Vegas Workflow
 

I tried loading some 5D footage into Vegas on my Q6600 system with a 4-drive RAID, and I don't get real-time playback on a full screen monitor. From experience, I could render out to uncompressed 60i (and, I assume 30p), and I can work with one stream, but not two.

Here are the options that I can think of:

1) Work with limited frame rate. (Yuk.) Pre-render areas of interest.

2) Get Vasst's Gear Changer. Render smaller proxies after capture, manually. (Wide DV?) Work with proxies, until the final render. Not too expensive, but not ideal.

* Get Cineform Neo HD for $600. Ouch. That's more than the NLE! I guess this is the Windows equivalent of ProRes.

* Work from a 4-drive RAID uncompressed. It's fine if you don't do any layering. The problems are disc cost, pre-rendering, and LOTS of data. Not good if you have lots of footage.

* I have Raylight from a previous HVX project. Can we encode MPEG-4 files to Raylight, and use this as an intermediate format? When rendering from "maximum quality" is there loss? Can we use this with Gear Changer to use Raylight only for proxies, but to render straight from the originals?

From RED's zone plate demonstration, we can see that the 5D MkII has chroma aliasing problems with high frequencies in the vertical dimension. It's possible that there is a way to apply an effect to the original footage to re-site the RGB pixels for highest quality. The workflow might be like this:

1) Capture.
2) Trim out the garbage and find the useful material as desired.
3) Apply the vertical HF fixer effect.
4) Render to the intermediate format (Cineform/Raylight/Uncompressed/WideDVSomethiingElse?)
5) Edit away, possibly with Gear Changer

Comments?

Although a bit crippled, vegas includes a free version of cineform. Here's one option then:

1. Drop all project-related clips into a vegas timeline.
2. Use a script that can make a region out of every event (a free script can be found deep in the bowels of the new vaast site, find free scripts or somesuch)
3. Do a batch render from vegas (built in script) to cineform intermediate (choose avi, then pick an intermediate preset, then click custom to tweak it as necessary) that will render out each region to an individual cineform clip (I actually downres to 720p30 cineform at this point because it's plenty good for most purposes).
4. Save the project file for making the intermediate clips (this will make sense in a minute)
5. Start a new project and work with the cineform clips, make sure project settings are correct (match the cineform files)
6. Edit and render out a final master, and save the final project file
7. Once complete, you can throw away all the cineform files, keep the original 5d files and the two project files (the final project and the project you made the intermediates from)

This is free (if you already have vegas), this gives you a real time edit, this allows you to archive only the original files, and leaves you in a spot where if you had to repeat the whole thing you could (just load up the intermediate project file, render out new intermediates, load up the final project file, and render out a new master). Alternatively, you could throw away the originals and the intermediate-generating project file, and just keep the cineform files and the final project file.

Sounds complicated, but it's actually not that bad.

Like you said though, GearShift is another viable solution. And, don't forget, you could go with the Neo HDV cineform app if you were willing to edit/final render in 720p -- that app is only $250 I think. Using HD Link in Neo HDV, you should be able to read in 1080p clips and convert them out to 720p clips, so that might work for you and is only $250 I think. Me personally, I've been using the above workflow without issue.


OOPS, UPDATE:

I forgot to mention that this workflow also has the advantage of being able to do a really rough edit BEFORE creating the intermediate files. I do this, and it's really handy. Say you've got a clip you know you won't use, then just delete it from the timeline. Or say you want to trim certain clips, like you have a ten min. clip but know you only want 1 min. of it. Then trim it. Once you have all the raw footage you know you need, THEN render out the intermediates. That can cut way down on the intermediate render times and extra disk space needed.

Thanks Bill!

I did a quick test, manually setting regions, then manually rendering them, and it worked well.

I tried rendering 1080p/30 Cineform files, but it crashed. Rendering 720p/30 was fine. The results were real-time enough (maybe some frames were dropped, I'm not sure) with two streams, and the visual quality was high enough, that I'm happy with the results.

