Quicktime Only 8 bit in CC? Is PPro is unsuable for professional finishing and online?

[Marcus Smith]

[Walter Soyka] “I did test your file, and I see the same thing as you do — but as above, I think that your test pattern will band with ProRes 422 with any system”

Thanks for checking. On our systems this test dpx does band a little on FCP, Davinci and Smoke, but in no way near as badly as it does on Premiere Pro. FCP Davinci and Smoke preview a dissolve on this dpx in an almost identical way, Premiere Pro on the other hand shows far more banding.

[Walter Soyka] “I tried it on my Smoke system, too, and got the same banded result.”

Were you checking on it inside a Smoke timeline, on your monitor or from a pro res HQ rendered file, outputted by Smoke? I’ve had trouble with Smoke’s output to Prores HQ file via the export function which I haven’t had time to work out yet, but Smoke views the gradient with almost unnoticable artifacts compared with PPro when previewing to the monitor and outputting to tape.

[Walter Soyka] “If you try non-synthetic footage, or if you change the design of your test pattern to eliminate the other variables, you’ll see Premiere outputs clean 10-bit ProRes.”

One of the main reasons I started this post is that since we’ve moved from FCP to PP for online, we are seeing banding on natural shots, skies, water and other gradients in Premiere Pro Prores 422 HQ sequences. You mainly see this type banding when a gradient is accompanied by a dissolve. These same aberrations are not present in Davinci, Smoke or FCP when dissolving on the same shot. Our online editors are having to dissolve noise into the shots to fix the problem!

It seems to me that from your tests the output module in Premiere Pro is writing a 10bit file fine, its just appears as though something nasty might be happening at the RGB to YUV conversion in PPro before it gets to the file output stage in the chain. This would explain why going to DNXHD 444 and and DPX works fine.

Walter I feel you’re still not convinced there’s a problem, so the best way for me to show you is with vectorsope screen shots.

File attached here:

https://www.hightail.com/download/OGhlRm8zTkFtUUc1eDhUQw

Included are two screen shots, both of two different pro res 422 HQ files. One is the pro res gradient rendered in PPro and the other the same file rendered in FCP. I’ve just done it quickly with the low res Davinci scopes but here you can clearly see the FCP gradient is alot smoother, colours are evenly distributed. In the PPro screen shot, you can see colours clumped together and a noticable degradation in signal quality.

Now the interesting thing is if you are in a PPro unrenderd timeline viewing the file, on the vectorscope the image will look perfect, like the FCP one. As soon as you render a preview and PPro reads off the Quicktime the degradation appears on the vectorscope.

Thanks again for you time!

Marcus.

[Walter Soyka]

[Marcus Smith] “One of the main reasons I started this post is that since we’ve moved from FCP to PP for online, we are seeing banding on natural shots, skies, water and other gradients in Premiere Pro Prores 422 HQ sequences. You mainly see this type banding when a gradient is accompanied by a dissolve. These same aberrations are not present in Davinci, Smoke or FCP when dissolving on the same shot. Our online editors are having to dissolve noise into the shots to fix the problem!”

Marcus, I will try to check this out sometime tomorrow and get back to you — let’s see if we can figure out what’s going wrong.

Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events

[Walter Soyka]

[Marcus Smith] “It seems to me that from your tests the output module in Premiere Pro is writing a 10bit file fine, its just appears as though something nasty might be happening at the RGB to YUV conversion in PPro before it gets to the file output stage in the chain. This would explain why going to DNXHD 444 and and DPX works fine.”

Ok, I do see what you’re talking about — something there is definitely not right, and it might be QuickTime-related after all. I am getting clean previews with DNxHD 220X, which should be 10-bit Y’CbCr, after setting the sequence to render previews at maximum depth.

Does DNxHD 220X render any faster for your setup?

Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events

[Marcus Smith]

DNXHD renders extremely slowly on 3 systems I’ve test it on. The 440 format renders 4 times as slow as 422 HQ Quicktime.

