FS7: prores, xavx-i and photo Jpeg

[Doug Jensen]

Oh wow, i have been sick from work for a while, so only done some small edits here and there. Still on CC 18. Upgrading now!

Thats gonna be a game changer for sure, for stock. Im at 300mbits fiber so don't care about size.

Outputting stock footage clips from Premiere will be a workflow nightmare because every clip has to be on it's own timeline. On the other hand, in Resolve all the clips can be on a single timeline and you just tell Resolve to render the clips as individual clips. Takes 5 seconds vs. setting up a render and sequence for every individual clip. Maybe not a big deal if you are doing a handful of clips but I'm often doing hundreds at a time. In fact, I'm working on 300 clips right now that I'd hate to have to render individually. I'd quit before I had to do it that way. Graded individually, yes, but not rendered one at a time.

BTW, all of these workflow tips and workarounds are covered in my "How to Shoot Stock" training.

 

[Martin.G]

Can Davinci export in prores though?

I only use Prores for delivery to stockfootage sites.

National TV channel takes DNxHQ and thats perfect.

 

[Publimix]

Doug, I guess you are right... again. The one million euro question still remains; what footage will be bought.

 

[Doug Jensen]

Can Davinci export in prores though?

On a Mac you can export to every flavor of ProRes imaginable (except maybe RAW) from Resolve. If you're on a PC, I have no idea if it can do any ProRes at all . . . but you will have my sympathies.

 

[Martin.G]

On a Mac you can export to every flavor of ProRes imaginable (except maybe RAW) from Resolve. If you're on a PC, I have no idea if it can do any ProRes at all . . . but you will have my sympathies.

Sorry if im coming off salty, thats not the intention. Just a little on the defence regarding machine and OS choice.

Im saving a lot of money building my own PC and i dont have stock footage in my business plan as of today. But i live near amazing nature like the Lofoten Islands, Atlantic Road, and the fjords in Norway. So i see it could be a little income with the b-roll i do on my own projects. If i where to pay the high premium cost of apple machines, i would never get the investment back in prores exports alone...

Premiere Pro should export nice and quick if i just set In and Out on each clip individually. Should go fast enough with a proper preset set up. And i do so little stock footage i dont need batch process of hundreds of clips in one go. Not yet at least.

 

[Publimix]

Can Davinci export in prores though?

I only use Prores for delivery to stockfootage sites.

National TV channel takes DNxHQ and thats perfect.

ffmpeg and User Interface myffmpeg (20 euro) converts from Davinci on a windows machine to Prores. And pretty fast.

 

[Martin.G]

Is that addons for Davinci?

 

[Publimix]

No Sorry, It can convert most codecs to most other ones. I use Davinci and ffmpeg. So you need one step extra. I render in Davinci and convert to prores.

I used it only once until now. But it works fine. And the footage was accepted. The myffmpeg is an User Interface, ffmpeg is a bit nerdish software with command line strings. But the User Interface works really fine.

 

[jcs]

For (at least) 1/2 the file size at better quality, you can use a QuickTime container with Cineform wavelet. Try the various options (Filmscan 1, 2, etc.) to find quality/file-size that works for you. If you need something smaller, MXF OP1A 300Mbps H.264 is much smaller and very high quality for 4K upload. Otherwise, high bitrate IPB 8-bit 420 H.264 can provide sufficient quality and much smaller sizes (if you don't need HDR (10-bit)). If you have a fast machine/GPU, H.265 while it takes longer to export, will provide another ~2x reduction in file size over H.264 at the same quality. Again, experiment with encoding options and bitrates to find acceptable quality. With hardware H.265 encoding (Intel QuickSync or GPUs), output times can be decent (where otherwise you might be better off using H.264).

 

[Cary Knoop]

For (at least) 1/2 the file size at better quality, you can use a QuickTime container with Cineform wavelet.

You keep claiming that but you are wrong.
Wavelets may arguably be slightly more efficient than DCT but it certainly not at least twice as much.

 

[jcs]

You keep claiming that but you are wrong.
Wavelets may arguably be slightly more efficient than DCT but it certainly not at least twice as much.

This matches my experience doing export tests: https://blog.frame.io/2017/02/13/50-intermediate-codecs-compared/

Why is high bitrate H.264 (10 or more bits, 422-444, ALL-I) more efficient than ProRes? Specifically, what is different about the algorithms? What about ALL-I H.265 10-bit vs. ProRes?

How does DCT compression work vs. wavelet compression, mathematically & algorithmically?

