Midi timing suggestion on recorded VST instruments.

Isn’t this exactly what I said above?

But you leave out one part. If the midi data from a track played against a project is now moved forward (to the right on the timeline) to sync to the grid, if you play the project back using all the same tracks and VSTi you used to record in the first place, the midi data will now trigger that same VSTi late on the beat, no?

Is it most important for the midi data to look right or to sound right in the context of the project in which it was recorded?

I think everyone can figure out that answer.

Now, if you are recording multiple passes, each with a different VSTi, some of which have the highest PDC latency in the project and you want to create a new midi file of all the parts for export for other uses, then, yes, you will have to adjust all the tracks to the beat. And, for that purpose, presumably Cubase could keep track of some adjustment for each midi track to sync each one individually to the grid.

Como

I made a simple video as a contribution to the smart one’s that one day are going to solve this problem.

http://vimeo.com/38791238
Password: 2012

  • Given to readers of this thread.

Please tell me if I have violated any rules, and I will immediately delete this Video.
My purpose is only to make clear what I wish to be fixed since there is a huge confusion about this.

You have illuminated an issue that makes me feel a lot better about my playing. For years, I have thought that trying to compensate for latency has made me play everything early, as per the MIDI results…now I think perhaps not. Hmmm.

Thanks NYC fellow composer!

Looking at your specs that says, you’re using a fine MAC we can also presume that all the discussion about windows and timing and stuff is quite irrelevant here.

Well that’s it! Exactly what I am experiencing.

We’ll see. I’m going to try to reproduce tonight. How are you recording your audio-are you sending the track to a group?
Also, just for consistency, what sound card are you using and what was your buffer setting when you did the video?

VERY informative.

And, just to be clear …

If I’ve understood correctly, the reason your playing was on the beat (the click) - in spite of latency - was that you were doing what would be expected of a musician, and playing ahead of the click to make the instrument sound together with the click?

Some organists have to do that all the time, in effect always playing with tens of miliseconds latency - eg, apart from a lag between the pressing of a key and the corresponding pipe(s) beginning to sound, the organ console can be tens of feet from the pipes - meaning tens of miliseconds just for the sound (once started) to reach the organist’s ears. It’s a skill acquired through playing (eg pianists transferring to the organ learn to adjust to the lag and can then play as if there were no lag), and …

… as this thread shows, this skill is clearly just as relevant to DAW users.

Unfortunately, apparently, Cubase currently seems fixated with logging a MIDI note at the instant when it’s played, rather than at the moment when Cubase outputs the start of the sound of the note from the VI. For those players who have acquired the skill to make the VI’s sound co-incide with other audio, in spite of the latency, it would be useful if Cubase could delay the placement of the MIDI note to the instant when the sound of the note starts. Then (if I’ve properly understood), for playback, the normal DAW methods of compensating for the different latencies of different VIs (and FX) would ensure that the MIDI notes would then play in time with what was already recorded (instead of early, as in the video).

1 Like

Because it must. First the note input and then the sound output. If it logged it as the sound you hear begins, then when you played back what you just recorded, your playing would have to be behind the beat of what you had heard while recording.

Read this carefully and think about it: Midi timing suggestion on recorded VST instruments. - #74 by chase - Cubase - Steinberg Forums

I don’t disagree that Cubase could and should track the PDC to make adjustment available. But I do disagree that the default behavior should be any thing different than it now is.

I watched the video and the behavior is expected. The slight difference between the playback of the VSTi against the click is the difference between where it should be to remain in sync with the project, i.e., click, and the playing.

And, I’m done posting on this topic.

Good luck, all!

Como

Say, if you use an external synth or keyboard, using that sound to listen on while recording on your VST instrument track, then I agree with you Como. And it has to be said, when recording on midi tracks routed to outboard midi gear, it shouldn’t be different from now.

But regarding VST instruments - I might think even the default behavior should be different from what it is now.
However, if they would change the default behavior or not is not so important to me, I could live with “activating” the functionality for this. I’d probably celebrate it with sparkling wine and have activating cake with my friends.

This explains a little about how the test was done:
I don’t think it’s so important for the matter but it
could be useful if you’d like to test for yourself.

  • First I created a VST instrument track, a group channel, and an Audio track.
  • On the VST instrument’s output channel I activated a “send” (copy) to the group channel.
  • Then I set the audio channel’s input to the group channel .
  • I record enable both the VST instrument track and the audio track.

This test is made using a Frost Autumnleaves PC and the soundcard is a Focusrite Saffire Pro 26 I/O. Windows 7 64 bit running Cubase 32 bit. The latency on the soundcard is set higher than usual to give a more clear picture of the problem: The ASIO Buffer size is set to 40 ms giving me an Input latency on 41.6 ms and an output latency on 45.5 ms.
I have to say though, that even on my soundcard’s smallest ASIO buffer size setting, 4 ms giving the input Latency 5.6 ms and the output latency 9.5 ms, this is a problem too. I wouldn’t bother to lay time on this thread if I didn’t feel that way. Notes still doesn’t playback accordingly with how it was played and all data seems to be placed a bit too early in comparison to the “take”.

