Hello All, I've now rer
 

Hello All,

I've now reread this thread, which is growing by the day!

You all want some solutions to problems here, so this post will propose some solutions. You can let me know here if they will significantly help.

First, a brief introduction: There's a two-way road between MIDI and notation. MIDI to notation is called "transcription". Notation to MIDI is called "musical interpretation". We should be clear about the distinction when we discuss and solve these problems. I will propose some solutions that deal with both transcription and interpretation in a uniform way that deals with both directions between MIDI and notation.

Ian reports that Composer’s transcription is inconsistent. Transcription is inconsistent if, given exactly the same MIDI file, the resulting notation is different. I highly suspect that Herbert experience of inconsistent transcription is due to different Transcription Options settings being applied at different times. With exactly same set of Transcription Options and the same MIDI file, the same notation should be produced. It's possible that some additional factor alters the resulting notation; but I have not ever observed exactly the same Transcription Options and same MIDI file produce different notation. I think the problem here is that the user is unknowingly applying different Transcription Options at different times.

My proposal for solving the problem identified in the previous paragraph is that the Transcription Options dialog will let the user set one or more “transcription profiles”. Take a look at Sherry’s screen shot of the Transcription Options dialog box near the top of this thread. There would be an extra place in the Transcription Options dialog box where the user could say, “Please remember exactly all of the checkmarks and settings in this dialog box, and give those settings the name ‘My Options.”

Now that you have a precise set of Transcription Options associated with the options profile name ‘My Options’, you can confidently recall those options when re-transcribing a MIDI file, with assurance that the MIDI file will transcribe to exactly the same notation.

Going the other direction, from notation to MIDI, is called “MIDI interpretation”. It has been pointed out in this thread that Composer does no MIDI interpretation, while other notation apps out there do. That’s quite true. That will change. In the last 18 months, I’ve already done a lot of unpublished work in the area of MIDI interpretation of notation. That includes interpretation of markings such as accent marks and dynamic mark. To be honest, I had not planned to also include swing rhythm interpretation, but that will be easy to add. These interpretation features will be delivered in Composer 3.0, which release date I cannot commit to yet.

Included with the interpretation feature will be the equivalent of the ‘My Options’ proposed above for the Transcription Options dialog box. The user will be able to control the musical interpretations with quite a few knobs. The user will be able to save distinct sets of musical interpretation knobs settings with one or more ‘My Interpretation’ names. Thus, there would be a close parallel in how the user can save and recall options for both directions between MIDI and notation: transcription and interpretation.

The above solution does not claim to solve every problem related to transcription that has been discussed in this thread. If we admit that perfection is impossible in transcription and interpretation, would you find the above solution a big step forward for Composer 3.0?

Cheers
-- Mark

P.S. Sherry sent me an email
 

P.S. Sherry sent me an email pointing out that it was Ian, not Herbert, who reported inconsistent transcription. So, I edited the above post accordingly. If you got an email notification of the post, the email will say Herbert instead of Ian.

Cheers
-- Mark

Hi Mark, You said: “I hi
 

Hi Mark,

You said:

“I highly suspect that Herbert experience of inconsistent transcription is due to different Transcription Options settings being applied at different times.”

No, I have not complaint about any inconsistent transcription, but pointed to a consistent reinterpretation of my notation while editing a .not file. Please read again my post of Monday, February 23, 2009 – 5:47 am.

Here is a step by step procedure to demonstrate what I mean.

1. Start a new song with a 5 line drum staff.
2. Place a 1/16 note for a tom at the end of the first measure.
3. Place a 1/16 note for a bass drum at the beginning of the second measure.
4. Convert the 5 line staff to single line staves
5. Convert the single line staves to a 5 line staff.

The result is that the tom has now become a grace note in the second bar.

Best wishes,

Herbert

Hi Mark Just read your last
 

Hi Mark

Just read your last post after sending my post.

Herbert

Dear Herbert I've just
 

Dear Herbert

I've just come to the forum after a busy day (UK time) and I'm reading the posts in chronological order. I've got as far as your February 25, 2009 - 3:25 am: post.

I'm am sorry you found anything in my post at which to take offence. I am totally at a loss at to what it might be, but I respect your knowledge and contributions, and I therefore express my regrets anyway.

Sincerely

Ian G.

Mark Continuing to work for
 

Mark

Continuing to work forward through the day's posts, as previously noted...........

Re your Wednesday, February 25, 2009 - 12:43 pm post:

I've already upset Herbert unintentionally; so I think I should point out that you are possibly (?probably) attributing to Herbert comments and faults that are mine, not his .

Ian G.

And Mark again Having now r
 

And Mark again

Having now read the whole of your your Wednesday, February 25, 2009 - 12:43 pm post, I have to say that all sounds brilliant. Can I just put in a plea that, where practicable, it should be possible to specify one's preferences by means of numerical values (e.g. perhaps the size of the 'envelope' within which Composer will deem midi sounds to be 'simultaneous' for the purposes of transcription)?

