And about the codec? Can it be considered as a professional codec?

My opinion is that it is a "usable" codec.

But "professional" is a fairly ambiguous term. If compared to non compressed codecs from very high end systems.... then no.

But.... so far it does handle a bit of tweaking in post.... and I've already made money doing "professional" projects with it.... so yes?

IMHO, the codec is excellent for the price, considering that it's real-time and battery-powered. The results are small enough to stream onto any decent CF card, yet nice enough to give a good result.

If you ask me, they nailed it.

Sure, REDCODE or equivalent would be better, but if you want that, sell the 5D MkII on eBay next year and get a Scarlet...

How would the codec compare with HDV?

the bitrate is higher than HDV, so the quality is probably a bit higher. H.264 usually comes out great.

How would the codec compare with HDV?

Quality from the 5dmkII's going to be significantly higher than HDV:

HDV uses mpeg2 compression, the older standard that DVDs and original DV uses. HDV 1080i uses a recording data rate of 25 Mbit/s (3.125 MB/s) while HDV 720p records at 19.7 Mbit/s (2.46 MB/s).

The 5DmkII uses mpeg4 based compression, which is newer, higher quality for the bitrate, and is used in things like DirecTV's newer HD channels and AVCHD. The 5Dmk2 has a recording data rate of somewhere around 47 Mbit/s (5.6 MB/s).

Ergo, it's using a more efficient codec (higher quality) at a higher bitrate (higher quality).

For comparison, newer AVCHD cameras like the HF11 have a super-high quality mode that uses AVCHD's highest bitrate setting, 24 Mbit/s, but that's more than 20 megabit less than the 5DmkII's laying down for the same resolution.

The 5DmkII's going to be the highest compression quality you can get without spending significantly more money.

much better than the hdv I get out of my archived z1u and hv20 footage.

not quite at artifact free when tweaked as my ex1 ( 5d2 footage can get a bit of banding if not converted to prorez before tweaking or adding grads )

The 5DmkII uses mpeg4 based compression, which is newer, higher quality for the bitrate, and is used in things like DirecTV's newer HD channels and AVCHD. The 5Dmk2 has a recording data rate of somewhere around 47 Mbit/s (5.6 MB/s).


Actually I'm pretty consistently seeing 39 Mbit/s, which is still quite a bit more than HDV.

The only drawback of the codec I've seen so far is that it's not treated as a native format in FCP so you don't get any realtime transitions, effects, etc. Converting to prores works well, unfortunately I prefer to work on my laptop and the conversion is slower than real time. I expect my workflow will be cuts-only native, then media manager transcode to prores only the clips used in my rough and then finish to prores master.

Actually I'm pretty consistently seeing 39 Mbit/s, which is still quite a bit more than HDV.

Yea, that's why I said somewhere around, I think the initial reports were math inflated. In any case, even at the same bitrate MPEG4 would look a whole lot better than MPEG2, and the 5D even outdoes the max of AVCHD by 15+ megabit.

h.264 allows very advanced efficient coding. You can get great results in 15Mbps. But that requires a top-class fixed hardware (typically 2RU) or slower than real-time software implementation.

The h.264 encoded in the 5D MkII likely doesn't implement many of the tricks that the format offers, but at 47Mbps, it doesn't have to.

It's definitely better than HDV, which is MPEG-2, has fewer bits-per-second, and is only 1440 wide.

Something to be aware of when checking out sample clips online is that many people are intentionally pushing the limits of the camera in low light because it's so good at that. However this does potentially introduce multicolor noise, especially in shadows, which is easy to mistake for compression artifacts. If you really want to evaluate the codec quality (independent of the camera noise) you should look for well-lit or daylight clips shot at something less than 1600 ISO.

Good point, Evan,

Also, one should look at content with smooth gradients as well as other video lots of texture/detail. Also, study images with little motion, and others with high motion and fast shutter. This will give you an idea of where the codec shines, and where it falls apart.

Frankly, the D90's poor compression of texture is what drove me to the 5D MkII.

Does video from 5D Mark II records in sRGB color spase, is it 0-255 ??

My colleagues see it 16-235 in vector scope and wave form, thus crushed blacks and clipped highlights, what we see in pretty much all examples presented.

On Windows platform they switch to Enable Direct3D video acceleration in QT.

Apparently, a plugin for OS X platform has to be developed to utilize full image quality of the camera..

Does video from 5D Mark II records in sRGB color spase, is it 0-255 ??

My colleagues see it 16-235 in vector scope and wave form, thus crushed blacks and clipped highlights, what we see in pretty much all examples presented.

On Windows platform they switch to Enable Direct3D video acceleration in QT.

