Re:Video generation

[Flaming Hakama by Elaine]

From: Adam Spiers <address@hidden>
Subject: Re: video generation
Hi all,

...

>You could try that or use Knut's tool.  I'd probably use Knut's if
>I were starting today since it seems more recent / actively developed.

Knut's tool certainly looks worth investigating, and it's certainly
possible his approach is better than the one in ly2video in some
respects.  However please don't write off ly2video just because I
haven't done much with it recently (except for today).  It has some
nice features, like multiple scrolling modes, and the ability to
synchronise the video perfectly with live performances with variable
tempi.  In the long term it'd be great if the LilyPond community
converged on a single best-of-breed solution.

...

 

>>> To generate score videos like Bach: Invention 1, Wolf: Resignation and
>>> Peterson-Berger: I Fyrreskoven
>>> that give a different color to notes and rest while they are active you
>>> might use the attached files if
>>> you run a unixoid system like linux.
>>
>> This is a very interesting approach - thanks a lot for sharing!  I
>> particularly like the highlighting of notes currently being played.
>>
>> Two questions:
>>
>> 1. Please can you put the source files in a git repository and publish
>> it somewhere (e.g. GitLab.com), to reduce barrier to collaboration?
>> 2. Would it be possible to generate video which is perfectly
>> synchronised with a live performance with variable tempi?
>
> A patch for his changes in LilyPond is in the issue tracker at
> <https://sourceforge.net/p/testlilyissues/issues/5162/> but it seems
> basically abandoned after discussion.  You might want to look at the
> Rietveld issue to see what it would require taking it forward.

Thanks for the pointer.  That's just for the hooks though; there's
still the mkvideo script and videohelper.ily, so I think it would help
to place those in a git repo somewhere, together with a README which
explains how to use them (including links to the patches and code
review).

 I wanted to follow up and say that I investigated Knut's video code a bit.

The main observation I have is that this approach is bound up with audio generation.  I could be interested in helping pare it down to allow creation of the videos without audio, or by specifying an existing audio file.

Likewise, the overall interface and project notes setup could use some attention to decouple some of the logics.  And I'd be more in favor of implementing the business logic in a language other than bash (although Knut's bash-fu is quite impressive!)

However, I also realized that I am not as interested in the paginated approch.  So, for the moment I am investigating other approaches for obtaining a scrolling score.  My first attempt is to use Premiere Pro to do the animation.

I was hoping to use lilypond to export a single-page SVG, which I could, but it was not recognized as the proper file format by Premiere.  There were two minor errors when I ran the lilypond-generated svg through an online validator, but it did validate.  I'll have to look more into that, since svg would be ideal.

Then I tried a one-page png, which also exported fine, but was too large for Premiere to import.  So, I ended up formatting the score onto four pages and importing the four pngs into Premiere. 

There are clearly issues with this scrolling approch (static clefs and key signatures would be necessary) and the png quality is pretty crappy.  Likewise, the lack of indicator can make comprehension difficult.

But it was certainly a fun effort.

https://www.facebook.com/david.elaine.alt/posts/10156701810739782

In any case, I agree with some of the discussion in the ticket https://sourceforge.net/p/testlilyissues/issues/5162/

The biggest disagreement was about the color issue, but it seems like there was agreement that "it would be possible to allow color to be either a list of three numbers in the range [0..1] or a list starting with some symbol followed by other values."

There seemed to be agreement about the possibility to pass some code directly to the postscript prolog simply by string-appending it to ps-user-prolog.  This feature might also be used for other purposes, e.g. watermarking and page decoration.

There seemed to be agreement that a hook mechanism discriminating between backends will allow people to implement something other than a PostScript backend, for example try doing animated SVG.  These would be happy to have the possibility to use overloaded-rgb, and probably need an extended svg dump-page. So they probably would ask for a predicate use-svg-dump-page-replacement? in videoextensions.scm and a test for that predicate in the dump-page procedure in framework-svg.scm.

Thanks, 

David Elaine Alt

415 . 341 .4954                                           "Confusion is highly underrated"

address@hidden

self-immolation.info

skype: flaming_hakama

Producer ~ Composer ~ Instrumentalist

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-