Comments

jabloomf1230 wrote on 6/14/2009, 9:53 AM
The system always reports the codec that is presently installed to decode the video clip.
LarsHD wrote on 6/14/2009, 10:23 AM
OK, so it is reporting what version of *decoder* is present then.. ok.

Is it possible to see which version of CIneform *encoded* a specific Cineform AVI file?

Lars
321 wrote on 6/14/2009, 10:41 AM
Good question. I've been wondering if encoding has changed in the last several versions of Neo Scene.
LarsHD wrote on 6/14/2009, 10:53 AM
I've only experienced three different versions of their codec. All three have been different.

Last one plays best and has fixed the chroma sideways shift/skew. So while truly offering a big improvement I ind it unusuable because it produces a green color cast and the luminance levels are going down for each generation.

Lars

jabloomf1230 wrote on 6/14/2009, 3:07 PM
You'd better better off posting this kind of information on the Cineform message board. Most of the points that have raised, are only marginally related to using Vegas.
LarsHD wrote on 6/14/2009, 3:16 PM
I think you're right. I'm probably better off stopping posting about Cineform altogether I think.
jabloomf1230 wrote on 6/14/2009, 5:36 PM
That wasn't my point. I think if what you are discussing has some direct connection to Vegas, it makes sense to post it here. For example, your thread about XDCAM HD 422 is really helpful and it makes a suitable intermediate codec for people who are stuck with h.264 video clips, like for example, the high bit rate Canon 5d2 MOV files. This workflow makes Vegas Pro 9 a good candidate for those people with such hybrid DSLRs and a Windows PC. A lot of them are well-versed in photography, but may be somewhat intimidated by video editing.
David Newman wrote on 6/14/2009, 6:32 PM
There was a Vegas issue we have now addressed, that is what Lars was discussing. I re-uploaded Neo Scene 1.3.2.117 about an hour ago (you can use the same link.) It has codec version 5.1.0, it turns out the release didn't get the finished calibration data after all -- it now does. Wondered why Lars's tests didn't match my own findings.

David Newman
CTO, CineForm
LarsHD wrote on 6/15/2009, 5:30 AM
David...

So had I waited a few hours I would have gotten a 5.1.0 instaed of the 5.0.9 I got. You canged versions pretty fast there... ;)

So is 510 different from 509 in terms of color cast (towards green) and the drop in luminance level?

Is the 5.1.0 now a "final release" that will not be changed next hour again...?

Lars


JJKizak wrote on 6/15/2009, 5:56 AM
LARS:
Please don't get Dan upset. Thanks.
JJK
LarsHD wrote on 6/15/2009, 6:44 AM
I don't think David gets upset. Just look at the degree of constructive work David is getting done here lately. His product is improving all the time it seems like. With such a basically good / excellent codec his interest and my interest is the same - getting it to run as smoothly as possible with the highest possible quality... I'm sure he understands that.

He's getting test info from me, he's making new versions all the time.. he's getting new feedback, he makes another version... Can it be better and sweeter? ;)

Lars
David Newman wrote on 6/15/2009, 10:53 AM
Lars,

I do find your style a little abrasive, many of your responses are like you are talking to other companies who don't listen (let alone respond,) for CineForm it like you are shouting in our ears. I stopped visiting here so often as I get enough yelling once every couple of days or so (also this site don't give email notifications, so out of sight, out of mind.) So quiet down a little, if you find something you don't like, inform us politely, and we will see if we can fix it, but the sky is not falling.

The new version 5.1.0 is working nicely in Vegas, I will show you some multigen tests in a moment.

David Newman
CTO, CineForm
LarsHD wrote on 6/15/2009, 11:10 AM
David,

I am a customer of yours. I have customers too, but I never make personal remarks on forums about them.

Would it be possible to focus on the real technical aspects of this you think? I have donme my very best to provide you with as exact, technical and precise info as possible.

Please think for a moment about the time I have spent testing your codec and how I therefore perceive your resonse above.

====================

1. I *still* get green color cast. After multiple generations, I can clearly see a green cast. it is extrenmely easy to see and repro. Can you not see this? If you cant, I can email you examples of 1 - 5 - 10 -20 generations so you can see this. Please let me know. If you cannot see it with your eyes, the vecorscope very clearly shows how the colors are drifting away towards green.

2. After multiple generations I can see that the overall level / luminance level is going down. Can you see this too?

================

My only reason for spending the time with these tests is that I find your codec basically good. But not usable yet until colors are right. Colors and levels are not yet under full control. They must be for professional use in architecture, interior, fashion etc.

Are you willing to communicate only test results and technical aspects and be as specific as possible?


Best,
Lars



PS.

Thinking about it... If you are happy with the codec - as you obviously are - I think it will difficult to communicate this issue?

You are deviating from the technical issues into personal emotional areas instead of being "scientific". Which is what I would like a codec manufacturer to be... ;)

If you had the color / levels under full control you would of course not need to deviate from real technical issues into this emotional talk.

I get the message.... I'll just drop this.

