Encore - Swing Playback - removal

[q]

"When all you have is a hammer, everything looks like a nail."

Part of the issue here is to imagine what Align Playback was originally for. I suspect it was originally for fixing up measures where editing had left the noes with anomalous start times.

Then a secondary use has presumably been, after import of a passage via MIDI, to make the timings match the inferred ("guessed" ) notation.

Doug and Eveline,

I believe Encore should interpret swing on-the-fly, and not change the original, underlying, raw data.

You're raising new and old questions:

What was Align Playback designed to do?

Did it ever successfully fulfill that goal?

Does it fulfill that goal now?

Is it attempting to serve two masters:


a) "undo" swing"

b) fixing anomalous input from real-time input or MIDI import?
Can it successfully do both, or do we need two separate tools?

Is any further function or flexibility warranted? I think it's clear, like a broken record, I keep voting "Yes, we need a choice with Align Spacing and Align Playback":


a) preserve durations — the default (and the only option at present) which moves the Note Off time value

b) preserve note off time values (presently unsupported) which alters the note's duration by adjusting the Note on time value only. For Swing Playback this is a more complex issue, but cause "the first note of pairs" must gain a longer duration, but perhaps the user might want the second note of a swung pair to incur a shortened duration by retaining the Note off time value.

To be able to undo Swing Playback, and getting back the original MIDI Data, two conditions must be met:

1. the Swing Playback function uses a reversible algorithmto calculate and alter the note start times and note stop times from the original note start and note stop times
2. we need a new function 'Undo Swing Playback' (or a name like that), that applies the inverse algorithm.

Indeed, when swing is unwanted, can we revert to prior state? Can we adopt adopt face values, or some version of face value? Should all three be viable options?

As I've said before, an underlying issue is that Align Spacing and Swing Playback alter the "raw data." We don't know for sure, but this probably means that Encore has no original data available to revert to.

Thus, the large remaining questions are:


Should Align Playback and Swing Playback alter "raw data"?

Should Encore interpret swing on the fly, leaving the underlying straight data unaffected, and available for reverting to? That would resolve all revert issues. (I've already stated my positive vote for this at the beginning of this post.)

q

[Doug Kerr]

Hi, q,

Doug and Eveline,

You're raising new and old questions:

What was Align Spacing designed to do?

There has been no discussion here about Align Spacing. Perhaps you have in mind Align Playback.

My guess: It is intended to align the start of notes with what the notation implies. Period.

If part of its duty was to "fix up" durations, how could it possibly know what would be "appropriate"? (Recall than in many cases these durations came from a live performance played in through MIDI.) Therefore I reject the notion that part of its original mission was to "fix up" durations.

Again, "align" is the key here.

Did it ever successfully fulfill that goal?

See next question.

Does it fulfill that goal now?

Predicated on my conjecture as to its mission: Yes, evidently. (Haven't tested every situation.)

Is it attempting to serve two masters:

a) "undo" swing"
Not as far as I'm concerned; the manual, however, says it is.



b) fixing anomalous input from real-time input or MIDI import?

[q]

Hi, q,

There has been no discussion here about Align Spacing. Perhaps you have in mind Align Playback.

Yes, of course. I meant Align Playback. Thanks for pointing out that error. I've updated the original post.

q

[q]

My guess: [Align Spacing] is intended to align the start of notes with what the notation implies. Period.

That's what I'd think — and it's the feature I keep requesting. Maybe all this quagmire goes away when we get a piano roll view in Encore and a real Quantize dialog. But a simple fix is warranted in the meantime.

Unfortunately, in it's current incarnation Align Playback preserves note duration, thus messing with the Note off time value ... and that's the only option. No way to "align the start of notes with what the notation implies. Period."

Encore has a Durations dialog for setting durations to one's intent ... but users have no way for "quantizing" attachs only (i.e. Note On and only Note On.) And I REALLY need that.

It common for MIDI quantization functions to allow the user to specify if quantization affects the time value of the attach alone, the position of entire note (i.e. Note On and Note Off) , or the time value of the Note Off only.