Apparently, a plugin for OS X platform has to be developed to utilize full image quality of the camera..

It is 8 bit sRGB. The A/D converter is 14 bit.

As quicktime is apples product, it's unlikely it needs a plugin for OSX. All recent macs use the graphics card to decode h.264. Perhaps windows machines need a little help from the plugin.

Don thank you,
mu colleagues had a hard time to get full 0-255.
Is there a workaround, to you see it, are you on Mac?

Don,
thank you,

my colleagues had a hard time to get full 0-255, on different editing platforms,
it seem to always cut it off to 16-235

Is there a workaround, do you get 0-255, are you on a Mac, which editing program?

sorry for double post..

We just got one yesterday, and after a bit of testing, I would say that although the bit rate is higher, it's just about the same as HDV in quality--maybe slightly better, and I would compare that to Canon's HDV implementations, not Sony or JVC.

I think it has a higher bit rate than prosumer AVCHD because it's not doing quite as much efficiency encoding as, say, the 24mbs AVCHD of the HF11.

It's definitely decent quality, but I see artifacts on certain scenes with fast movement, just like HDV.

Oleg,

On NewTek SpeedEDIT I can see the full 0-255 range. With a little contrast adjustment it can be made to fit into 8 bit RGB without clipping. Maybe there's a setting on other products to interpret the colorspace correctly and get the whole range. I've been seeing a lot of clipped video going around so it appears to be a problem for most users.

Luis, thank you!

I can't wait to see a solution, fix, on FCP platform !

I think it has a higher bit rate than prosumer AVCHD because it's not doing quite as much efficiency encoding as, say, the 24mbs AVCHD of the HF11.

It's definitely decent quality, but I see artifacts on certain scenes with fast movement, just like HDV.

How surprising, 24mbps on HF11 better than 40mbps on 5DMII ?
Maybe that's because the HF11 has variable bitrate, I'm not sure 5DMII has it.

Do you have a screenshot with movement artifacts ?

I think it has a higher bit rate than prosumer AVCHD because it's not doing quite as much efficiency encoding as, say, the 24mbs AVCHD of the HF11.



From specs H264 should be better quality as its designed to need more processing power. It would have been a lot easier to implement HDV.

Isn't avchd and h264 implementation also?

It would have been a lot easier to implement HDV.Nope, because then it would have to be a tape-based camera, per the provisions of the HDV consortium.

What about XDCAM, isn't a faster yet similar codec than HDV? Can we expect Sony to offer it with their new photo camera line?
How would that compare with the Canon codec? (maybe I m getting little bit to speculative)

The 5D Mark II has a few things that I'm disappointed in. The codec isn't one of them. The only appreciable step up for me would be a compressed RAW option. h.264.MOV, HDV, AVCHD and XDCAM are all just different flavors with the similar bit-depth limitations.

How well will the HF11 30p AVCHD edit with the 5DII in FCP? I assume colors can be made to be pretty close. Is FCP all happy with AVCHD now?

Seems like a nice camera to have with the 5DII. I've never even seen one of the HF models.

From my observation, the codec seems really nice, except it doesn't really hold up well in the shadows, which is, incidentally, where digital hold most information. The solution is simple though. If you want something good for post grading, make a picture style that expands the shadows and compresses highlights a bit to make this information fall in more useable segments of the codec. This is particularly useful for very contrasty situations such as backlit scenes.

Here's such a picture style, in case anyone's interested.

A few weeks later . . . do you stand your prior statements about the quality of the codec?

My only testing so far with the video was a night shoot of local Christmas lights and some statues in a veteran's memorial.

The codec handles individual frames pretty well, so that a still pulled from the video would look good (I saw some slight banding in a clear blue sky as the gradient changed on the horizon, but nothing earth shattering). But from my tests, it's fairly clear that this codec is not optimized for motion as there is very slight mosquito noise even in shots that are on a tripod if there's high contrast AND high detail at the same time. Panning across a bronze, lit statue at night, for example, where there's a lot of luminance contrast and fine detail in the wrinkles of clothing and texture on the statue reveal this problem. Even some areas of grass lit by the vapor lamps showed this when the tripod was completely locked down. There's just a jittery quality to the fine details that appear to be due to shifts in the compression from frame to frame that I find annoying.

But from my tests, it's fairly clear that this codec is not optimized for motion as there is very slight mosquito noise even in shots that are on a tripod if there's high contrast AND high detail at the same time. Panning across a bronze, lit statue at night, for example, where there's a lot of luminance contrast and fine detail in the wrinkles of clothing and texture on the statue reveal this problem. Even some areas of grass lit by the vapor lamps showed this when the tripod was completely locked down. There's just a jittery quality to the fine details that appear to be due to shifts in the compression from frame to frame that I find annoying.
What's the best post-production work-around in your opinion?

