Return-Path: Received: from mail-pb0-f41.google.com ([209.85.160.41] verified) by media-motion.tv (CommuniGate Pro SMTP 4.2.10) with ESMTP-TLS id 4873525 for AE-List@media-motion.tv; Tue, 16 Oct 2012 02:53:28 +0200 Received: by mail-pb0-f41.google.com with SMTP id rq2so5030392pbb.28 for ; Mon, 15 Oct 2012 17:55:46 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20120113; h=from:mime-version:content-type:subject:date:in-reply-to:to :references:message-id:x-mailer:x-gm-message-state; bh=3PMFCu3+Md8U9/nfbiixaHujcLoOSll89g18c2i7eb8=; b=pgWfVSdj9ul2Tu5HIxjl34HY33YQcBonP+gwbceFUBVOGOOkctVN/OZ9XzAw0w00+d jFwjuTUdo8voegOyLLkmbixRUINL9xzG/BZ8Wqziq/UKvj+xKeAQqy75dYu+j8A5w9g0 7iRbogpP3yEdqkTMQH/FJQzh2BCgX7z4/mG4U2CQO8lWduCH7DpKma7OrWLMkuZH752r l97mDCRuhThC5Emd2+s2y70y5RPR+lGxIqxNOw9+FbspJT0oekbaye892yFSZDeYXVgR FrnV+rvHSxvAGPMrse5EyFL67HY5wHz5tFt8GL3M6T4UrcDHsfpymVvHl69QpnlnlIEr P8TA== Received: by 10.68.234.71 with SMTP id uc7mr41831832pbc.72.1350348946547; Mon, 15 Oct 2012 17:55:46 -0700 (PDT) Return-Path: Received: from [10.0.0.137] ([65.223.58.66]) by mx.google.com with ESMTPS id m5sm9879272pax.10.2012.10.15.17.55.34 (version=TLSv1/SSLv3 cipher=OTHER); Mon, 15 Oct 2012 17:55:42 -0700 (PDT) From: Phil Spitler Mime-Version: 1.0 (Apple Message framework v1283) Content-Type: multipart/alternative; boundary="Apple-Mail=_A7E49E91-45A8-4296-B3A7-9AFEC003CA60" Subject: Re: [AE] RED in CS6 Date: Mon, 15 Oct 2012 17:55:54 -0700 In-Reply-To: To: "After Effects Mail List" References: Message-Id: <3C527EB6-D6B2-4DA9-9036-B655EA7DE4B4@bonfirelabs.com> X-Mailer: Apple Mail (2.1283) X-Gm-Message-State: ALoCoQlZuHlK21MLQ93Lsy3Uiw5Yw1mLeJJIYKocWvb7X/RLvmWfiE2511UhImAPNGkYwHbT3+ab --Apple-Mail=_A7E49E91-45A8-4296-B3A7-9AFEC003CA60 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=iso-8859-1 Scratch that.... the test without scaling is also taking FOREVER. Phil Phil Spitler | Creative Technologist | Bonfire Labs | t : = 415.394.8200 m : 415.571.3139 Website | Facebook | LinkedIn On Oct 15, 2012, at 5:51 PM, Phil Spitler wrote: > Thanks Brian. >=20 > We are delivering 1920x1080 but are doing some serious pushes on the = footage. >=20 > It seems like the scaling could be a big issue as I just rendered a 5k = clip and it is 3 times the speed as rendering the same clip at 1080 = (after scaling the footage). >=20 > I will look at the debayeing settings too to see what I can figure = out, >=20 > The reason I wanted to keep the files R3d was so that I could do the = color and keying with maximum information in the file. >=20 > More testing to come. >=20 > Thanks. >=20 > Phil >=20 >=20 >=20 >=20 >=20 > Phil Spitler | Creative Technologist | Bonfire Labs | t : = 415.394.8200 m : 415.571.3139 >=20 > Website | Facebook | LinkedIn >=20 > On Oct 15, 2012, at 5:45 PM, Brian Higgins wrote: >=20 >> Phil, what resolution is your deliverable? Jack is right on the = money. Just because you *can* debayer on the fly, doesn't mean you = should. Heck...if you're not doing push-ins and repos, you probably = don't even need to do a full debayer! If you have fast and large enough = storage, do the debayering when you do the conform and write out = DPX/TIFF/EXR sequences of what you need. You'll be much happier! >>=20 >> Here's a test for you to try if your deliverables are (only) HD like = mine are. It's somewhat footage dependent, but might be enlightening: = Load the same r3d file twice. Debayer the first copy at full 5K, then = debayer the second at half premium. Drop them both into 1920x1080 = comps, and scale them so they fill the HD frame. Now toggle between = them and see how much difference you can spot. I can sometimes see a = tiny, tiny sharpness improvement on the full-debayered version, but a = lot of the time I can't. Now put a little sharpening (should be = standard operating procedure on Red and Alexa footage) on both comps and = see if you can tell the difference. If you've got 200% blowups to deal = with or are mastering at greater than HD resolution you obviously need = the full debayer, but if you aren't, you might not. >>=20 >> As to your render times...have you tried rendering a chunk out of = RedCine? A full debayer should only take a few seconds per frame there, = even without a RedRocket. >>=20 >> $.02, >> Brian >>=20 >>=20 >> On Mon, Oct 15, 2012 at 7:16 PM, Jack Tunnicliffe = wrote: >> Depending on what you are doing with the Red files it's sometimes = better to render it out as an image sequence, say 16 bit tiffs or = something similar so you computer isn't doing the debayer for every = frame while you are working with it. I find the only reasonable = explanation for sticking with the R3D data is so you can get at the raw = data at the initial stages should there be under exposure, over = exposure, etc. You can find detail that would otherwise be lost, but = really after that, why not convert it to a more AE friendly format.=20 >>=20 >> Jack Tunnicliffe >> Java Post Production >> www.javapost.ca >>=20 >>=20 >>=20 >> On Oct 15, 2012, at 6:09 PM, Phil Spitler = wrote: >>=20 >>> Wow, working with 5k HD Red footage in CS5.5 is REALLY painful. >>>=20 >>> Has this been improved in CS6?=20 >>>=20 >>> It must be the RED decoding that is taking the time as I use 5k = plates from our 3d guys without this issue at all. >>>=20 >>> I did a simple garbage matte / Primate key and resize and it is = saying 2.5 hours to render a 10 second clip. >>>=20 >>> Plus the machine is really un-responsive.... >>>=20 >>> This is on an 8 core Xeon with 16GB RAM running CS5.5 >>>=20 >>> Any idea? >>>=20 >>> Thanks >>>=20 >>> Phil >>>=20 >>>=20 >>>=20 >>>=20 >>>=20 >>>=20 >>> Phil Spitler | Creative Technologist | Bonfire Labs | t : = 415.394.8200 m : 415.571.3139 >>>=20 >>> Website | Facebook | LinkedIn >>>=20 >>=20 >>=20 >>=20 >>=20 >> --=20 >> brian higgins | creative director >> Sol Design >> 312.706.5500=20 >> higgins@soldesignfx.com >> soldesignfx.com >>=20 >>=20 >>=20 >=20 --Apple-Mail=_A7E49E91-45A8-4296-B3A7-9AFEC003CA60 Content-Transfer-Encoding: 7bit Content-Type: text/html; charset=iso-8859-1 Scratch that.... the test without scaling is also taking FOREVER.