I'll probably get the gear shifter, because I like the idea of rendering from the originals. Vegas generally uses all cores during the render, so it should work okay. When previewing, Vegas seems to want to use one core on my system.

I tried DVFilm Maker (comes with Raylight), but it failed to render the files. The README has a number of detailed steps that I didn't follow on the first pass. I'll try a more careful pass when I get a chance.

Regardless, the free Cineform method is VERY workable and looks good.

Thanks for the detailed instructions!

Saw elsewhere that David Newman indicated HDLink will only work in Aspect and up for using this file at 1080i.. I took that to mean NeoHDV will not work for full 1080i. But I took what I thought was Chung's raw file, and still couldn't convert, even to 720.

It did work inside Vegas though.

I could be wrong, but I believe I was told by Cineform that Neo comes with a decoder that can decode any cineform file, but the ENCODER is limited by which version you buy. That's why I said you could only use Neo if you were willing to drop to 720p with your intermediate and then render out from there.

NEO HD has a maximum encoding rez of 1920 x 1080. NEO HDV has a maximum rez of 1440 x 1080 (1.33 PAR). The former is recommended for h.264 files, but the latter will still work, as long as you set the resize option to 1440 x 1080 in Cineform HDLink. You could also drop all the way down to 720p, as you stated. BTW, the basic limitation is being able to decode the original file, so you better have a good h.264 encoder installed. Adobe Premiere CS3 or later comes with the MainConcept h.264 decoder and a nice freeware decoder is CoreAVC HD. And then there's QT. Nero also provides one, but I don't recommend installing Nero, just to get the decoder.

Bill, can i ask you question - Why do you use the Cineform files as proxy's? I thought the Cineform files themselves would be good enough for a final render?

I'm with Mike on this question.
As I understand it, the Cineform files are significantly better all the way 'round for editing, grading and finishing. I'm wondering why one would go back as well.
I know I'm missing something...

Not mentioned is the 30 fps vs. 29.97 fps issue. I suppose it's best to keep everything in 30fps until the final render. That way you can go back and forth between any proxies/intermediates and the original without having to worry about keeping track of rates.

Regarding inters/proxies, I personally prefer to do the final render from the original (or uncompressed video), rather than from an inter, because it potentially removes a processing step from the equation. An inter can never have higher quality than the original - just higher efficiency.

Also, the original is 1920x1080. The free Cineform inter in Vegas only supports 1280x720 (and maybe 1440x1080), but not full res video.

That's not to say that Original -> Cineform -> Final isn't good enough for a given product.

Then again, I recently finished a "scientific" video production. Step 1 was to decompress everything. I then converted everything to a common size and frame rate (1080p). The next time I saw any compression was at the final render. Oh yeah, and it was boring. ;)

IEC - News releases > nr3708 - World's most boring DVD and Blu-rayTM to measure energy efficiency of new generation TVs

BTW, I used Vegas for all editing and level matching. I used After Effects for all size and rate conversion on that project.

Jon, thank you in advance!

DO you get a 16-235 in vector scope and wave form, (what produces common for many camera clips crushed blacks and clipped highlights) or full 0-255 in Vegas?

Vegas interprets 5D Mark II clips incorrectly. So, others NLE on PC platform.

Let's test 5D Mark II RAW-sample from dpreview: http://s3.amazonaws.com/movies.dprev...kII/Video2.MOV

Let's use Vegas Pro 8c standard project "HD 1080-24p (1920x1080; 23,976 fps)" with "Frame rate" set to 30. Also let's set "Preview Window Quality" to "Best-Full".

