GH2 transcode with 5D2RGB
- Thread starter Mike_Bluestone
- Start date
Phil Rhodes
Well-known member
It's not that expensive to do anymore. If you're happy to work in 8-bit YUV (under 100MB/sec) it can be done redundantly with four disks.
But here's the thing about redundancy: if you keep only your uncompressed media on the fast storage, then you've got the compressed originals elsewhere, and if the RAID falls over you can simply retranscode. Obviously, I wouldn't recommend that approach for material which is harder (or more expensive) to recover, but that is one of the incidental conveniences of the DSLR workflow.
P
But here's the thing about redundancy: if you keep only your uncompressed media on the fast storage, then you've got the compressed originals elsewhere, and if the RAID falls over you can simply retranscode. Obviously, I wouldn't recommend that approach for material which is harder (or more expensive) to recover, but that is one of the incidental conveniences of the DSLR workflow.
P
original AVCHD
AVCHD after 5D2RGB
Shot with a GH13 and 2x anamorphic. Stretched in AE CS5. Maybe it's too soft to give you an idea. I think it's a significant improvement. Look at the difference in detail on the road. (Mind the red flare, that looks like a reflection on the road.)
I know it's not a GH2, but then i don't have one
Last edited:
vcassel
Veteran
View attachment 28290
original AVCHD
View attachment 28291
AVCHD after 5D2RGB
Shot with a GH13 and 2x anamorphic. Stretched in AE CS5. Maybe it's too soft to give you an idea. I think it's a significant improvement. Look at the difference in detail on the road. (Mind the red flare, that looks like a reflection on the road.)
I know it's not a GH2, but then i don't have one
No, that's fine. It is a significant improvement and it looks great! I think I'll be hijacking my girlfriend's Mac until this gets ported.
Thanks. Just reward her with kisses and love, and i'm sure she'll agree.
Danielvilliers
Veteran
View attachment 28290
original AVCHD
View attachment 28291
AVCHD after 5D2RGB
Shot with a GH13 and 2x anamorphic. Stretched in AE CS5. Maybe it's too soft to give you an idea. I think it's a significant improvement. Look at the difference in detail on the road. (Mind the red flare, that looks like a reflection on the road.)
I know it's not a GH2, but then i don't have one
It is very difficult to make a judgement because the anamorphic lens strech etc which produce a very soft image. The only thing I see is a lower contrast image, but it can be me not seeing the difference.
If you zoom in a lot, the added detail is clearly there. Even at 100% the footage looks a bit sharper.
I too am interested to see the effect on GH2 footage.
The ProRes file is a lot larger than the .mts file. But then that's normal.
Maybe it's possible to correct the colour the original like the 5D2RGB sample, but my guess is that you'd have to work with "two gammas", because you can see the darker areas are lighter, but the lighter areas are less blown out and even a bit darker. I think it's easier to add greater contrast than to remove it. I tried matching the two with one instance of levels and it wasn't possible. Maybe with curves?
But the added detail is what makes 5D2RGB a preferable option imo.
Last edited:
JamesRitson
Member
Transcoding with ffmpeg (using the ffmbc method I posted previously) yields a similar result, which is to say an image that is flatter in contrast:
(AVCHD)
(decoded with ffmpeg into DNxHD)
-----
(AVCHD)
(decoded with ffmpeg into DNxHD)
The footage was shot in extended tele mode, accounting for its apparent softness.
Although a flatter picture might seemingly afford you more leniency for grading, I do not believe that was how the footage was intended to be captured nor viewed. As noted in the above comparison pictures, the perceived increase in quality is simply related to the flatter contrast. In fact, because of this shallow contrast, my initial perception of the ffmpeg-decoded image is that it looks softer.
There is not that much extra detail hiding in the shadow areas of the image. AVCHD, I believe, being an implementation of H264, works strongly on a perceptive quality basis, and so does not capture these areas in any great detail, leading to this "mud" issue that is so frequently debated. Using the cinema picture preset seems to abate this problem somewhat, taking better advantage of the codec's available bit rate.
