cfenc - command line Cineform encoder

[marksfink]

Not sure what the interest level may be, but thought I would at least make it known that I wrote a command line Cineform encoder for Linux. Mac is also supported, but the main purpose and value is Linux, for which no native Cineform encoder has been available up to now.

You can find it here (with full details, please visit the site before asking questions). It's free/OSS of course.

https://github.com/marksfink/cfenc

If you have interest, feedback, questions, feel free to ask!

[LigH.de]

Interesting project. If it can be linked as encoder back into ffmpeg, then it can also be a free alternative for Windows users, as much as users of other OS.

[El Heggunte]

HolyWu has released a Windows build

https://forum.doom9.org/showthread.php?p=1906671#post1906671

[LigH.de]

Anyway ... using ffmpeg as I/O module in cfenc is one way; using cfenc as encoder in ffmpeg is a different one, and the more versatile.

[marksfink]

There was a ticket open to add Cineform encoding to FFmpeg. It was opened a while ago. It was closed on the basis that Cineform source is tied to x86 architecture and FFmpeg supports many architectures -- therefore they won't add it to FFmpeg. So rather than fight that, I just wrote my own program. I'm all for adding Cineform encoding to FFmpeg but I thought it would be an enormous amount of work to support architectures other than x86, so not practical.

[LigH.de]

Thanks for the explanation then...

[Cornucopia]

Not to tell you what to do, because I understand that these sorts of changes aren't trivial, but why not CHANGE the Cineform source so it isn't tied to x86?

Scott

[El Heggunte]

The truth is, many /most professional programmers in the OSS universe are lazy but do not have the courage to admit it...

And even when someone offers them some money, their laziness ++ their hypocrisy may speak louder ↓↓

https://sourceforge.net/p/gpac/discussion/287546/thread/43881fa9/

&&

https://forum.doom9.org/showthread.php?p=1878320#post1878320

Last but not least...

[El Heggunte]

There was a ticket open to add Cineform encoding to FFmpeg. It was opened a while ago. It was closed on the basis that Cineform source is tied to x86 architecture and FFmpeg supports many architectures -- therefore they won't add it to FFmpeg.

source→→ https://trac.ffmpeg.org/ticket/6863

[Cornucopia]

Yup.
My experience in IT tells me that response to the ticket is BS. Shame.

Thank you, marksink, for your efforts!

Scott

[richardpl]

Cineform code is BS. It can not be included in FFmpeg as is. Live with it.

[Cornucopia]

Who said anything about as-is?


Scott

[El Heggunte]

Cineform code is BS. It can not be included in FFmpeg as is.

Thanks for stating the obvious.

And for confirming that being a highly-skilled programmer /developer / software """engineer""" cannot stop anyone from being a sophist

[richardpl]

OT: We accept donations in time and money!

[El Heggunte]

I don't know what Paul B. Mahol wrote in the post above because his name is on my Ignore List

For now I say only this:

in 2017 he could have given me a straight-to-the-point answer, but he chose to write a trollish one instead.

[richardpl]

Welcome to my ignore list.

[LigH.de]

Could we please return to the topic? Or is this already "cabin fever"?

[marksfink]

I found it was very easy to add, so I created a branch called "alpha" on the Github repo for an update that supports alpha channels.

I will merge it into the master branch once I get confirmation from at least one person that it works correctly. I did some basic testing and believe it is good.

This is in response to an issue opened in the Github repo, so at least one person cares about it!

It works very simply: if the input video has an alpha channel then it automatically encodes to RGBA. No command-line option for it. I was surprised to find that the Cineform SDK does not support YUVA, which is why we encode to RGBA and don't bother with a command line option.

[marksfink]

Posted version 0.2:

Support "-i -" for piped input (previously required "pipe:")
Support BT.2020 coefficients for conversions
Add passthrough for v210 -- this was an oversight on my part, no harm caused previously, just more efficient now