I gave it my best shot. You're happy with codec as it is and I don't think any more testing and/or communication will improve anything.

Good luck David! You wont hear me mention the word Cineform again.



PS.
You are mentioning codec version 5.1.0 . The build number on your site and of the downloaded version is the same as I have, but there it says 5.0.9. All this adds to my confusion. Have you released a new version that has the same version numer, same build number but includes a new codec number? Should I download and install the same version # and build # again and a new codec version will pop up? Hmmm...

Laurence wrote on 6/15/2009, 11:51 AM
Some of us here have been with Cineform since it added to Vegas. At the beginning there was no other possible way to do HD other than with Cineform. Now there are other options, but for many things, using the Cineform codec still is the best.

David has been responsive to our needs here since the beginning, and has built up quite a bit of loyalty from many of us in the process. While I am grateful to Lars for pointing out the multi-generational errors, I would also like to note that Cineform has taken note and fixed or improved many little issues over the years when they were pointed out with private messages or with very low key posts in this (and the DVInfo.net) forums. An example of this would be how HD Link is able to capture and convert video that is in smart-render Vegas format. That was an idea that I suggested early on and was added without any disparaging of either Vegas or Cineform in the process.

As you have noted in your other posts Lars, without a codec like Cineform, one needs a really fast drive array to get good results over a number of generations. That and an obscenely large amount of disc space, even before considering backups. With Cineform I can get results extremely close to this with just a plain old eSata or USB2 hard drive off the shelf from my local electronics discounter. I work entirely from a laptop with external eSata and USB2 drives and for my workflow at least, Cineform is invaluable.

I hate to think that somebody casually browsing this forum, or somebody new to this forum, or somebody who has come here by way of a Google search, might get the idea by reading some of our recent posts that Cineform is inferior to XDCAM, or that Cineform is not a good practical substitute for working with uncompressed video. The fact is that Cineform is a wonderful codec that can be used pretty much any time you would normally expect to need uncompressed. The current temporary glitches are just that: temporary glitches in an exemplary product that many of us love. Keep in mind that us Vegas users are just a small subset of the Cineform user base, and that many of the issues that we are running into are unique to us since Vegas is the only editor still using VFW and and multiple color space conversions.

How is Cineform as a company? Let me explain it to you this way, I like my car, but Isuzu won't move the cupholder to a more convenient place if I call them up. I like my motorcycle, but Honda has no interest in my opinion about how high a windshield should be. Cineform is one of the best companies I've ever dealt with. Here I am, a some little mostly hobbyist level guy and yet they listen to me and actually act upon what I say! Even my wife and kids don't listen to me, but Cineform does! I love those guys!

What I'm saying is test away, point out things that need fixing, point out viable alternatives that might be well suited for certain people or occasions, but keep in mind that you are dealing with a company that actually listens to you and will go way beyond what anyone else will in support of their product. Be polite and give them a chance to fix any problems before write them off or suggest that anyone else do so. I have no doubt that in the end, you'll be a fan too.

LarsHD wrote on 6/15/2009, 12:09 PM
Laurence: "give them a chance to fix any problems before write them off "

----------------------------------

That's exactly what I have done. By providing way more info that is "normal".

I have been very appreciative in regards to the improvements that have been made. I have xplained how I have done my tests so that others can do them and so that there can be a reasonable level "scientific" approach to the whole thing.

But colors and levels *continue* to be wrong and that's not very good for the work I do.

There's nothing wrong with fast drives and plenty of disk space. And a codec-less workflow isn't bad either. And MXF isn't bad either.

You're working with a laptop. Maybe color accuracy isn't so important in your work, For me it is. So Cineform may be the perfect product for you. My work is different. I don't work on a lap and Cineform may not be what serves my needs.

I'm giving up :)



Take care. I love uncompressed AVI and fast hard drives! And this thread is getting to emotional now...

The rather severe chroma sideways shift of version 4.8.0. was obviously something most were happy with and loved... (?) they , CF, released it... and were probably happy.... but when I wrote about it, it actually got fixed...

I just don't like finding out real things and useful information and then beeing barked at!

I really regret I found about the various Cineform weaknesses now... I should have kept that info for myself and just smoothly transitioned back to uncompressed AVI... :( And everybody would have been happy with that severe choma shift (?).


Best & I will never ever respond to anything in this thread again... Back to the beach.

Lars
apit34356 wrote on 6/15/2009, 12:26 PM
Well Laurence, Ladrs comments about Cineform codec is not that out of line, especially compare to the beating SCS gets. Compare the expense of both products and that that the codec does just one function which it should do well. If Dan is unhappy then he has a serious problem, because markets in general are unstable and competitive. With invest capital or bank loans or credit lines hard to get or maintain, daily business operations are getting tougher, staffing issues become bigger. Dan in the past has been a positive asset to this forum as well as Sony's paying customers to Cineform, if Dan doesn't need this market, Sony Corp can always help little SCS with codec issues, especially since the new Sony Division align is suppose to make them stronger..... ;-)
apit34356 wrote on 6/15/2009, 12:34 PM
"really regret I found about the various Cineform weaknesses now... I should have kept that info for myself and just smoothly transitioned back to uncompressed AVI... :( And everybody would have been happy with that severe choma shift (?)."