Our very own MasterTracks Pro offers the first two options (right in the upper left hand corner of it's Quantize dialog):



Maybe that's all Align Playback and Swing Playback need. A simple radio pair:

• Attacks only
• Entire note

q

[polarbreeze]

To be able to undo Swing Playback[/color], and getting back the original MIDI Data, to conditions must be met:

1. the Swing Playback function uses a reversible algorithm to calculate and alter the note start times and note stop times from the original note start and note stop times
2. we need a new function 'Undo Swing Playback' (or a name like that), that applies the inverse algorithm.

kind regards
Eveline

I REPEAT - IT USED TO WORK JUST FINE. I've used the swing function many, many times over several years and I've always been able to reverse the swing and bring it back to straight 8ths, by simply doing "align playback". PLEASE JUST PUT IT BACK THE WAY IT WAS!!!!

[q]

I REPEAT - IT USED TO WORK JUST FINE. I've used the swing function many, many times over several years and I've always been able to reverse the swing and bring it back to straight 8ths, by simply doing "align playback". PLEASE JUST PUT IT BACK THE WAY IT WAS!!!!

I think the illustrations clearly show that Encore does straighten swung attacks. Are you saying that it used to get the durations right too? If so, can you run Eveline and my tests through 4.5. Then attach the resulting MIDI file or post an image as we have.

q

[polarbreeze]

Part of the issue here is to imagine what Align Playback was originally for. I suspect it was originally for fixing up measures where editing had left the noes with anomalous start times.

I suppose that the purpose of "Align Playback" is (and always was) to make the note starting times and durations (modified by the "% duration" parameter) sound exactly "as written" because if the notes have been entered real-time their "raw" values will be something different from the "ideal" values implied by the notes as-written. If that is the case, then it's not surprising that "align playback" always did (until now!) obligingly undo the effects of the "swing" function.

There remains the mystery regarding at what point Encore "forgets" the "raw" values. Do they stay there for ever, thus making "revert to raw" an available escape route to put things back the way they were when entered? Or do some functions (for example "swing" or "align" or "change duration"...) overwrite those raw values for ever?

[polarbreeze]

Now, back to "unison" sequences and the infamous MIDI straddle:

As before, we see that Align Playback does what I think it was intended to, not useful in "reversing" the swing.

Doug

So you are saying that up until this latest revision, Encore was actually misbehaving all the time but by some lucky chance this misbehaviour had the beneficial effect of undoing a swing? I don't think that passes Occam's razor test - you're arbitrarily re-inventing what you "think it was intended to do". Surely the simpler and more likely explanation is that IT USED TO WORK FINE BUT NOW IT'S BROKEN.

[q]

Now, back to "unison" sequences and the infamous MIDI straddle:

As before, we see that Align Playback does what I think it was intended to, not useful in "reversing" the swing.

Doug

So you are saying that up until this latest revision, Encore was actually misbehaving all the time but by some lucky chance this misbehaviour had the beneficial effect of undoing a swing?

I haven't tested the result of reversing swing in 4.5, but I can tell you, it sounds plenty straight on the Mac. You don't really notice the problem unless you listen for the durations, which ARE messed up. I can't speak for Windows behavior, but the illustrations look similar.

q

[polarbreeze]

I think the illustrations clearly show that Encore does straighten swung attacks. Are you saying that it used to get the durations right too? If so, can you run Eveline and my tests through 4.5. Then attach the resulting MIDI file or post an image as we have.

q

I no longer have the old version installed so I can't test it any more. I don't know whether it somehow restored the "raw" values or whether it introduced "as written" values (see note below) but either way I am 100% certain that it corrected both the attack time and the duration because if it had not done so this would have become extremely obvious over the hundreds of time I've used the function.

Note: As mentioned in my other post, I'm betting that it actually introduces "as written" values, which is just fine as far as I'm concerned but if it is restoring the "raw" values, then that's fine with me too.

If this whole are of functionality is up for debate, don't forget to include the "revert to raw" function. What does that do, and when, and how does it collaborate with the other functions. And "guess durations"...? These things are all interrelated.

[q]

Part of the issue here is to imagine what Align Playback was originally for. I suspect it was originally for fixing up measures where editing had left the noes with anomalous start times.

