Why AviSynth?

[retrogrouch]

I was having a discussion with a friend who just got a workprinter. I told him about the stuff I had seen on this sight about Avisynth, and How great the results looked. I must confess that I know about as much about computer programing and software as a chimp knows about the space shuttle. Our conversation boiled down to this (him- film professional) "I guess I'm not clear on the advantages of working in a text based system where you can't see the effects of the changes you're making like you would in any other NLE, apart from the fact that it's open source and free (and I'm pretty sure that there are free NLEs out there now too). What am I missing?" Like a half baked teenager, ( even after some googling) my answer sounds allot like " cause it's cool" Can some one tell me why this program seems popular, in the days of high power NLEs?

Thanks

RG

[grainy]

I come from the same monkey-based tech school, and am also curious about this...

[carllooper]

Being a programmer I can answer this one.

Once you understand the relationship between a piece of code and what it does to an image you no longer need to see the image on the screen. You can see it in your minds eye, from the code, what the result will be. A bit like the traitor in The Matrix. And it actually becomes much faster (and repeatable) to alter the image to what you want, by just changing some lines in the code, or some numbers, than to work through a graphic user interface. You can still see the results of the changes of course.

But far more importantly you can start writing your own programs. A graphic user interface program has a fixed set of pre-written programs and a limited ability to alter the behaviour of those programs (sliders etc). The program is locked into the assumptions of the programmer who wrote it. Now more often than not this is just fine. The programmer has already solved your problem and packaged it with a nice user interface that is easy to use. But it's when you have a new idea, that goes beyond the assumptions of a given hardwired program, that something like Avisynth comes into it's own. You can write up your own program.

Most high end NLEs assume your source material is already digital or already in a form compatible with the digital. AviSynth emerged in the transition from analog to digital, as a tool for processing analog video in a way that would then look good in the digital domain. Deinterlacing analog video for example. Or denoising analog video. But more than that - coming up with ways to do it better, without knowing in advance how that might be done. The ability to create your own NLE to suit your own particular problems. Now of course, a solution to your particular problem becomes a solution to others with the same problem. You can redistribute your solution to others through a common system: AviSynth. It's about keeping assumptions to a minimum and the potential at a maximum.

Avisynth then became something very useful for transitioning film into the digital domain. There is a unique set of problems to be solved with film. You can bypass the mainstream system with it's humungus costs and produce your own cheap-as-chips but beautiful solutions.

There is of course a learning curve. Not everyone's cup of tea. But it's definitely my cup of tea.

Carl

[retrogrouch]

Thanks for the quick and to the point Answer. I suspected that this tool is not a replacement for an NLE, but more like a filter in your work flow. Small gauge film making has it's own special problems that it doesn't share with digital, or pin registered big gauge formats. Because of this, and the fact that small gauge is a comparative tiny nich market ( maybe inhabited by quixotic masochists) the big software producers would not include tools for us. I guess the big question for us non computer jedi types, is how hard is it to take one of the scripts written by somebody else and use them?

Thanks for reading

RG

[carllooper]

I guess the big question for us non computer jedi types, is how hard is it to take one of the scripts written by somebody else and use them?

Yes, that's a more difficult question to answer of course. The thing is to give it to a go, without too many expectations. If you follow the getting started guides, without any particular problem in mind to solve, you'll be in a much better position to understand the system. The guides are very easy to follow. Starting is the most difficult because you'll typically go into a new domain with a particular problem in mind and can get fustrated very quickly - that you are nowhere near getting close to the solution you are after. The trick is not to have any problem in the first place.

May the force be with you.

ps. I don't use AviSynth anymore (left it behind quite a few years ago now) so I won't be much help with particular advice on AviSynth. I now write all my image processing programs directly in languages that faciliate native binary code: C/C++. But I recommend AviSynth for anyone starting out in image processing programming. And there are heaps of great programs written for AviSynth. It is very powerful. VideoFred, for example, is someone who has taken AviSynth into the stratosphere and beyond.

[retrogrouch]

Thanks Again Carl,

yup, I am sure that anyone who has read this far into this thread is a fan of Freds work. Thanks for the tip on going to the Avisynth sight. I've started "bootstrapping " my way through it, to get some answers. many of the concepts are familiar to me, and it seems some what demistitified ( of course I know how Houdini got out of the box. Could I do it?....)
One of the things that has amazed me about the avisynth clips I have seen has been the sharpness of the image before and after. I have read about the stabalization plug in in fred's script , DePan ( global motion compensation of neighbor frames for strong temporal denoising and clip (film) restoration.) and was wondering how well it measures up against the stabilizers in FCP. Is the software taking the images beyond the level of steadyness that can be achived with a when maintained camera and projector?

Here's a side by side I found on Youtube just for DePAn
Cheers,

AP

[Lunar07]

Excellent point!
I use POVRay tracer, for example, to create height fields from video files or overlay video frames on any geometric structure i can come up with. All in text!
I even wrote a GUI in Delphi that interfaces with AVISynth, that acts as a frame server by the way.
How about writing a C++ extension and compile it as a Python PYD to interface Python with AVISynth.
The sky is the limit.
And indeed, once you do it, you start seeing the effect in your mind. You morph it to your need. You AUTHOR the effect and the transition etc....
Kind of like manipulating the RGB values in a Triangular RAW format of images.
Definitely worth the effort to learn. NLE's are secondary on my PC. Their restrictions bug me.
But indeed one has to develop a taste for it.
By the way, text based processing is not restricted to Video processing. There is even a MIDI processing scripting language. Music scripting extensions to MOD and MIDI composing software.
There is a whole world of text based processing packages out there.

Being a programmer I can answer this one.
Once you understand the relationship between a piece of code and what it does to an image you no longer need to see the image on the screen. You can see it in your minds eye, from the code, what the result will be.[........]
Carl

[VideoFred]

One of the things that has amazed me about the avisynth clips I have seen has been the sharpness of the image before and after.

If you are referring to my scripts, it's a simple trick: 3x unsharp mask, radius 3,2,1 large radius first. You can do this in VirtualDub too, with the unsharpmask plugin filter.

I have read about the stabalization plug in in fred's script , DePan ( global motion compensation of neighbor frames for strong temporal denoising and clip (film) restoration.) and was wondering how well it measures up against the stabilizers in FCP. Is the software taking the images beyond the level of steadyness that can be achived with a when maintained camera and projector?

The trick here is to do clever motion estimation. Not on the entire frame, but on a certain region. This can change from scene to scene, but in general I use about 2/3. The benefit of Depan: it runs in one pass and in real time. And with Avisynth we can do before/after comparisons. And it shows some interesting info about how the stabilizer works.

And then there is the speed factor: VirtualDub starts at once, it does not even has to be installed. A simple stabilizing script takes about 1 sec to load in VDub.

many greetings,
Fred.