[LAU] AVSynthesis videos on-line

[Florin Andrei]

Dave Phillips wrote:
> Florin Andrei wrote:
>>> ... using Jean-Pierre Lemoine's AVSynthesis.
>>>     
>> Very nice. I tried this software a while ago but it was a pain to 
>> install. Hopefully that will change.
> 
> It's very specific about its requirements, e.g. Csound 5.08 compiled for 
> double precision, which means the user is probably going to have build 
> some stuff.

That's fine. I can put custom things in /opt if need be, I do that all 
the time with SVN snapshots and stuff like that. I'm not a coding expert 
but I'm comfortable with the GNU build tools.

Is there an AVS Install HOWTO for Linux?

> Thanks again, Florin. Have you considered writing a HOWTO or a dedicated 
> Web site for this information ? It's obvious you know it very well, and 
> it would be great to have a 1-stop web site for it.

This is standard knowledge in the field of video processing. Anyone who 
has spent some time with cameras and NLEs and stuff like that will tell 
you the same things.

What I could do is write a video HOWTO specifically for AVS, to allow 
people to create videos with this application at the best quality 
available with free software. You know, make a DVD or a BD that looks as 
good as possible, or a clip that will look good on Vimeo.
And then that could be included in a more general "AVS for Linux HOWTO", 
with installation procedures and stuff like that. But I can only take 
care of the video part.

But for that to happen I need to install AVS first and play with it 
myself. I tried that a while ago and gave up in despair. :-/

Whipping up a quick clip is easy. Making a video that looks good is not 
so easy. The situation is complicated by the fact that AVS was designed, 
from a video perspective, a bit simplistically: it pretty much assumes 
integer frame rates. That's fine for computer video. It's not fine for 
video made to be watched on a TV: NTSC video is not 30 fps, it's 29.97; 
"film" video (film-looking progressive digital material) is not 24 fps, 
it's 23.98.
Europeans (some of them at least) are in an easier situation - the PAL 
frame rate is not fractional, it's 25 fps exactly.

We discussed this before, I'm just re-hashing it for the benefit of 
everyone else - with AVS you can either choose to ignore the numbers 
after the decimal point and decide that NTSC is really 30 fps, and then 
you either have to de-tune the audio track or risk audio/video 
desynchronization. Or you can run the video track through a complex 
filter to adjust the frame rate and the results may sometimes be less 
than ideal if you don't use a very good filter.

> Btw, the Vimeo site converts my videos, but I don't know to what format. 
> Given what you've seen and understand of the site, what would be my 
> best-choice options for rendering the AVSynthesis images to video ?

There's no simple answer.
You have to come up with a procedure to generate the video file that 
will be uploaded. They have a few recommendations...

http://www.vimeo.com/help/compression

...which essentially boil down to "H.264 with AAC at 30 or 25 fps", but 
then turn around and say they convert everything to 24. I guess some 
experiments are required.
Eugenia Loli-Queru has several HOWTOs regarding the different formats 
recommended for Vimeo...

http://www.vimeo.com/forums/topic:3671

...but those are geared towards people using typical NLEs and stuff like 
that. Even so, it's a bit of a hit-and-miss - her Vegas HOWTO didn't 
work for me very well with material from an AVCHD camera, it created 
choppy video. I didn't try yet other methods.

http://www.vimeo.com/1224139
http://www.vimeo.com/1223670

See? Even when doing "classic" video, things can still go wrong. With 
computer generated video the complexity is even higher.

There's a bunch of interconnected decisions that need to be made, going 
back to the AVS parameters, and that's why I'm saying I'd like to try 
this software myself before blurting out random advice that may not be 
so helpful after all.

Tools such as mencoder create the illusion that video is easy. Yes, bad 
video is easy. Good video is not, you have to deal with fractional frame 
rates, interlacing, video encoders that create non-standard video (hello 
ffmpeg!), etc.

> Just to clarify: AVS does no rendering to video at all, it simply 
> creates the image sequence. I have control over the recorder frame rate 
> and the image width & height, that's all that's available to me in AVS 
> itself.

Yup, that should be enough. The collection of images just needs to be 
fed into the processing chain, and then all kinds of things can happen: 
cropping, resizing, interlacing, changing the frame rate, etc. Once 
you're in the uncompressed domain, almost anything is doable.

-- 
Florin Andrei

http://florin.myip.org/