There are statements in this thread speculating that the 5D MKII shoots in sRGB (16-235). I thought that issue was resolved before and that the camera shoots video in cRGB (0-255):

http://www.dvinfo.net/conf/944824-post22.html

The problem is that the QT decoder is clipping the histogram and producing sRGB in most NLEs. If you install the DirectShow h.264 (MPEG4 Part 10) decoder, CoreAVC, you can decode the 5D MKII MOV files as cRGB in any software package that doesn't force the use of QT for decoding, such as Cineform.

Also, MPEG4 is a far more efficient codec than MPEG2. What this means is that for a given bitrate, MPEG4 will always be of higher quality than MPEG2. But there are various implementations of the MPEG4 standard, so people see different results when making comparisons between the two codecs. The point of the MPEG4 standard was to make more compact files and video streams than MPEG2, not higher quality. The limitations were either the capacity of an optical disk or the bandwidth available for cable or satellite broadcasting. To really compare specific implementations of the two codecs, you would have to conduct a double-blind viewing test to see whether a large population of viewers find the level of compression artifacts acceptable or not.

I think the best and easiest workaround for getting the full range clip into finalcut is recapturing from the 5d´s hdmi output.

I playback all videos on my cf card with the slideshow (only movies) function of the 5d. I convert the hdmi signal from the 5d with a convergent design convertor to hd-sdi and capture with a blackmagic hd card directly into the prores codec.

That means the process takes as long as the clips on your cf card last.

converting with any other method in compressor or color is not faster.

Works perfect for me. Easiest approach would be using a capturing card with a hdmi input like the blackmagic intensity. Then you don´t have to convert the hdmi signal to hd-sdi.

Regards
Holger

I think the best and easiest workaround for getting the full range clip into finalcut is recapturing from the 5d´s hdmi output.

I playback all videos on my cf card with the slideshow (only movies) function of the 5d. I convert the hdmi signal from the 5d with a convergent design convertor to hd-sdi and capture with a blackmagic hd card directly into the prores codec.

That means the process takes as long as the clips on your cf card last.

converting with any other method in compressor or color is not faster.

Works perfect for me. Easiest approach would be using a capturing card with a hdmi input like the blackmagic intensity. Then you don´t have to convert the hdmi signal to hd-sdi.

Regards
Holger

Why? Isn't it the same than a simple transfer from the CF card to your NLE?

No its not the same

If you transfer from your CF card directly into Finalcut (or any other NLE using Quicktime codecs) you get the strange effect of clipped blacks and whites cause not the whole 0-255 range of the signal will be used. Only 17-235 will be used.
In short, that means clipped blacks and whites unless you use one of the software workarounds posted in an other thread in this forum.

If you recapture from the hdmi output of the 5d this effect will not happen.

I think the software workarounds take longer and are more labour intensive. At least if you want Apple Prores as your target format.

I think the best and easiest workaround for getting the full range clip into finalcut is recapturing from the 5d´s hdmi output.

I playback all videos on my cf card with the slideshow (only movies) function of the 5d. I convert the hdmi signal from the 5d with a convergent design convertor to hd-sdi and capture with a blackmagic hd card directly into the prores codec.

That means the process takes as long as the clips on your cf card last.

converting with any other method in compressor or color is not faster.

Works perfect for me. Easiest approach would be using a capturing card with a hdmi input like the blackmagic intensity. Then you don´t have to convert the hdmi signal to hd-sdi.

Regards
Holger

Hi Holger,

I took a look at the specs of the Blackmagic Intensity Pro and it seems that it captures 1080i 59.94fps. Is this an issue, given that the 5D files are 1080p 30fps?

Also, what is the difference in file size?

I've heard that capturing in this manner makes real time editing and effects a lot easier so I'm intrigued by this option, as well as the advantage of recovering the full range.

(I took a look at the specs of the Blackmagic Intensity Pro and it seems that it captures 1080i 59.94fps. Is this an issue, given that the 5D files are 1080p 30fps?)

No its not an issue, because the pictures are already progressive when they come out of the 5d. So you see no difference.

(Also, what is the difference in file size?)

That´s a huge difference. I can put about 40 minutes of video on a 16GB card in the 5d.
When exported those 40 minutes take about 4 times more harddisk space (64 GB)

(I've heard that capturing in this manner makes real time editing and effects a lot easier so I'm intrigued by this option, as well as the advantage of recovering the full range.)

Yes working in prores does really work very well and fast (at least on a quadcore macpro).
Working in the H264 codec that the 5d outputs is painfully slow.

