[Matplotlib-devel] switch default to h264

[Eric Firing]

It's all incredibly confusing, as the wikipedia article shows.  The 
first problem is the distinction between the container format and the 
codec; both have to be recognized and handled properly by an 
application.  The file extension generally is more indicative of the 
container than the codec.

MPEG-4 seems to be an umbrella that can cover a wide variety actual 
codecs, one of which is h264, and that at least implies a file format 
("version 2").

Specifying the mpeg4 codec in animation, with avconv or ffmpeg, seems to 
result in the use of a codec "mpeg4 (Simple Profile)"--but I suspect 
this might depend on how avconv or ffmpeg was built.  Specifying h264 
results in use of the newer and more efficient h264 encoder, which is 
what I am proposing to make the default.

One problem: h264 (at least as implemented in ffmpeg and avconv) can't 
handle frame dimensions that include an odd number of pixels, but mpeg4 
can.  Regardless of whether h264 is the default, we should trap this 
error at an early stage; otherwise it leads to a non-obvious failure mode.

I need to do a little more testing before submitting any PR.  I want to 
identify the conditions that were leading to video files that were 
completely corrupted.  When I was trying to get things working I was not 
keeping track of what I was doing, so I need to backtrack and make a set 
of examples.

Eric

On 2016/08/15 7:38 AM, Benjamin Root wrote:
> As I stated in my book:
>
> "While I will not claim to possess anything more than a passing
> familiarity with movie formats, I recognize that neither do many other
> people. For those who have had experience editing and creating movies
> and understand these concepts far more extensively, you can skip this
> section; be assured that I could not possibly add any new knowledge. For
> those reading on, there are likely to be some over-simplifications. As
> such, this section is only intended to provide enough information to
> successfully render animations and avoid pitfalls and should not be
> considered authoritative."
>
> My quick research at the time found that while mpeg4 and h.264 weren't
> strictly equivalent, for all intents and purposes they could be
> considered as such. And it is my understanding that avconv and such were
> treating it as the same (unless explicitly told not to). If I am wrong
> in this regard, please refer to the previous paragraph and kindly file
> an errata for my book.
>
> Cheers!
> Ben Root
>
>
> On Mon, Aug 15, 2016 at 1:23 PM, Chris Barker <chris.barker at noaa.gov
> <mailto:chris.barker at noaa.gov>> wrote:
>
>
>
>     On Mon, Aug 15, 2016 at 6:21 AM, Benjamin Root <ben.v.root at gmail.com
>     <mailto:ben.v.root at gmail.com>> wrote:
>
>         In my research for the animation chapter of my book, I also came
>         to the conclusion that the `mpeg4` codec in a .mp4 container was
>         the best bet for out-of-the-box compatibility across platforms.
>
>
>         Ben -- is that H.264? It seem "mpeg 4" is not well defined. from
>         Wikipedia:
>
>
>
>     "MPEG-4 is still an evolving standard and is divided into a number
>     of parts. Companies promoting MPEG-4 compatibility do not always
>     clearly state which "part" level compatibility they are referring to."
>
>     Sigh.
>
>
>
>          The one place where that format didn't work was MS PowerPoint
>         on Windows (but, oddly enough, it worked on PowerPoint on MacOSX).
>
>
>     That's because PPT punts to the "system" for decoding -- it has
>     little if any built-in support. A freakin' nightmare.
>
>
>         Note, I did that test a couple of years ago on a Windows 7
>         machine, so I don't know if that limitation is still valid.
>
>
>     I think H.264 is getting better and better supported.
>
>     Question: how is MPL planning to support it? Is there a
>     not-too-hard-to-install cross platform lib???
>
>     -CHB
>
>
>
>     --
>
>     Christopher Barker, Ph.D.
>     Oceanographer
>
>     Emergency Response Division
>     NOAA/NOS/OR&R            (206) 526-6959 <tel:%28206%29%20526-6959>
>     voice
>     7600 Sand Point Way NE   (206) 526-6329 <tel:%28206%29%20526-6329>   fax
>     Seattle, WA  98115       (206) 526-6317 <tel:%28206%29%20526-6317>
>     main reception
>
>     Chris.Barker at noaa.gov <mailto:Chris.Barker at noaa.gov>
>
>
>
>
> _______________________________________________
> Matplotlib-devel mailing list
> Matplotlib-devel at python.org
> https://mail.python.org/mailman/listinfo/matplotlib-devel
>