I suppose that the purpose of "Align Playback" is (and always was) to make the note starting times and durations (modified by the "% duration" parameter) sound exactly "as written" because if the notes have been entered real-time their "raw" values will be something different from the "ideal" values implied by the notes as-written. If that is the case, then it's not surprising that "align playback" always did (until now!) obligingly undo the effects of the "swing" function.

Howard,

What you've described should be a feature!

Because it makes perfect sense ...

And that feature should modify (like what you stated) durations according to a "percent of face value"" parameter. Really, what else would the user expect? Initially, and for many years, I naturally assumed Align Playback would do what you describe, but in certain instances it clearly failed this definition on some account. Here's why:

By preserving durations Align Playback does not set the duration to face value, a percent of face value, nor a percent of the previousor existing duration. Align Playback adjusts the attach and note off by the same amount. That's all it does, and that's all its presently capable of. At least, that all that the user guide reveals (see below.)

My only other concern: If Align Playback gets a "percent of face value"" parameter (like the one in the Duration dialog) preferably it will be an independent value that does affect the Duration dialog setting; nor will the Duration dialog setting affect the value in the Align Playback dialog. All wishful thinking.

There remains the mystery regarding at what point Encore "forgets" the "raw" values. Do they stay there for ever, thus making "revert to raw" an available escape route to put things back the way they were when entered? Or do some functions (for example "swing" or "align" or "change duration"...) overwrite those raw values for ever?

We'll never know the underpinnings of all this until development reveals. In the meantime, I'm sure we're less helpful until Gvox shows a card.

Indeed. It's my understanding that swing and align overwrite raw data. (So among other things, the name is not apt.) but then, I guess I think of the raw data as the original data. Anyway, on page 146 of 2009 3.1 User Guide:

Align Playback changes the MIDI start times for notes within the selected area to exactly match the screen representation. Align Playback can be used to change a real-time or imported MIDI file to perform closer to the transcription obtained. Align Playback can also be used if extensive editing for a section requires re-establishing the correct note order and timing.

Align Playback does not affect durations.

The last sentence is where we find, in somewhat ambiguous terms that Align Spacing also adjusts the Note Off time value, by the exact amount that it adjusts the Note On time value — thus preserving the duration.

q

[Doug Kerr]

Now, back to "unison" sequences and the infamous MIDI straddle:

As before, we see that Align Playback does what I think it was intended to, not useful in "reversing" the swing.

Doug

So you are saying that up until this latest revision, Encore was actually misbehaving all the time but by some lucky chance this misbehaviour had the beneficial effect of undoing a swing?

I'm not saying that at all. I'm saying I don't think it is misbehaving now (and it wasn't in 4.5.5 either). And I have every reason to believe that is what it was originally intended to do. And I don't think that part of that was undoing swing.

I don't pretend to understand why it formerly did well for you in reversing swing. It could for example mean that the Mac version formerly worked differently. Or maybe it really wasn't "fully" reversing the swing, but your situations were such that the "partially reversed" situation sounded OK.

Here's why I think it its original purpose was just what the name says, to align the notes (that is, their starting times) with the notation: if it was also designed to do something else (let's say "undo swing"), I can't imagine any way that could have worked. How would it deduce what the pre-swinging durations of the notes were?

Why do I emphasize the starting times in aligning the notes with the notation? Well, for example, as we adjust the play durations of the notes, the one fixed thing is where the starting time falls in the measure. When we say that a certain eigth note falls on the third beat, we mean it starts there. And so forth.

But anything is possible.

All I know for sure is what I see.

In any case, if indeed the function is intended to do something different from what it is doing today (and was doing in E/W 4.5.5), but which it was in fact doing properly in an earlier Mac version you were using, Matt should be able to find that out by review of the development records (perhaps even by just reverse-engineering the source code in that Mac version) and will be able to put the algorithm in V5.x back to that.

Do you know the most likely version number of the Mac version that worked well for you in this regard? That might help Matt.