Why does DCP use wavelet (JPEG2000) vs. DCT? https://comprimato.com/jpeg20001-better-compression-scalability-6-perks-jpeg2000/

Why did Apple ban/block Cineform (wavelet)? https://nofilmschool.com/2018/11/why-apple-dropping-cineform-and-dnx-final-cut

What evidence can you provide that shows that frame.io is incorrect?

 

[Cary Knoop]

What evidence can you provide that shows that frame.io is incorrect?

Where does it state in that article that Cineform compression is twice as efficient as ProRes?

With respect to the subjective difference in the comments:

Q: what is the difference between 3-MINOR vs 2-MINIMAL in lossyness? their is NO 1-NONE Lossyness in Cineform?

A: The differences are complicated and somewhat subjective. It also depends on the particular images that need to be compressed. I created those general categories as a rough benchmark, but there unfortunately aren't any agreed-upon benchmarks.

There exists several objective comparison methods (using for instance PSNR Cineform is only marginally more efficient than ProRes) and for a good subjective test only a double bind test would be useful.

And from the same frame.io website (take a look at the phrase bolded by me):

"The only significant difference between the two for a proxy workflow is the fact that you may have trouble creating ProRes on a PC, while DNxHD is very easy to create cross-platform. PCs can easily play back and edit ProRes files, but you can’t encode new ProRes files on a PC as easily as DNxHD, and so some editors prefer a DNxHD workflow for that reason.

A third good option is Cineform, which was acquired by GoPro but is still supported by all of the creative tools. Cineform offers similar options to DNxHD and ProRes, and it has shown slightly better image quality in some tests, but not by a huge amount."

https://workflow.frame.io/guide/proxy-codecs

 

[Cary Knoop]

Why is high bitrate H.264 (10 or more bits, 422-444, ALL-I) more efficient than ProRes? Specifically, what is different about the algorithms? What about ALL-I H.265 10-bit vs. ProRes?
[/url]

H.264 and ProRes are both based on the same DCT technology but H.264 has some additional optimizations (for instance CABAC) that increase efficiency somewhat but takes more decoding time. The difference between H.264 and H.265 is significant but decrease fast for all-intra and high bitrates. In fact, all codecs converge their efficiency towards higher bitrates. The often quoted 'H.265 is twice as efficient than H.264' only applies for low bitrate encodings for higher bitrate encodings the efficiency advantage is far less and for high bitrates (in such cases as for intermediate codecs) the advantage is small.

 

[jcs]

Where does it state in that article that Cineform compression is twice as efficient as ProRes?

With respect to the subjective difference in the comments:



There exists several objective comparison methods (using for instance PSNR Cineform is only marginally more efficient than ProRes) and for a good subjective test only a double bind test would be useful.

And from the same frame.io website (take a look at the phrase bolded by me):

"The only significant difference between the two for a proxy workflow is the fact that you may have trouble creating ProRes on a PC, while DNxHD is very easy to create cross-platform. PCs can easily play back and edit ProRes files, but you can’t encode new ProRes files on a PC as easily as DNxHD, and so some editors prefer a DNxHD workflow for that reason.

A third good option is Cineform, which was acquired by GoPro but is still supported by all of the creative tools. Cineform offers similar options to DNxHD and ProRes, and it has shown slightly better image quality in some tests, but not by a huge amount."

https://workflow.frame.io/guide/proxy-codecs

From: http://www.dvxuser.com/V6/showthrea...upport&p=1986787487&viewfull=1#post1986787487

From the chart: https://blog.frame.io/2017/02/13/50-intermediate-codecs-compared/

...............................................................Mbps
Cineform 10-bit UHD Film Scan...3840 x 2160 310 140 Variable 3 Minor 4:2:2 10-bit Yes
ProRes 422 HQ (UHD)................3840 x 2160 707 318 Variable 3 Minor 4:2:2 10-bit No*

707/310 = 2.28

Can you provide frame examples that prove the above is false?

Can you provide a link to PSNR or other algorithmic quality tests showing that ProRes/DCT vs. Wavelet does not perform as reported by frame.io (either Cineform or JPEG2000)?

 

[Cary Knoop]

F
...............................................................Mbps
Cineform 10-bit UHD Film Scan...3840 x 2160 310 140 Variable 3 Minor 4:2:2 10-bit Yes
ProRes 422 HQ (UHD)................3840 x 2160 707 318 Variable 3 Minor 4:2:2 10-bit No*

So because both are '3 Minor' you conclude Cineform is twice as efficient as ProRes? :shocked:
I rest my case.

