rendered audio from midi out of sync in cubase and reaper

[DevonB]

This sounds very close to, but not exactly, like a persistent Cubase midi problem (known bug actually) that persisted up until Cubase 6. This caused me lots of grief, but it was a very real problem that I had to address with every new Cubase install up to v. 5. It may help, it may not, but it's worth a look. I originally posted about it here: http://www.kvraudio.com/forum/viewtopic ... idi+timing

Best of luck
-B

Nice advise, thanks.

Devon

[tonytee335]

I have now uploaded a video to youtube demonstrating the aforementioned issues. It can be viewed here:-

http://www.youtube.com/watch?v=RHfYsgMl ... e=youtu.be

Thanks to everyone who's posted constructive help and advice, I really appreciate it. I have tried everything suggested here and nothings worked, its not a latency problem, as you will see on the video, the midi playback is fine, but the timing of the render is all over the place.

I find it amusing though that I'm the one who has been labelled hostile when the first reply to my post basically said I had crossed over to the land of absurdity. Is that how you normally respond to new members just looking for a bit of help? I'm not flustered, just surprised by the hostility I received when trying to solve a problem.

[tonytee335]

'You also say it happens from the minute you start with the most basic setup/process possible.

Are you saying that if you were to get given a brand newe PC with nothing on it, then installed an operating sys and then Cubase/Reaper or anything else and recorded something with any of your VSTi then renedered it to audio you would have the same problem?'

Yes, That is exactly what I have done, the pc that I did the first lot of tests on is brand new, I installed all of my software on it, then backed it up, and restored the factory settings, which is basiclly just preinstalled with Windows 7 professional, basic stuff that Scan the builders put on like Acronis, and Nero. I installed Cubase 5, Pro Tools 10, Reaper, Audio Interface drivers and the plugins. Same on my laptop, restored back to factory settings, and the above software installed.

[SJ_Digriz]

I have tried everything suggested here and nothings worked, its not a latency problem, as you will see on the video, the midi playback is fine, but the timing of the render is all over the place.

You've tried everything but sharing one of your project files to see if anyone else on the planet can duplicate the same problem with the same project. Link up a Cubase project.

And why did that video look like we were hunting for bigfoot?

[bazwillrun]

I've watched around 40mins of the vid.
Nothing here makes sense whatsover, there is no logic here.

sorry to repeat myself but I cant see anything else to do, except as others have said, post a project and lets see if its repeatable and manifests itself to any of us.

post the ones you did in the vid.

at the very least we should by a simple process of elimination be able to eliminate factors to leave ones that are common across all equipment and software, and like any scientific experiment its only fact when its reproduceable and repeatable on demand.

you are saying it happens virtually all the time no matter what combo of software, hardware, interface, Vsti etc etc etc.

Having seen the vids from what i can see its milliseconds, is it audible ?
In cubase cant you quantize the audio instead of cutting and repositioning.
not being funny, but are you maybe worrying over nothing ?

I'm no expert in any of this but if it was some other subject, the same methods to track down the problem would apply.

i'm at a loss

[BertKoor]

Milliseconds off... That sounds like the midi events are only occurring at the boundaries of the audio buffer. Fair enough...

[whyterabbyt]

1) you're predominantly talking about what you can see. what can you hear?

2) you're predominantly talking about what you can see when you zoom in. exactly how far are you zoomed in, ie what sort of time issue are you actually talking about? a few samples? a few milliseconds?

[bazwillrun]

dont know if you guys have watched the vid, but to me it looks about as zoomed in as you can get and thats why I asked if he could here the "problem" because to me it looks like milliseconds and seems to be pretty academic.

"but the timing of the render is all over the place" is the description hes used in an earlier post.

i just recorded in cubase 5.1.1, 1 bar of 4 bass midi notes and exported to audio and it plays back fine, zoomed in to audio wave form by the absolute max, cant zoom in any further and the notes are a tiny tiny amount offset but surely not by an amount thats audible.

This issue is purely academic, I cant see by the miniscule amount the render is out that it can make any real world difference

what do i know though ?!

[whyterabbyt]

The 'timing' of already-rendered audio is sample-accurate. Thats immutable. One audio sample per timing subdivision. It runs at the sample rate, and as such, visually, its timing can be displayed exactly if you soom in sufficiently.

The internal timing of MIDI events, however, has nothing to do with the sample rate. Ultimately, it has to be quantized to that sample rate for playback purposes, but it doesnt run nearly that fast.

MIDI events are driven by a clock which is measured in pulses per quarter note. In a modern DAW ppqn is typically set at values like 480, although that's something which can be altered. The further complication of that is that ppqn is derived from the tempo, not the sample rate.

And then you have quantization on, to 64ths, or 32nds, or 16ths.

Just imagine how these different layers of timings lay over each other as though they were on stacked grids. As you go from one grid to the next, imagine you have to shuffle your timings around so that events are grid-accurate.

Imagine you map events onto a grid thats got a resolution of, say, 480 pulses per quarter note at a tempo of, say, 120bpm but the events are being quantized to a further grid of, say, 32nds then that has to map on yet another grid of 44100 samples per second.

So your quantized note sits on a grid mapped to a grid mapped to a grid, and none of those grids exactly overlaps in integer divisions, and the highest possible accuracy an event can have is at least an order of magnitude less accurate than audio samples can be.