I can live with this strange approach of taking the nonlinear clips on the cf card and redigitizing them as in the old hdv times. You don´t have to convert anything and you don´t loose information.

Well, it appears that there still are some questions, as to whether the 5D MKII shoots in cRGB or sRGB. Here's another discussion:

CineForm Insider (http://cineform.blogspot.com/)

What's the best post-production work-around in your opinion?

I'm not sure there is a work around for the problems I mentioned. I must say, however, that my second set of tests, shot yesterday in daylight, show substantially less of the fine jittery noise than I noticed in some of the night shots. I won't know what factors most contribute to its presence until I've done more filming under a variety of conditions and carefully inspected the images afterward.

I downloaded Cineform Neo HD and will likely end up buying it, at least by the time I get a RED Scarlet later this year. When perfected, this workflow will, I believe, represent the best post production experience with Canon files. Full res Cineform encodes played back at full frame rate on my Core 2 Duo/2 GB system, and the dynamic range and gamma were properly handled. Of course it's not cheap ($500).

However the current version of Neo HD will not handle the Canon audio and also will not process files in the included batch processing utility (HD Link) at native 30.00 fps. It worked to encode to 29.97 fps, which is obviously preferable as a mainstream frame rate, but I noticed some mild softening of the image when doing this, so I don't think they will have their Canon 5d support nailed down until the next version of Neo HD is released.

Right now I'm using the Vegas workflow outlined by Keith Paisley in post #45 of the http://www.dvinfo.net/conf/canon-eos-5d-mk-ii-hd/139418-vegas-workflow-3.html thread. So I'm using the stripped down Cineform encoder that comes with Vegas and working at 720p resolution.

I'm not sure there is a work around for the problems I mentioned. I must say, however, that my second set of tests, shot yesterday in daylight, show substantially less of the fine jittery noise than I noticed in some of the night shots. I won't know what factors most contribute to its presence until I've done more filming under a variety of conditions and carefully inspected the images afterward.

I downloaded Cineform Neo HD and will likely end up buying it, at least by the time I get a RED Scarlet later this year. When perfected, this workflow will, I believe, represent the best post production experience with Canon files. Full res Cineform encodes played back at full frame rate on my Core 2 Duo/2 GB system, and the dynamic range and gamma were properly handled. Of course it's not cheap ($500).

However the current version of Neo HD will not handle the Canon audio and also will not process files in the included batch processing utility (HD Link) at native 30.00 fps. It worked to encode to 29.97 fps, which is obviously preferable as a mainstream frame rate, but I noticed some mild softening of the image when doing this, so I don't think they will have their Canon 5d support nailed down until the next version of Neo HD is released.

Right now I'm using the Vegas workflow outlined by Keith Paisley in post #45 of the http://www.dvinfo.net/conf/canon-eos-5d-mk-ii-hd/139418-vegas-workflow-3.html thread. So I'm using the stripped down Cineform encoder that comes with Vegas and working at 720p resolution.
And at native 1080p?

(I took a look at the specs of the Blackmagic Intensity Pro and it seems that it captures 1080i 59.94fps. Is this an issue, given that the 5D files are 1080p 30fps?)

No its not an issue, because the pictures are already progressive when they come out of the 5d. So you see no difference.

(Also, what is the difference in file size?)

That´s a huge difference. I can put about 40 minutes of video on a 16GB card in the 5d.
When exported those 40 minutes take about 4 times more harddisk space (64 GB)

(I've heard that capturing in this manner makes real time editing and effects a lot easier so I'm intrigued by this option, as well as the advantage of recovering the full range.)

Yes working in prores does really work very well and fast (at least on a quadcore macpro).
Working in the H264 codec that the 5d outputs is painfully slow.

I can live with this strange approach of taking the nonlinear clips on the cf card and redigitizing them as in the old hdv times. You don´t have to convert anything and you don´t loose information.
Noticeable blocking? How do you deal with it?

converting with any other method in compressor or color is not faster.


Depends on your workflow and shooting ratios. If you rough cut in the native files to eliminate footage you don't need then the conversion process may be significantly faster than capturing everything upfront. Also, using native files means you can have someone doing that rough cut while you continue shooting - you don't need to tie up the camera waiting for it to capture, and you don't need to have enough CF cards to cover your entire day's shoot.

And at native 1080p?

Hi, Mathieu.

When using the trial version of Neo HD, I encoded to full res, 1920 x 1080p intermediates.

The "light" version of Cineform that comes with Vegas doesn't support encoding to resolutions above HDV (1440 x 1080, 1.3333 anamorphic pixels). I didn't try encoding to that max resolution and instead opted for square pixels at 1280 x 720, but I presume the 1440 x 1080 anamorphic works and that you could retain the full vertical resolution that way.