LarsHD, your info has useful, so don't refrain from posting info. Your structure of posting data may be tough for some, but its just a style, so, everyone get over it.
The info is far more important that style, except in Apple's case and we are not Apple fanboys, right? Maybe right ;-)
Laurence wrote on 6/15/2009, 12:54 PM
Well Lars, I'd rather you posted in an abrasive style than not at all. Your threads have been invaluable to me. I really love the XDCAM codec and I had hardly touched it before your posts. I've been using Cineform lately mostly as a bridge between VirtualDub and Vegas. I expect however that VirtualDub has exactly the same issues as Vegas with multiple generations of Cineform since it also uses RGB color space to work and a VFW pipeline.

In the long run what I would really like to see is a "Lars approved" Cineform codec regardless of how how gentile the thread may or may not be before this is achieved.
David Newman wrote on 6/15/2009, 2:59 PM
Here are latest data from my tests and why I'm puzzled by claims of visible color shifts.

Here is the first frame of the original sequence.

For test multi-generation I used the convolution kernel set to unity (1.0 in the middle), that prevents Vegas from smart rendering. Smart rendering will give you loss between generations.

Same frame after 10 generation of Neo Scene progressive encoding at "High"

The quality stats:
>psnrbmp32 orig00.bmp NS10.bmp -stats
Luma Red Green Blue (dB PSNR)
46.62,44.60,45.89,44.88
Luma Red Green Blue (average LSB offset)
-0.31,0.21,-0.51,0.13

A luma PSNR of 46+dB is pretty good for codec going through color space conversion and 8-bit I/O. The LSB offset is who much the image has moved on average per channel after 10 generations. A value of 1.0 is a single step 256 possible values, so a value of -0.31 is not much.

Repeating the same test for XDCAM 50mb/s 4:2:2 HQ MXF

We see lots of macro blocks.
>psnrbmp32 orig00.bmp MXF10.bmp -stats
Luma Red Green Blue (dB PSNR)
42.46,38.90,42.60,39.38
Luma Red Green Blue (average LSB offset)
-0.69,-1.37,-0.42,-1.37

The lower PSNR numbers is likely due to the macro blocking which DCT codec can be prone. There is more color shift shown in the MXF exports.

For comparison here is Neo 4K, running 4:4:4 in Filmscan 2.
>psnrbmp32 orig00.bmp N4K10.bmp -stats
Luma Red Green Blue (dB PSNR)
52.19,49.19,51.05,49.24
Luma Red Green Blue (average LSB offset)
-0.00,-0.00,-0.00,-0.00

While 50dB+ is excellent in the 8-bit I/O is the limiting factor, the same sequence in a 10/16-bit app gets 60+dB luma PSNR.

While doing a multiple generation test in which nothing is changed in the image is not all that valuable, as smart rendering will prevent any losses for unchanged frames. The real test for any codec is to change the image in each generation (like you would in post), this would be color correction, titles, keys and layering etc, yet as we still need to visual compare each generation, I have used a spatial shift (2 pixels down and 2 to the right.) This is a really hard thing for any losy codec to handle, as everything moves, it would be the extreme/worst case scenario. In this situation, no codec gets to "settle". The shift is by 2, is so that 4:2:2 chroma artifacts do not interfere.

So you can compare here is the shift by 2 ten times through uncompressed.

After 10 generations with shift of Neo Scene progressive encoding at "High"
>psnrbmp32 origmoved.bmp NSmoved10.bmp -stats
Luma Red Green Blue (dB PSNR)
43.98,40.85,43.34,41.25
Luma Red Green Blue (average LSB offset)
-0.45,0.10,-0.65,0.02

Color levels are hanging in there but PSNR drops as the image got softer. 10 generation with spatial shift is not something you would want to do with Neo Scene (Neo HD and Neo 4K fair much better.)

Repeating the same test for XDCAM 50mb/s 4:2:2 HQ MXF
>psnrbmp32 origmoved.bmp MXFmoved10.bmp -stats
Luma Red Green Blue (dB PSNR)
38.14,31.85,40.55,33.27
Luma Red Green Blue (average LSB offset)
-2.27,-5.73,-1.00,-4.74

Some poor numbers here. Color is way off, and so is PSNR.

For comparison here is Neo 4K, running 4:4:4 in Filmscan 2.
>psnrbmp32 origmoved.bmp N4Kmoved10.bmp -stats
Luma Red Green Blue (dB PSNR)
49.37,47.10,48.36,47.15
Luma Red Green Blue (average LSB offset)
-0.01,-0.01,-0.01,-0.01

I think these result show that the codec is performing well and better than the codec we where asked to compare with. We continue to improve the quality whenever we can.

For those wanting to do their own testing, here is the PSNR tool used to generate the quality metrics.

Thanks for reading.

David Newman
CTO, CineForm