Power to your elbow (s) !

Ian G.

Hello, all I think I'm
 

Hello, all

I think I'm up to date now!

It seems to me that part - just a part - of the 'problem' we've been masticating so thoroughly is (paradoxically) that Composer is normally such an excellent transcriber. When a bar or two are not transcribed as one might like, one need is to 'see the midi', and what might help with that is a notator which is far cruder than we expect Composer to be. I've been looking around, and there's a little freeware sequencer called Anvil Studio, which I wouldn't want to have to look at for more than a few minutes at a time, but which I think might serve as a sort of microscope to 'check the midi'. What I envisage is to use Composer's excellent copying tools and the 'new from clipboard' route to make a 'mini' midi file to put 'under the microscope'.

That's the plan, anyway/ If it provers useful, I will report back.

BW to all

Ian G.

Hello Herbert,

<blockquote><hr size=0><!-quote-!><font size=1>quote:</font>

1. Start a new song with a 5 line drum staff.
2. Place a 1/16 note for a tom at the end of the first measure.
3. Place a 1/16 note for a bass drum at the beginning of the second measure.
4. Convert the 5 line staff to single line staves
5. Convert the single line staves to a 5 line staff.

The result is that the tom has now become a grace note in the second bar.<!-/quote-!><hr size=0></blockquote>That's a good trick you played on Composer to trip it up. Composer's counter move in this game is to go to the Setup / Transcription Options command, and turn off the checkmark for Transcribe Ornaments / Grace Notes. Now the 16th note is restored in the 5-line staff as a 16th note rather than grace.

Your move next ;-)

Cheers
-- Mark

Mark Thanks for your Februa
 

Mark

Thanks for your February 26, 2009 - 11:28 pm: suggestion. Yes, I can see the logic of that and it well be helpful. I'll bear it in mind next time I'm in 'that position', and let you know.

Thanks for all your attention to this issue.

Ian

Mark I'm going to attac
 

Mark

I'm going to attach a short *.not file. Let's ignore its history for the moment except to say it now LOOKS as a* .not as I would like to see it.

Sound-wise, it is a mixture of swung and straight time values. (That is because of the history we're ignoring). If I had input it all directly, it would sound ALL straight time.

What would you advise as the best next step to get it sounding all swung - i.e. as you or I would play it, within the swing-eighths convention ?

Sincerely

Ian <center><table border=1><tr><td>http://www.notation.com/discus/icons...hment_icon.gifst_looksgood
st+looksgood.not (26.5 k)</td></tr></table></center>

Hello Ian, I started to wri
 

Hello Ian,

I started to write up a tedious set of instructions for how to use piano roll to do this. You can take a pair of notated 8th notes, and hit the "33" key to change them to a notated and performed swing. That much is easy to do, except that you have to do it a pair of notes at a time, which is tedious. It gets even more tedious, however, to then keep the performed swing rhythm in tact while restoring the written-out swing to a simple pair of 8th notes again. This was so difficult to do, that I recommend that you don't even try it. If I can't figure out how to do it easily, there are only a few other people who have a chance of pulling it off (eg, Sherry and David).

So, I hate to say it, but there just isn't a good way to instruct Composer to interpret the 8th note pairs as a performed swing rhythm. That is a very common need for jazz. Other notation programs support this. So should Composer.

In the upcoming version 2.6, I'll add a swing interpretation feature that will make it easy to do this. When you see that feature in 2.6, you can say to yourself, "I got that feature added to Composer!"

It's task #2082.

Cheers
-- Mark

By the 'Crann Rule of Prop
 

By the 'Crann Rule of Proportions' (cf her post in this thread Friday, February 20, 2009 - 9:26 am ), a lot of people will be very happy with 2.6 ! Look forward to it.

Thanks

Ian

Hi, Mark: Is task #2083 
 

Hi, Mark:

Is task #2083 "get some sleep"?

David

Dear Mark, and all I shall
 

Dear Mark, and all

I shall attach 2 files, and *.mid and a *.not.

The latter is the (auto-) transcription of the former, with transcription settings 'remove overlaps of notes' and 'do not remove rests'. I haven't edited the notation at all.

I'm particularly interested in the 'Soloist' track. And so much of the transcription is very good, exemplary in places - see, for instance, bars 7, 12, 24, 25, 26, 27, 37, and 38 , for some clear notation of quite complex stuff.

BUT: look at bar 4 beat 3 (4-3);13-4;15-4;16-3;26-3;28-3&amp;4;29-2;32-1 and 33-3. I think they're all examples of the same thing - sequential notes transcribed as chords. In most cases, the time separations involved as quite as large as those correctly separated in the 'good' bars cited. Can you recommend a transcription setting that would 'catch' these ?

Also, look at the rhythm in bars 34 - 36. To my ear, the soloist is playing across the beat in pretty much equal steps - but that isn't what the transcription says. What are those triplets doing, for instance ?

Mark: I'm not getting at Composer. It may be something in the midi file. I'm just trying to understand !

Regards

Ian G.


