Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

CineForm Goes Open Source

cineform.blogspot.com

Post reply on HN
Print view View original

11–19 of 19 posts

Re: CineForm Goes Open Source

#11

Post by ttoinou »

dkh wrote: post #10
muizelaar wrote: post #7

Earlier quoted context omitted.

ffmpeg has had a CineForm decoder since Jan 2016: https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/cfhd...

Yes, but not an encoder, and that's the piece preventing it from being used to its potential. It's also nice when you no longer have to reverse-engineer the decoder. FFmpeg is one of my favorite things on earth, but minor differences between its implementations of closed-source codecs unfortunately prevent it from being a viable alternative in high-end post-production in many situations. ProRes encoding in FFmpeg for…

Can we find thoses quality control tests online ? I'm very interested in knowing what's wrong.

Re: CineForm Goes Open Source

#13

Post by m-p-3 »

ttoinou wrote: post #12
dkh wrote: post #9

This is very exciting news. I've been a huge proponent of CineForm in the production pipeline for several years now. Glad this finally happened.

Aren't the files too big ?

For distribution, yes. But for production and post-production that's fine.

Re: CineForm Goes Open Source

#14

Post by ttoinou »

m-p-3 wrote: post #13
ttoinou wrote: post #12

Earlier quoted context omitted.

Aren't the files too big ?

For distribution, yes. But for production and post-production that's fine.

I understand it's an intermediate codec. It seems that it's even heavier than ProRes XQ so I'm wondering but yeah, hard drive are cheap now

Re: CineForm Goes Open Source

#17

Post by 0dan0 »

ttoinou wrote: post #14
m-p-3 wrote: post #13

Earlier quoted context omitted.

For distribution, yes. But for production and post-production that's fine.

I understand it's an intermediate codec. It seems that it's even heavier than ProRes XQ so I'm wondering but yeah, hard drive are cheap now

Generally CineForm is lower bit-rate for the same quality as ProRES. See https://blog.frame.io/2017/02/13/50-intermediate-codecs-comp... for an independent analysis.

Re: CineForm Goes Open Source

#18

Post by 0dan0 »

sturmen wrote: post #4
j_s wrote: post #2

I am interested to see how the patent side of this shakes out, it seems to be a minefield for video codecs. Is there some scenario where going open source exposes the authors/companies involved to potential trouble? Releasing under Apache License Version 2.0 seems to cover the normal bases with its Grant of Patent License. What are the compelling reasons to choose to use the MIT alternative?

It would seem that CineForm is inherently simple, so perhaps it does not infringe on any of the "more clever" codec's inventions (AKA patents). I imagine GoPro ran it through legal before doing this.

Was the idea. CineForm was developed to avoid patents in 2001, by using technology that patents had already expired: 2-6 blockless wavelets, RLE and Huffman entropy encoding -- this is about that is used to make CineForm. Simplicity is what drove performance, wavelets over DCT is what helps in quality.The patent mind-field is in distribution codecs that strive for the lowest possible bit-rate, which was not a design goal for CineForm. Declaimer: I'm the author of the linked article.

Re: CineForm Goes Open Source

#19

Post by 0dan0 »

petedoyle wrote: post #8

Anyone have any thoughts on whether this could be implemented on the GPU? (e.g. CUDA) Especially encoding.

The 2-6 integer wavelets would be very fast on a GPU, although they are almost too simple. GPUs struggle more this the entropy encoding/decoding, which is what a prevented a GPU port in the past. Do any GPUs have generic Huffman cores now?
Post reply on HN