[theora] python bindings to libtheora

[Ondrej Certik]

On Sun, Jun 14, 2009 at 4:24 AM,
ogg.k.ogg.k at googlemail.com<ogg.k.ogg.k at googlemail.com> wrote:
>> Currently it can read any ogv file (that seems to work quite well) and
>> also encode to ogv (this mostly works, but there are some subtle
>> issues to be fixed, I suspect I still have some bug in the way I half
>> the dimensions of the Cb and Cr planes, so when you read ogv and write
>> ogv, the resulting image is not exactly pixelwise exact).
>
> If you are decoding to YUV, and reencoding, I think it's unlikely you'd get
> the same exact result.

Indeed, I fixed all the cropping/offset things and I do not get the
exact results, even if I encode it to quality=63. I don't see any
difference by my eye though, so it's good enough, but if I make the
difference of the arrays, the actual numbers (in the YCbCr planes)
differ roughly from -6 .. 6, mostly by +-1 or 0.

>
> Also, in the thread you linked, you used ogmmerge; while I have never
> used it, ogm files are said to be a kind of bastardized ogg, so I'd try
> oggz tools from liboggz (eg, oggz-merge), which are quite good for
> handling Ogg streams. Also oggz-validate can check your input streams
> for common problems, to check if the source program's buggy.

Thanks for the tip --- yes, oggz is indeed much better.

Ondrej