And remember, the writer of the other frame.io article seems to agree with me, I quote:

"A third good option is Cineform, which was acquired by GoPro but is still supported by all of the creative tools. Cineform offers similar options to DNxHD and ProRes, and it has shown slightly better image quality in some tests, but not by a huge amount."

 

[jcs]

H.264 and ProRes are both based on the same DCT technology but H.264 has some additional optimizations (for instance CABAC) that increase efficiency somewhat but takes more decoding time. The difference between H.264 and H.265 is significant but decrease fast for all-intra and high bitrates. In fact, all codecs converge their efficiency towards higher bitrates. The often quoted 'H.265 is twice as efficient than H.264' only applies for low bitrate encodings for higher bitrate encodings the efficiency advantage is far less and for high bitrates (in such cases as for intermediate codecs) the advantage is small.

Comparing JPEG, ProRes/DNxXX (basically > 8-bit JPEG), H.264, and H.265, the main differences for ALL-I (intraframe) is complexity and optimization of DCT encoding macroblocks (IPB / interframe is a completely different case; still superior to wavelet and fractal methods. AI deconstruction will blow everything away (coming in the near future)).

JPEG has fixed 8x8 blocks up to sophisticated tree-based 64x64 blocks (CTU) for H.265. CABAC won't affect image quality (as does improvements with macroblock processing), only file size. Note as macroblock size processing variability increases, quality and efficiency increases. What is the (equivalent) size of the (effective) 'macroblock' for wavelet?

We don't really care about the high bitrates, we're looking to optimize for visually lossless at the lowest possible bitrate. For max efficiency, H.265 IPB/interframe currently is king. For visually lossless with least artifacts, wavelet provides a sweet spot option and that's why JPEG2000 is used for DCP and is becoming more popular. H.265 ALL-I intraframe is likely more efficient, but will have macroblock artifacts not present in wavelet. The other issue is H.265 ALL-I is expensive to decode vs. wavelet.

 

[jcs]

So because both are '3 Minor' you conclude Cineform is twice as efficient as ProRes? :shocked:
I rest my case.

And remember, the writer of the other frame.io article seems to agree with me, I quote:

"A third good option is Cineform, which was acquired by GoPro but is still supported by all of the creative tools. Cineform offers similar options to DNxHD and ProRes, and it has shown slightly better image quality in some tests, but not by a huge amount."

The frame.io quote you've repeated is for low bitrate proxy, and they're not comparing bitrate/efficiency.

Those chart elements are equivalent. Both ProRes and Cineform are ranked the same, along with the necessary bitrates to get those results. I couldn't see a difference when testing myself, have you done any tests yourself?

If we can't see a difference, where both are ranked as "3 Minor", isn't that a valid comparison, from which we can then look at bitrate to measure efficiency?

Where's the link to PSNR tests you mentioned, that will provide quantitative proof if matching quality: "3 Minor" is not sufficient?

What is your intention in this discussion? This might be helpful: http://www.dvxuser.com/V6/showthread.php?363972-How-to-win-every-debate

 

[JCummings]

Wow. It's a nerd fight.

 

[Doug Jensen]

Im saving a lot of money building my own PC and i dont have stock footage in my business plan as of today. . . . Premiere Pro should export nice and quick if i just set In and Out on each clip individually.

One clip at a time? There's no way I could do that for hundreds of clips. My god, the tedium and hours of wasted time would kill me. As for investing in the right hardware and software, I could post some stuff about "time being money", "stepping over dollars to pick up dimes", " you have to spend money to make money", and things like that, but only you know what your time is worth. Frankly, in the big scheme of things, my MacBook Pro cost me nothing when I amortize it's huge productivity advantages over the 3-4 years it will serve me before I decide to trade up to a faster and better model. And then I'll hand it down to a friend or family member who will get another few years out it. My current Mac cost me about $4500 last year and you couldn't have gotten me to take a Windows machine if you offered me a similar computer (on paper at least) for free. Why? Because I'd pay for it in lost productivity and loss of capabilities, such as not being able to spit out ProRes files right from Resolve with no extra dicking around. As you will learn, Premiere cannot do what Resolve can do and you will pay for that in time lost. Don't forget to factor that into consideration. Yeah, I'd rather spend $4500 on a Mac than take the best PC you could offer me for free.

Sorry if I'm going to come off as salty, too. That's not my intention either. Just a little on the defense regarding machine and OS choice! I'm just giving you the advice I would give to a best friend. Take it for whatever it is worth, but it is offered in friendship.

 

[Cary Knoop]

What is your intention in this discussion?

To let the audience know that: Cineform gives the same quality as ProRes with half the bitrate, which apparently you spread around on this forum as some kind of mantra, is absurd.