<center><table border=1><tr><td>http://www.notation.com/discus/icons/mime_midi.gifBehind Blue Alice - mid
Behind_Blue_Alice.MID (16.0 k)</td></tr></table></center>
<center><table border=1><tr><td>http://www.notation.com/discus/icons...hment_icon.gifBehind Blue Alice not
Behind_Blue_Alice.not (156.6 k)</td></tr></table></center>

Hello Ian, I see that the r
 

Hello Ian,

I see that the relatively poor transcription is happening in the Soloist (BB) track. There's a good way to improve its transcription.

The problem is that the average note in the Solist track lags behind the beat by 29 MIDI ticks. I'll calculate below what that lag is in milliseconds. It turns out that this is such a big lag that it messes up the transcription.

We can fix the problem by shifting all of the Solist notes to the lest by 29 MIDI ticks, so that on the average they land exactly on the beat. Then we'll retranscribe the Sololist track with much better results.

Finally, if for some reason really want the notes to lag, on the average, by 29 MIDI ticks, as a last step you can add back the 29 MIDI ticks to each note.

The whole procedure is actually fairly easy. Here's how it works.

First, let's analyze the problem by looking at the piano roll notation.

1. Choose the option to see the as-performed piano roll rectangles.
2. Select the entire Soloist track.
3. Use the Convert Region to Selection of notes command.
4. Then shift the notes (both attack and release) by one tick to the left. The button in the piano roll pallette shifts notes by 15 ticks at a time. If you use the keyboard shortcut, Shift+LeftArrow, then the ticks are shifted one tick at a time. Use the keyboard shortcut in this example.

Here's what you'll see:

http://www.notation.com/discus/messages/32459/37129.jpg

I'm looking above at the rhythm transcription problem in measure 13. I'm stacking the deck in this write-up to show you a place where this procedure improves transcription. This procedure doesn't fix the problem in measure 4. This procedure doesn't promise to fix every problem. Rather, this procedure will improve the overall results of the transcription.

Notice that Composers reports that the note attack offset is +29 MIDI ticks. That means that the average note in the Soloist track lags 29 MIDI ticks after the beat. Visually, that means that the piano roll rectangles tend to lie to the right of the music notes in the score. You can see that above.

Aside for the mathematically inclined reader: How long is a MIDI tick? It depends on the tempo. In Notation Composer, there are 480 ticks per quarter note (which is common among many MIDI applications, but some apps using different resolutions.) The tempo of this song is 105 quarter notes per minute. You can see that next to the metronome icon in the toolbar. At that tempo, there are 105 x 480 = 50400 ticks per minute, or 840 ticks per second. One tick lasts 1/840 seconds = 0.0019 seconds, or 1.9 milliseconds. So, the average lag of 29 ticks is 28 * 1.9 = 34.5 milliseconds! That is a long lag!

I'm curious, Ian. Did you record in this Soloist track, or did BIAB generate it?

The next step is to fix the lag, temporarily or permanently, whichever way to desire. Click the Shift Note Left button 29 times, or type Shift+LeftArrow 29 times. Now you'll see Offset Attack: +0 reported in the status bar, which means that on the average the notes are lined up with the beat.

http://www.notation.com/discus/messages/32459/37130.jpg

Now, re-transcribe just the Sololist track. Use the Format / Re-transcribe command. Don't forget to choose the Swing style for this jazz-swing style piece:

http://www.notation.com/discus/messages/32459/37131.jpg

Here's what you see now:

http://www.notation.com/discus/messages/32459/37132.jpg

Turn off the piano roll, such as by clicking the Notes and Rest tab, to see the notes without the messy piano roll:

http://www.notation.com/discus/messages/32459/37133.jpg

If you want to restore the original performance, just reverse the earlier step, by selecting all of the notes in the Soloist track and shift their attacks to the right by 29 ticks.

Cheers
-- Mark

Dear Mark Thankyou for taki
 

Dear Mark

Thankyou for taking so much trouble. It's absolutely fascinating.

Yes, it is a Biab-generated 'solo', and I think your diagnosis is a remarkable vindication of Biab's musicality - what you are demonstrating empirically is the mechanics of a 'laid-back' feel. I've just paused in writing this to listen again to the midi file without any score or visual parallel, and I think it sounds very musical, in its type. Wouldn't you agree ?

As to the *.not file / transcription, well, refer to my earlier descriptions of the different ways I might use the file. I find the process you lay out perfectly acceptable if it's what the situation requires. I think in practical terms, I'd probably clone the stave, and have one for sound and one for score, muting or hiding according to my purposes.

It's also interesting that the process shifts all notes equally - so the 'distances between them' are unchanged, and yet the transcription changes. This seems to imply a note's relation to the (metronome) beat is more influential in the transcription than its relation to the preceding and following note(s). Is that right ? If so,(a) is the 'snap to grid' setting relevant, and (b) might a 'tweak' to make the 'distances' relatively more important be a future possibility ?

Intriguing stuff. Thanks again for your trouble.

Ian G.