And you have to do some rounding at each new level of 'grid' to get things to map. always forward? always backwards? to the nearest value? the latter means some events get moved forward and some get moved backwards.

These non-integer relationships mean that your MIDI events are essentially at a much cruder resolution than your sample rate. They might look onscreen like they should map onto exact samples, but they dont. They dont even map onto a convenient multiple. And they've been rounded to the nearest ten (or whatever, could be twenty, could be fifty) samples.

All this means one simple thing; zooming in at a sample-based level to view the picture of an audio waveform of a synth, and comparing that to a picture of where the MIDI event looks to be is misleading because that MIDI event is NOT clocked on the same timescale as the audio.

The view of the MIDI event timing is an abstraction of data running at a subdivision of a clock which is unrelated to, and far slower than, the sample rate, and the mapping of that abstraction onto the underlying sample-accurate timing is done by approximation (ie rounding).

A picture of an event at an approximation of its timing against a 'hard' timeline like the sample position does not indicate that the sample data is innacurate against the MIDI. It indicates that the MIDI event is not accurately-enough timed to display it absolutely on that audio timeline.

So you're looking at something and seeing it the wrong way. Your MIDI isnt right while your audio is wrong. Your MIDI cant be timed accurately enough to be right in that view, only approximated. However the MIDI is being interpreted timing-wise on bouncing, its being done so accurate to the MIDI clock, which has a lower timing resolution than the sample rate. Its only as 'right' as the MIDI can be, which is less accurate than audio can be.

The fact that you've quantized your MIDI timing, thus potentially shifting your MIDI events forwards and backwards to a timing grid which may not be an integer division of your sample rate OR your ppqn only exacerbates that, IMO.

[bazwillrun]

So if I've understood that correctly, the OP is chasing his tail on this then !
and like i said but without knowing the exact reasons its all pretty academic and makes no difference to the resulting audio

[whyterabbyt]

So if I've understood that correctly, the OP is chasing his tail on this then !

Im inclined to believe so.

and like i said but without knowing the exact reasons its all pretty academic and makes no difference to the resulting audio

I dont recall him talking about anything other than how it looks onscreen, so yeah.

[musical android]

the danger of digital music-
to forget to listen with you ears.
the jazz musicians of yesterday would be laughing their asses off!!!

[bazwillrun]

From the 2nd para of the OP

" I was listening back to the project recently, and found that one of the tracks was glaringly out of time. When I zoomed in on the project, I was horrified to find that all of the waveforms that had been rendered from midi were all out of time, the first beat after the midi note, then the beats afterwards either in front of or after the midi note, drifting out of time inconsistently as the track moves forward."

According to that he says he heard it was out then zoomed in.

all i can say is he must have damn good hearing

[jancivil]

said I had crossed over to the land of absurdity. Is that how you normally respond to new members just looking for a bit of help? I'm not flustered, just surprised by the hostility I received when trying to solve a problem.

You described a practically impossible scenario, and it turns out you are focused on an impossible expectation, having described it in the farthest reach.

"the danger of digital music: to forget to listen with you ears.
the jazz musicians of yesterday would be laughing their asses off!!!". That nails my attitude.

I saw a bit of your video. You are dwelling in an absurd place with this. You spent nearly eight minutes describing in detail irrelevant information, before demonstrating it at all. I guess you think people have tons of time to waste on something that no one else considers an issue. It's extremely obsessive behavior. I don't need to insult you, this is not hostility, it's that I think you need a dose of reality from someone and I'll take the risk of someone reacting poorly to help you out in this way.

You are zooming in all the way or close to it, in most of what you're trying to show, and remarking 'way way off the beat'. Visually you are forcing it to appear as far off as you can, which is evident to all.

The midi 'sounds on time to me' but there is no mention of the audio 'sounding' out of time, you are focused on what seems to be like one sample difference occasionally in a depiction of the waveform. In an absolutely quantized to a low-resolution midi. It seems clear you're going to some lengths to prove your issue. You have demonstrated now that the problem is in your understanding and a particular, unrealistic expectation out of that.

What Whyte said needs to be heeded, particularly as hard-quantized as you're doing where the numbers will never be close to samples, it's a round peg in a square hole problem.

To be perfectly frank it looks like you've found the perfect way for you to avoid making music.

[paveloso]

Well, I watched your video (25 minutes of it anyway). I must say there is not a single problem with your rendered audio, man.

You zoom all the way in to prove your point that notes are WAY behind the beat, but you never play the rendered audio to actually hear them. If you did, you would realize that they sound perfectly in time!

The time difference is so incredibly negligible that I am astounded it has stopped you from making music for months! I mean, seriously, if you tell me you can actually hear it and not only see it in ultra-zoom-in then you must be some kind of cyborg with diamond-cut precision and bionic hearing.

You could buy another 10 computers, another dozen sequencers and a gazillion interfaces only to end up with the same results.

Why? Because, as you have already found yourself, there is no technology available that will give you the ridiculous precision you "need" for some reason beyond my understanding.

Are you making music? Or are you trying to achieve alien, nonhuman accuracy for some secret purpose?

I really don't mean to be rude, but I must agree that this looks a lot like obsessive-compulsive behavior.