@Joel: I watched the video and I can confirm that the scenario you shown, I have been plagued with that for many years. You learn to get by, but this needs some developers attention for sure. No matter what system MAC/PC the same results have been like that for years. Back in the SX2,3,4,5 days, Many keyboardist would complain to me about this very situation. You can fix the notes but, why should you have too…

But with the buffer substantially lower, the amount you’re off is much less, yes?

Yeah Joel! Thanks for the video. Now, somebody earlier said:

“I watched the video and the behavior is expected”

Man, did you watch the same video? Do you expect the sequencer to record with bad timing? Do you expect you have to move your performance EVERY single take you make? Jeez…

Depends what he saw. Maybe you missed something. I mean to put that into perspective, for myself I can sometimes worry about a brick without noticing that the whole wall is wonky.
I think now, and lean towards Como, that this is a quantisation problem more than a latency problem.
First off though, not all Cubase users have to move their performance after each take, certainly I don’t and if I do it’s usually because I played it that way.
Check the quant settings and see that they are exact. Either you may have made changes to the templates (while experimenting with a new Cubase) or the quant templates you have downloaded are not perfectly set. ie: 1/16 may have an offset set for a looser feel and as people frequently overestimate their timing talents (mine are extremely good but not as good as I think all the time) any discrepancies are compounded and very annoying to correct.

Conman,

With all respect: If at this stages of the thread, and i know you´ve been active on the other one about this in the forum too, you come here and talk about quantization i think i am gonna consider you a troll buddy.

So :wink:

Unless a tornado draws me back into the discussion again,
I’d like to say just a few more words.

The video example is recorded to showcase the problem,
and is therefore performed with a higher latency than usual.
Maybe there was some misunderstanding earlier but
to make clear, I don’t use any quantize there.

When recording VST instruments with lower latency, the problem shrinks accordingly,
but doesn’t disappear. I often have to move forward stuff to compensate.
Especially when it comes to live playing and I don’t want to quantize,
but also at times when I intend to quantize right.

Seems like there’s a bunch of people out there who’d love to see this work properly.
It could be a setting in the preferences at least. Maybe most people wouldn’t bother,
while others would see it as kind of groundbreaking.

Let’s see what they say at the support.

Thx!

/ Joel

Hi All

It’s a sticky little problem this, isn’t it?

The main issue that is blocking some readers from fully grasping what is going in is, as I stated earlier, that Cubase IS COMPENSATING FOR LATENCY ON PLAYBACK. By this I mean, if you step imputed a bunch of notes on the grid, it would play in time with the click. Is this just by magic? No. Cubase plays the midi data EARLIER THAN DISPLAYED IN KEY EDIT to compensate. It’s in the manual (somewhere).

Now, this (MIDI PDC) is clearly impossible during a MIDI recording. (otherwise Cubase would have to predict what you are about to improvise). The problem arises as the musician naturally played early to compensate. But Cubase then ADDS ITS COMPENSATION on playback once the recording is done and a midi part is created.

This is really the error going on here. Think about step imputed notes and why they do not play late.

What Cubase should really do is, once the recording is done, shift the midi DATA forwards (ie later) by the exact same amount as whatever it’s PDC engine was set to during record. In effect, Cubase should recognise that any midi events arriving at the midi port are early as a result of the player playing in time with what was audible at the time of record.

I really hope that Steinberb will pay attention to this very long standing and massively time consuming design flaw. Yes perhaps that’s the word to use if SB don’t recognise this as a bug.

Thanks. BB

Still there are some users claiming that they don’t have this problem.
That makes it so confusing.
I really want an official statement on this matter.

There is an article:

https://www.steinberg.net/en/support/knowledgebase_new/show_details/kb_show/midi-issues-on-windows/kb_back/2025.html?tx_p77sbknowledgebase_pi1[keyword_search]=MIDI

In this article Steinberg discribes a totally other scenario with this as possible outcome:
(taken from the article)

Shifted MIDI events while recording (events are recorded too late or too early)

If this is true then the OP and all the others like Joel, goodbyenine, Edcubasero and me (hope I didn’t forget someone) are totally wrong and this is a mididriver issue.

So what the hell is going on here??
Please please can a moderator chime in and make an end to this discussion?

Greetz Dylan.

FWIW
When I do the test and use an external apreggiator (not synced by Cubase), to input the MIDI data (to avoid human correction) the outcome is the same as the video. MIDI roughly between 700 - 800 (yes it variates) samples (800 / 44.1 = 18 ms) earlier then audio on a 120BPM project with an ASIO buffer size of 512.

they already chimed in in the other thread. this is Marcus reply:
We don’t have found any misbehavior at this moment. The UAD- Cards and VST Plugins don’t show any issue related to the Midi- Recording.

And Chris just added this link:

https://www.steinberg.net/en/support/knowledgebase_new/show_details/kb_show/midi-issues-on-windows/kb_back/2025.html?tx_p77sbknowledgebase_pi1[keyword_search]=MIDI%20Treiber


So, they didnt even recognize this as something related to Cubase. They didnt bother to ask for specs to the people who suffer from this either. I sincerely doubt they are ever gonna admit its a Cubase bug.