Vegas interprets origin .MOV-file incorrectly: blacks is crushed down; details on highlights is lost (see the sky):
http://www.videomax.ru/forum/uploads.../IMG_Vegas.jpg

Let's use Carbon Coder utility to covert origin .MOV-file to .AVI, CineForm HD v3.6.2 (1920x1080, progressive, 30 fps).
Now Vegas interprets .AVI-file more correctly: details in black areas; details on highlights appears:
http://www.videomax.ru/forum/uploads...IMG_Carbon.jpg

And let's look at Vegas histograms.
Histogram of origin .MOV-file shows gaps on color channels and abrupts on end's:
http://www.videomax.ru/forum/uploads...HIST_Vegas.JPG

Histogram of coverted .AVI-file without gaps:
http://www.videomax.ru/forum/uploads...IST_Carbon.JPG

As I see, The Carbon Coder utility is the only tool with correct interpretation of 5D Mark II video on PC platform.

Try MPEGStreamClip (Squared 5 - MPEG Streamclip video converter for Mac and Windows) - I used it to test conversion to ProRes on back-to-back histograms in FCP and get very similar results to the samples you posted. It's free and available for mac & pc.

Not sure I ever said I did? I might of said one or the other but not both (proxy or intermediate)? Anyway, no matter, when I use cineform, I render out from the intermediates (no proxies). Good render times too that way. But I've also experimented with poxies (a la gearshift) too. I do a lot of web stuff, and I feel cineform is perfect for that when needed. For those purposes, I edit in 720p30, but then I'll frameserve out the final render at half that size for flash encoding for the web, nice clean resize to 640x360 from those 720p cineform files. At that point, for archival, you have to decide whether to keep the originals or the cineform files. I tend to throw away the intermediates and keep the originals and retain my ability to repeat the whole workflow again in the future if needed.

Evan,
Thank you, it really works! Great news!

My Bad, Sorry Bill! - I should have read more carefully. I now see (#7) you keep your original 5D clips and project files for archival purposes. And you do render from your intermediates(#6)

Ya, Carbon Coder costs 5000$ ...I could buy a Mac for that price!

NewTek SpeedEDIT interprets files correctly if anyone is interested. To make the color space fit inside RGB colorspace just check the "Non ITU-R BT601 color" checkbox in the Layer Settings dialog :)

[QUOTE=Oleg Kalyan;978970]I see Carbon Coder, a universal transcoding application for $20 in a few places

Ohh... sorry, I wonder if the cost i saw on their website is a business license cost?

http://www.rhozet.com/rhozet_priceList.pdf

Can i ask a question Oleg - is it a effects plugin? Do i add it to my effects before transcoding to, say Cineform? Or do i need to use the standalone app?

In Vegas it's simple to translate from Computer RGB to Studio RGB. In an 8-bit project, apply the color corrector plugin to your track and select the "Computer RGB to Studio RGB" preset.

There has been a lot of conjecture that the 5D MkII was crippled for marketing reasons, but I'm thinking that this was a case of photo engineers making a video product. They did a great job with imaging, but didn't know how much they didn't know about video.

The evidence?
* 30p, rather than 29.94p (let alone no 24/25, etc) NOTHING is 30p in the video world.
* 0-255, rather than 16-235. Sure, it's better for images (a few more levels), but it's not standard video. If you want more levels, give us 10-bits.
*Nutty algorithms like making the shutter 1/focal length. It's bad enough that they didn't give us manual control, but to choose stuff like that is just plain off-the-wall.

In Japan, project teams have a lot on internal loyalty. I get the feeling that the video function was developed by the camera team with complete secrecy from the video teams. It might have helped politically, but led to some "odd" technical decisions.

It seems that "$20 Carbon Coder license" is not for business purpose. Maybe only for "home" usage (I hope you understand what it's means).


The Phozet Carbon Coder utility retains full level range of 5Dmk2 video (Upper histogram).
Other NLE's lose information: red areas on histogram's ends; gaps in the histogram. The result - too contrast picture, crushed down blacks, highlights without details.
http://www.dvinfo.net/conf/attachmen...dmkii_hist.jpg


I know only one way to resolve the "contrast" problem on PC platform:
- convert .MOV-files to .AVI-files by Phozet Carbon Coder;
- work with resulting AVI-files in your NLE.

