> I'm also a bit puzzled by people complaining about jitter. I don't have
Cubase is particularly bad when playing a soft-synth live, esp with larger
audio buffer sizes, because even though VST supports sample-accurate MIDI,
all note-ons are sent with timestamp of zero (the exact start of the
It's like trying to play drunk, like glue in the keys, I keep looking at my
fingers thinking "did my finger slip off that note?".
Playing a pre-recorded MIDI tract is different, timestamps are then
Why did Steinberg implement it like this?, I think it's a misguided attempt
at reducing latency. It's doesn't, the worst case notes are still delayed
exactly one 'block' period. There's no upside.
It's far better to have small latency and no jitter because your brain will
compensate very accurately for consistent latency, you will instinctively
hit the keys a fraction early. All will sound fine.
Jitter is baked-in timing error, once it's in your tracks you can't get it
out. Latency can always be compensated for and eliminated later.
The right way is to timestamp the MIDI, send it to the synth delayed by one
block period. Since audio is already buffered with the same delay, you will
get perfect audio/MIDI sync.
IMHO - After writing my own plugin standard, sample-accurate MIDI is no more
difficult to support than block-quantized MIDI.
> Message: 8
Linux-audio-dev mailing list