Phil

signature1 Phil Spitler  |  Creative Technologist   |  Bonfire Labs  |  t : 415.394.8200  m : 415.571.3139

Website | Facebook | LinkedIn

On Oct 15, 2012, at 5:51 PM, Phil Spitler wrote:

Thanks Brian.

We are delivering 1920x1080 but are doing some serious pushes on the footage.

It seems like the scaling could be a big issue as I just rendered a 5k clip and it is 3 times the speed as rendering the same clip at 1080 (after scaling the footage).

I will look at the debayeing settings too to see what I can figure out,

The reason I wanted to keep the files R3d was so that I could do the color and keying with maximum information in the file.

More testing to come.

Thanks.

Phil





signature1 Phil Spitler  |  Creative Technologist   |  Bonfire Labs  |  t : 415.394.8200  m : 415.571.3139

Website | Facebook | LinkedIn

On Oct 15, 2012, at 5:45 PM, Brian Higgins wrote:

Phil, what resolution is your deliverable?  Jack is right on the money.  Just because you *can* debayer on the fly, doesn't mean you should.  Heck...if you're not doing push-ins and repos, you probably don't even need to do a full debayer!  If you have fast and large enough storage, do the debayering when you do the conform and write out DPX/TIFF/EXR sequences of what you need.  You'll be much happier!

Here's a test for you to try if your deliverables are (only) HD like mine are. It's somewhat footage dependent, but might be enlightening:  Load the same r3d file twice.  Debayer the first copy at full 5K, then debayer the second at half premium.  Drop them both into 1920x1080 comps, and scale them so they fill the HD frame.  Now toggle between them and see how much difference you can spot. I can sometimes see a tiny, tiny sharpness improvement on the full-debayered version, but a lot of the time I can't.  Now put a little sharpening (should be standard operating procedure on Red and Alexa footage) on both comps and see if you can tell the difference.  If you've got 200% blowups to deal with or are mastering at greater than HD resolution you obviously need the full debayer, but if you aren't, you might not.

As to your render times...have you tried rendering a chunk out of RedCine? A full debayer should only take a few seconds per frame there, even without a RedRocket.

$.02,
Brian


On Mon, Oct 15, 2012 at 7:16 PM, Jack Tunnicliffe <jack@javapost.ca> wrote:
Depending on what you are doing with the Red files it's sometimes better to render it out as an image sequence, say 16 bit tiffs or something similar so you computer isn't doing the debayer for every frame while you are working with it. I find the only reasonable explanation for sticking with the R3D data is so you can get at the raw data at the initial stages should there be under exposure, over exposure, etc. You can find detail that would otherwise be lost, but really after that, why not convert it to a more AE friendly format. 

Jack Tunnicliffe
Java Post Production
www.javapost.ca



On Oct 15, 2012, at 6:09 PM, Phil Spitler <phil@bonfirelabs.com> wrote:

Wow, working with 5k HD Red footage in CS5.5 is REALLY painful.

Has this been improved in CS6? 

It must be the RED decoding that is taking the time as I use 5k plates from our 3d guys without this issue at all.

I did a simple garbage matte / Primate key and resize and it is saying 2.5 hours to render a 10 second clip.

Plus the machine is really un-responsive....

This is on an 8 core Xeon with 16GB RAM running CS5.5

Any idea?

Thanks

Phil






Phil Spitler  |  Creative Technologist   |  Bonfire Labs  |  t : 415.394.8200  m : 415.571.3139

Website | Facebook | LinkedIn





--
brian higgins | creative director
Sol  Design





--Apple-Mail=_A7E49E91-45A8-4296-B3A7-9AFEC003CA60--