P.S.
I have no NewTek SpeedEDIT, so can't comment.
The MPEGStreamClip utility was tested. It interprets 5Dmk2 video incorrectly (lose information the same way as others NLE's).

Jon, I tryed this way. It's not resolve the "contrast" problem of origin MOV-files.

The origin video and histogram:
http://www.videomax.ru/forum/uploads.../IMG_Vegas.jpg
http://www.videomax.ru/forum/uploads...HIST_Vegas.JPG


The origin video (translated from Computer RGB to Studio RGB):
http://www.videomax.ru/forum/uploads...oStudioRGB.jpg
http://www.videomax.ru/forum/uploads...oStudioRGB.JPG

The histogram remains the same form. It was squeezed, but without appearance of new details (see sky).



The video after Carbon Coder and histogram (see details on the sky):
http://www.videomax.ru/forum/uploads...IMG_Carbon.jpg
http://www.videomax.ru/forum/uploads...IST_Carbon.JPG

Look at appearance of histogram. Now we adds new areas on ends. And gaps disappeared.


Finally let's try "Computer RGB to Studio RGB" to Carbon Coder's video:
http://www.videomax.ru/forum/uploads...oStudioRGB.jpg
http://www.videomax.ru/forum/uploads...oStudioRGB.JPG

Feel the difference.

Ivan: Am I understanding this right: What your last example appears to show me is that the detail is not lost in Vegas, its just a matter of applying the proper manipulations through filters, or post Vegas processing to get to that point. Question I have is there away to do this in Vegas without resorting to Carbon Coder.

And by the way, is this OEM outfit legitimate ?

This is what SpeedEDIT looks like when the footage is first imported, the switch to correctly interpret color and the resulting image, all with RGB waveform to verify clipping. The big plus is that you can start editing right away instead of waiting for any transcoding, not to mention the quality hit/extra space.

Ivan,

Your examples shows the problem clearly. The source material is encoded as 0-255, but the Quicktime decoder seems to expect 16-235 AND it is stretching 16 to 0 and 235 to 255. The decoded clips in Vegas are clipped as soon as decoding occurs.

I verified the incompatibility between the 5D MkII output and Quicktime Pro by exporting as a PNG. I get the same results (clipped clouds) on the PNG as I do in Vegas.

If somebody has Quicktime Pro on a Mac, it would be interesting to see if this is a problem on that platform too.

Either the 5D is encoding to the wrong levels, or the metadata is incorrect. It's probably the latter, considering that other decoders can handle it.

Aside from buying new tools, it seems that one solution would be to create a custom profile in the camera. This will reduce dynamic range slightly (219 vs 255 => 14%), but will avoid clipping.

Here's the PNG export direct from QT. Note the flat/clipped clouds: http://colcrush.com/jon/MarketMan_Frame001.png

This is a long-standing issue with quicktime - it first showed up years ago when people noticed luminance shifts in renders with DV footage. The 'solution' was to give FCP the option to render in YUV and use the full range, but it didn't solve the underlying issue of quicktime forcing RGB into the 16-235 range. Now we're working with a camera that encodes to an RGB codec and the problem is just showing up - hopefully this will prompt Apple to do something about it, but considering how long it's been around without them really solving it I wouldn't hold my breath.

I tried TMPGEnc 4.0 Express (a reasonably flexible encoder of various formats for $99), but it had the same problem. It also relies on the Quicktime file decoder to open the .MOV file.

It seems that custom presets are the only workaround, short of buying a custom ($$$) Quicktime decoder.

It would be interesting to know if there was something in the file.MOV metadata that could be tweaked to make Quicktime decode it properly...

Jon, you are the Man! You can understand my poor English!

I think you are right here: "The source material is encoded as 0-255, but the Quicktime decoder seems to expect 16-235 AND it is stretching 16 to 0 and 235 to 255. The decoded clips in Vegas are clipped as soon as decoding occurs."

You wrote: "Either the 5D is encoding to the wrong levels, or the metadata is incorrect. It's probably the latter, considering that other decoders can handle it."

