[Stephen Dixon]

Nuh-uh. LZW is mathematically lossless, i.e. what you get out is bit-for-bit the same as what you put in.

LZW just removes redundancy in the file, without any processing of the image. That’s how it can be used for any file — LZW is the same compression used to compress a zip file. So it’s a lot faster than image-aware compression, but a lot less efficient. That’s why TIFFs are still way bigger than JPEGs, even with LZW on.

Stephen Dixon
Editor, Animator, Motionographer
Museum Victoria

[Stephen Dixon]

I think that’s exactly what the 3D extruder script does.

Stephen Dixon
Editor, Animator, Motionographer
Museum Victoria

[Stephen Dixon]

I’m pretty sure FCP does this, or at least it used to.

Stephen Dixon
Editor, Animator, Motionographer
Museum Victoria

[Stephen Dixon]

I’ve developed a technique where you can use text layers as expression editors. This has a couple of advantages: you can use one text layer to drive multiple layers and properties meaning that you only have to update the expression on all the layers. Also it updates as you type which is frankly awesome, and lastly you can see the expression in the comp window if you want, in whatever font you like with optical kerning and everything.

The biggest down-side of this technique is that you don’t have access to the pick-whip and the expressions drop down. If you’re a newb expressions writer then this is probably not for you, but for simple expressions that you want to apply to hundreds of layers it really is the bomb.

Basically it invoves using the javascript eval() function, that is sometimes used to drive expressions from external files. But there’s a trick to getting it to read text layers – you need to use the value property of a property. Full details on how to do it here: https://blob.pureandapplied.com.au/using-text-layers-as-expressions-in-after-effects/

exp = thisComp.layer("expressionSource").text.sourceText.value;
eval(exp)

Stephen Dixon
Editor, Animator, Motionographer
Museum Victoria

[Stephen Dixon]

It has always bugged me that there’s not a good 2D lookAt function in AE, so I made one:
https://blob.pureandapplied.com.au/lookat-me/

It can be used to create a better Auto-Orient function too.

Stephen Dixon
Editor, Animator, Motionographer
Museum Victoria

[Stephen Dixon]

If I was you I’d use ffprobe to find the dimensions and then use shell scripting – in the shell of yourchoice to make the decisions about output and filters before you get to ffmpeg. What you’re doing should probably work eventually, but it looks like you’re making it harder for yourself.

Stephen Dixon
Editor, Animator, Motionographer
Museum Victoria

[Stephen Dixon]

Stream numbering starts at 0. Check with ffprobe to make sure you’re using the right specifier for each stream.

Also, post the error.

Stephen Dixon
Editor, Animator, Motionographer
Museum Victoria

[Stephen Dixon]

I think it should be
ffmpeg -i input.mp4 -vf "fade=in:5:8" output.mp4

Many ffmpeg filters can be animated programatically over time, for example rotate can be applied thus: rotate=2*PI*t to rotate the image once every second (t is a constant representing the current time in seconds)

Stephen Dixon
Editor, Animator, Motionographer
Museum Victoria

[Stephen Dixon]

I think you may need to do it in two passes. First create a file with all the audio as separate streams, and then use that as the input to the next pass, where you can use the amerge filter to merge them into a single multi-channel stream.

https://ffmpeg.org/ffmpeg-filters.html#amerge and https://trac.ffmpeg.org/wiki/AudioChannelManipulation might be helpful.

Stephen Dixon
Editor, Animator, Motionographer
Museum Victoria

[Stephen Dixon]

Thanks for the tute. For a slightly more user friendly command line audio processor try sox. I usually extract audio, process in sox and remux.

Stephen Dixon
Editor, Animator, Motionographer
Museum Victoria