines 
I know can do that). Again, only one (best the "beat source") would need 
this cumbersome setup.
Examples: Both the tracks "Aspirin Age" and "Virtual Baggy Pants" on 
http://www.moinlabs.de/i_kqpda.htm were done using this approach.

Other approaches:
3. In a situation with only two participants (note that in this 
discussion, a "participant" is one client - which can very well feed the 
playing of more than one musician), one participant can for some time 
establish a groove and ignore what he hears from the other side (or at 
least not let himself be affected by its timing). The other participant 
can then simply play on top of that. This sounds good (meaning: in sync) 
for that other participant, not so much for the first one (or for anyone 
else who just connects to the server and listens).
Examples: the groovy part at the end of "The Milkey Way - Yivli Antare". 
Here, Charlie played the groovy bass part (not necessarily in sync with 
Ninjam's tempo), and I played on top of that.  "Advantages in Freeform 
Breeding" - here, both Rick Walker (percussion) and Krispen Hartung 
(guitar) were one participant, I was the other participant.
In both cases, the album uses the recording done on my side.

4. Also, you can simply ignore any beat (works best if playing 
non-beat-based music). This is the approach that was afaik used also 
during Warren's and Per's session at Y2K6 kyberloopfest. The funny thing 
here is that this sounds completely different (not necessarily in an ugly 
way) at the different participants'.
Examples: the rest of "The Milkey Way", "Below this Sneppah". There will 
also be a track on the second volume of the kybermusik series (due 1st 
half of 2008) where I'm gonna combine the independent recordings of each 
participant from a session I did with Tony K.

Things that can still go wrong:
   * as I said, when using the "two computer MIDI approach" AND (somebody 
else is also using it OR somebody else plays to the metronome), you have 
to compensate between the timing offset of both your computers.
   * in example 3, this normally only sounds good in exactly one place.
   * Ninjam doesn't compensate for latency induced by audio interfaces 
(although it could - this would be a nice thing one of you software 
developers could do).
   * players can be unable to play in sync
   * what is beautiful for one might be horrible for another one

BTW, all you software developers out there (including but not limited to 
Os):

I'd still like a Ninjam client OR a VST plugin (best both) that has the 
following features:
    * ASIO support (only for the standalone client) (priority 1 - this is 
already available in the normal client)
    * can output MIDI clock (priority 1 - this is available in the wasabi 
client)
    * has an additional setting for audio hardware latency to compensate 
for that (priority 3)
    * independent routing of audio output including metronome (priority 2)
    * running on XP platform (priority 1)

Looking forward to your implementations!

                Rainer