I also think that problem is with metadata.

I don't know who should resolve the "5D MOV metadata" problem. Either Canon, or NLE's manufacturers, or Apple with Quicktime.

I wonder what is a secret with that Phozet Carbon Coder, it can correctly decode 5Dmk2 MOV-files.

Ok so what does this mean? I'm confused. I dropped the raw 5D file in, set my project properties to match file...? I don't see any weird spaces in my histogram. Am i doing something wrong?

Ivan, your English is perfectly clear. Thanks for demonstrating the problem with the 5D MkII output.

Mike, I saw the same result (no gaps) at first. Make sure that your project is 1920x1080 and that your preview is Best and Full. Otherwise, Vegas will do some filtering and smear the values. Even without the gaps, you can see the peak at black.

DOH!!

I see it now! Thanks Jon!

Stupid Vegas, you'd think that the scopes would be independent of the preview quality!!! That should be a no brainer Sony!

[link to web site removed.]

No, $5000 is the normal cost. The reason it appears for just $20 on that web site is because they are making illegal copies and profiting from copyright infringement.

Sorry about posting the link, I saw it for $20 in a few places.
Did not know they were illegal.

So I posted this issue over at the Vegas forum, seeking a resolution to this clipping issue there within Vegas, but no one has really come up with an answer.

As I stated a bit further up in the thread, the "almost free" CoreAVC decoder can be used to replace the QT codec (but I would try the demo first, to make sure it works with the 5DMKII files):

CoreAVC.com: World's Fastest High Definition H.264 Video Software Decoder

(click on the first screen shot on this link)

As you can see, you can select either computer RGB or Studio RGB for both input and output. Note, that you have to make sure that the Direct Show file merit for CoreAVC is higher than that for QT, as the file merit dictates which codec is used by default by Windows. When you install CoreAVC it should automatically have a higher file merit than QT. CoreAVC will work with Premiere, After Effects, TMPGenc Xpress and any other Windows-based NLE that uses DirectShow. Unfortunately, Vegas uses the "older style" Video for Windows codecs (VFW) and it uses it's own built-in h264 VFW decoder. Vegas will not use CoreAVC.

this seems like an interesting alternative but what would be the most efficient way to work it into a Vegas workflow?

I was thinking there might be a way to manipulate the files into the timeline using a combination of avisynth and something that proxies the avisynth output into a standard avi file. To this end, I've tried 'MAKEavis' from ffdshow but it hasn't worked very reliably. One thing I'm unclear of is how reliably avisynth can parse the quicktime files and then how to properly promote CoreAVC so it is used instead of quicktime's own decoder.

Has anybody else figured this out? I want to avoid using intermediate files if at all possible because that would add several additional layers of hassle and resource intensiveness to the whole process. It's almost 2009. I just want to edit video and not jump through hoops to make stuff happen.

Wow this is tough. I'm looking to get my hands on the 5D to work with the EX1. However the post job looks very intimidating at the moment!

Cineform HD
 

I see where they have lowered the price on Neo HD from Cineform. It is know around 499.00 and 199.00 upgrade if you have had Neo HDV which is now being retired. This should enable you to turn the 5d files into cineform AVI's for PC or MOV for Mac.

Anyone tried mxing mxf from the XDCAM with the mov from the 5D2? Is it workable or does the timeline mess up?

I've never gotten Makeavis to work, with any footage at all, since I first heard about it. Thankfully, another method has since popped up that does exactly what you describe.

Not having access to a Mark II, I downloaded the original version of this video from the Vimeo page to play with, and though I can't get my quad core XP machine to play it back in realtime, I've successfully loaded said file in various applications normally unfriendly to Avisynth through the instructions I've laid out here: http://www.dvinfo.net/conf/canon-eos...ows-users.html

I have the Vegas Pro 8.0 trial installed and though not realtime, once finished building peaks the program seems to work with the footage at an almost-workable speed. Your definition of "almost-workable" may vary significantly from mine, of course, so I wouldn't get my hopes up.