[theora] mixing effects when joining videos

[Ondrej Certik]

On Tue, Jun 16, 2009 at 5:05 PM, Gregory Maxwell<gmaxwell at gmail.com> wrote:
> On Tue, Jun 16, 2009 at 6:45 PM, Ondrej Certik<ondrej at certik.cz> wrote:
>> Hi,
>>
>> what is the best way to go about mixing effects when joining two
>> videos, like crossfading?
>>
>> Once I have the individual images as numpy arrays, the mixing itself
>> is the easy part (I'll just use numpy + scipy for that, or any other
>> python lib). However, it's not clear to me how (and especially when)
>> to handle decoding and encoding properly.
>>
>> So lets say I create a video tutorial (screencast) and I have 3 ogv
>> files. Now I want to join them --- so if I want some mixing effect,
>> one way is to decode them, mix them + join them and then encode it as
>> one video. I can do that already. But every decoding and encoding
>> makes the image a little worse (am I right?), so what is the usual
>> practise?
>
> The recommended technique is to initially encode to lossless (if
> possible; usually for shorter material) or near lossless (i.e iframe
> dirac at zillion mbit/sec) and only do low quality transcodes at the
> final rate when everything is done.
>
> Failing that you don't have many other options. You can try to do some
> fancy stream splicing so that the unmixxed parts take no loss.

Ah I see, thanks. Btw, something like flac for video would be handy
here. I am saving my screencast to png images and it's very big. For
this particular purpose, I think just taking the difference of
successive frames and encoding that as png would save lots of space,
I'll try it.

>
>> Another question is about frame rate --- (e.g. one frame rate for my
>> web camera stuff and another for the screencast) I read that theora
>> can join them,
>
> Chained streams do not work in many players, unfortunately.
>
>
> What you can do is make a stream that operates at the least common
> multiple of the two rates, then emits duplicate frames for the the
> slower parts which are then encoded and decoded efficiently.

Right, I see.

>
>> so that's fine, but if I want some mixing effects? And
>> if I want to upload to youtube --- they will convert it to some other
>> format, will it still work?
>
> Last I checked you couldn't upload Ogg/Theora to Youtube. I have no
> clue how they would handle variable frame rate streams.

Actually it works perfectly now. Here are two git video tutorials that
I made yesterday and I encoded to theora from png images and then
uploaded to youtube:

http://code.google.com/p/sympy/wiki/GitTutorials

Ondrej