Best regards,

Doug

[q]

I think the illustrations clearly show that Encore does straighten swung attacks. Are you saying that it used to get the durations right too? If so, can you run Eveline and my tests through 4.5. Then attach the resulting MIDI file or post an image as we have.

q

I no longer have the old version installed so I can't test it any more. I don't know whether it somehow restored the "raw" values or whether it introduced "as written" values (see note below) but either way I am 100% certain that it corrected both the attack time and the duration because if it had not done so this would have become extremely obvious over the hundreds of time I've used the function.

I just tested in Encore 4.5.6.4 Mac. (I don't have any earlier versions installed.)

Here's playback in MasterTracks.

http://www.theoreticallycorrect.com/gvox/SwungUnSwungEncore45.mov

• The first measure is click-entered eighths, also entered into Measures 2 and 3.
• The second is swung/unswung eigths at 67%.
• The third measure is swung/unswung eigths at 75%

The artifacts introduced by swinging or swinging/unswinging sound very familiar to me. I hardly used 4.5 at all, so I think we're still hearing the effects of Encore 3 algorithms. Pretty flaking stuff.

q

[q]

It's common for MIDI quantization functions to allow the user to specify if quantization affects the time value of the attach alone, the position of entire note (i.e. Note On and Note Off) , or the time value of the Note Off only. Our very own MasterTracks Pro offers the first two options (right in the upper left hand corner of it's Quantize dialog):



Maybe that's all Align Playback and Swing Playback need.

A simple radio pair:

• Attacks only
• Entire note

For the record, I VERY MUCH get the results I'm looking for when quantizing attacks-only in various sequencers. This is true whether I'm applying swing, or fine tuning straight playback.

My main point here is that a slightly misplaced real-time attack is not necessarily followed by an equally misplaced real-time release That's the notion that Encore seems to believe! i.e. if the user misses the attack, they'll miss the release by the same amount. (I rarely find that to be the case, except on VERY short notes.

MasterTracks is smarter and more cooperative in this regard. And I assume it delivers a Gvox approved result. So please bring it on to Encore, in some simple fashion, and soon. If you want to get fancy about it later, that's fine.

As mentioned, for now, all we need is a radio pair in the align dialog, and adequate smarts beneath the hood:

• Attacks only (new feature)
• Entire note

Secondly, I rarely adjust the time values of note off after real-time recording, except when I hear an unpleasant and inappropriate overlap or gap. The way it is now, Encore creates and inserts unpleasant and inappropriate overlaps or gaps, and there's no in-Encore solution. And even if there were ... why make me go through that when align "attacks only" is the direct solution?

q

[Doug Kerr]

I had earlier reported (near the beginning of this Megillah) that, having applied Swing Playback to a passage, its prior state could not be restored by either of the techniques commended by the manual: application of Swing Playback with a swing parameter of 50% or application of Align Spacing. (This evaluation still seems to be true, as will shortly appear.)

Much of the further discussion (and testing) has centered on the behavior of the Align Playback function. I thought it might be useful to look a little more into the application of Swing Playback to a passage that had already been adjusted by Swing Playback. My first tests were (for obvious reasons) with the re-application of Swing Playback with a parameter of 50%. The sequence was eighth notes with duration 100% of face (a critical issue here), and the initial swing at 8th note, 75%.

It turns out that, in this case the result is exactly the same as if Align Playback had been applied. In neither case does this correspond to the original state of the note playback (note even close).

The whole story is shown here:

Swing_test_118-comp01.gif (21.67 KiB) Viewed 1926 times

I do not purport to be able to reconstruct this, which would require a generalized knowledge of the Swing Playback algorithm (which I hardly have).

I do not at all mean to suggest that the results of the two scenarios would be exactly equivalent for other situations. In fact, I know they will not be.

The more general question

The more general question is "what happens when Swing Playback is applied with one set of settings and then applied with another", allowing various values of the initial note durations.

This is more difficult to predict, since I am at this time only comfortable that I know the behavior of the algorithm for initial note durations of 100% of face, and only for orthodox original situations of the two notes of the pair.

Note that, among other things, for initial note durations less than 100% of face, even with orthodox starting sequences the result is "curious" (and presumably anomalous), but I haven't cracked the malgorithm yet.

Doug