PPro also can’t tell which flavours of DHXHD are very well and footage exported from other programs, say Davinci, require rendering, despite being of the same type of footage. Also I’m not convinced it’s a great intermediate format on OSX as I’ve seen gamma shifts come out of it in PP export. This may change in the future.

In regards the to PPro degrading the colour space it is definitely a QT related bug, as I’ve tested all the different flavours of QT and get the same results. I’ve onlined features in CS6 and used the Blackmagic RGB 10 bit format, which worked great. Now even that format bands. Don’t have CS6 currently installed to test it now.

What video output card you using?

Thanks,

Marcus.

[Walter Biscardi]

[Marcus Smith] “Is anyone using PPro for professional online and finishing or only for offline?”

I love assumptions like this. Well Quicktime might only be 8bit in CC so immediately it’s unprofessional for delivery.

I’ll say we’ve been delivering to PBS, NBC News, ABC News, SkyNews and more for the past two years with projects that are mastered directly out of PPro. So I’m going to say Yes, It’s Professional.

Walter Biscardi, Jr.
Editor, Colorist, Director, Writer, Consultant, Author, Chef.
HD Post and Production
Biscardi Creative Media

Southeast Creative Summit, Oct 25-27. Register Now! Save with code: creativecow2013
Foul Water Fiery Serpent, an original documentary featuring Sigourney Weaver. US & European distribution by American Public Television
MTWD Entertainment – Developing original content for all media.
“This American Land” – our new PBS Series.
“Science Nation” – Three years and counting of Science for the People.

Blog Twitter Facebook

[Marcus Smith]

[walter biscardi] “I love assumptions like this. Well Quicktime might only be 8bit in CC so immediately it’s unprofessional for delivery.”

Walter, did you not see the question mark in that sentence? I don’t see how an honest question about the whether the use of PPro is a valid online tool is an assumption, especially given the gravity of the mentioned problem in an OSX platform.

I also fail to see how your statement contributes to the issue at hand other than talking about the shows you have worked on.

Thanks,

Marcus.

[Marcus Smith]

[Walter Soyka] “Does DNxHD 220X render any faster for your setup?”

On your suggestion I went back again to test the 220x codec with some surprising results. I tested one minute of Prores HQ and DNXHD220x footage rendered in 220x, 440x and Pro Res HQ timelines with and without GPU.

The findings were that 220x actually renders faster than HQ, which I was rather happy to find out! The other interesting thing was that the DNXHD timeline renders the HQ faster than its own codec. Still though, the 220x timeline can’t recognise its own format, ie if you render out 220x and bring it back into PPro, it still requires rendering.

So for now 220x is the best options for online. One hour of Prores HQ should take about 20 – 30 minutes to render, which isn’t the end of the world. I’ve published my rendering results here: https://www.hightail.com/download/OGhmeEVRUzhWRCtKUmNUQw

What is the best way to inform Adobe of these bugs?

Thanks again for your help,

Marcus.

[Walter Soyka]

I’m glad we found a workaround that suits you.

[Marcus Smith] “What is the best way to inform Adobe of these bugs?”

File a bug report [link]. It looks like a pretty faceless form, but it actually goes directly into a quality engineer’s inbox. I’d suggest you link to this thread for additional context in your BR.

Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events

[Tapio Haaja]

I was testing this same issue last spring and my assumption is that Premiere shows Quicktime ProRes files on monitor or through video I/O (we hava AJA IO XT) as 8-bit but actually processes files as 10-bit. So there’s no way to monitor actual 10-bit output of Quicktime ProRes files even through video I/O. If I remember correctly there’s even slight quality difference of video I/O when video is paused and playing. DNxHD wasn’t showing any of these problems.

Best
Tapio Haaja

Development & Production Manager / Promotions / MTV MEDIA (Finland)

[Chris Gunningham]

Hi,

I realise this is a very old post but I’m seeing banding rendering to pro res 422 hq in 2015.2. Much less in 4444.Does anyone have any updates on workarounds or is dnx the way to go.

Cheers Chris