If I were to be cynical, I would actually argue that flattening the contrast makes AVCHD footage worse for colour grading. It brings out the noise in the shadows, not to mention any compression artefacts that are present (such as blocking and banding), and once you then crush the blacks and tweak other tonal areas during grading, these issues will be more prominent than if you had stuck with the original colour "profile" (for want of a better word).
Having said all that, though, I'd sooner use a decent decoder library like ffmpeg than rely on Quicktime's API, which I gather people are using, as they rewrap their .mts files to .mov files using Clipwrap?
I share your concern about the issue being prevalent with the GH series simply because people will continue to use Quicktime. Because Quicktime, with its inferior H264 decoding, has become popular in its use for transcoding H264 footage, born from Canon's use of the MOV wrapper with their DSLR range, there is a noticeable difference when that same footage is decoded with an API that makes fewer compromises for speed and stability. Windows 7, and possibly previous OS iterations with the correct updates, has strong support for AVCHD through its Directshow API, and the results during real time playback are much closer - possibly identical - to how the clips should look.
The issue is, of course, finding a user-friendly way of using superior decoding tools to transcode your footage. Not everyone finds batch files and command line applications so endearing! 5dtoRGB is certainly a step in the right direction, and would enjoy a boost in popularity with a Windows version. At the current time, if people want to achieve a good result it would certainly be no skin off the nose to give ffmpeg (or ffmbc) a try. It's well documented and far from daunting to use!
Apologies for the essay..
(AVCHD)
(decoded with ffmpeg into DNxHD)
-----
(AVCHD)
(decoded with ffmpeg into DNxHD)
The footage was shot in extended tele mode, accounting for its apparent softness.
Although a flatter picture might seemingly afford you more leniency for grading, I do not believe that was how the footage was intended to be captured nor viewed. As noted in the above comparison pictures, the perceived increase in quality is simply related to the flatter contrast. In fact, because of this shallow contrast, my initial perception of the ffmpeg-decoded image is that it looks softer.
There is not that much extra detail hiding in the shadow areas of the image. AVCHD, I believe, being an implementation of H264, works strongly on a perceptive quality basis, and so does not capture these areas in any great detail, leading to this "mud" issue that is so frequently debated. Using the cinema picture preset seems to abate this problem somewhat, taking better advantage of the codec's available bit rate.
If I were to be cynical, I would actually argue that flattening the contrast makes AVCHD footage worse for colour grading. It brings out the noise in the shadows, not to mention any compression artefacts that are present (such as blocking and banding), and once you then crush the blacks and tweak other tonal areas during grading, these issues will be more prominent than if you had stuck with the original colour "profile" (for want of a better word).
Having said all that, though, I'd sooner use a decent decoder library like ffmpeg than rely on Quicktime's API, which I gather people are using, as they rewrap their .mts files to .mov files using Clipwrap?
I share your concern about the issue being prevalent with the GH series simply because people will continue to use Quicktime. Because Quicktime, with its inferior H264 decoding, has become popular in its use for transcoding H264 footage, born from Canon's use of the MOV wrapper with their DSLR range, there is a noticeable difference when that same footage is decoded with an API that makes fewer compromises for speed and stability. Windows 7, and possibly previous OS iterations with the correct updates, has strong support for AVCHD through its Directshow API, and the results during real time playback are much closer - possibly identical - to how the clips should look.
The issue is, of course, finding a user-friendly way of using superior decoding tools to transcode your footage. Not everyone finds batch files and command line applications so endearing! 5dtoRGB is certainly a step in the right direction, and would enjoy a boost in popularity with a Windows version. At the current time, if people want to achieve a good result it would certainly be no skin off the nose to give ffmpeg (or ffmbc) a try. It's well documented and far from daunting to use!
Apologies for the essay..
Last edited:
I wouldn't be surprised if the poor results of the Quicktime decoder turned out to be another symptom of the infamous Quicktime Gamma Bug:
http://provideocoalition.com/index.php/aadams/story/quicktime/
This bug afflicted Windows users of Adobe CS3 and CS4 as well, since Premiere used Quicktime to decode imported MOV files. With CS5, however, Adobe created a decoder plug-in architecture that bypasses Quicktime entirely. By default, CS5's bundled MainConcept H.264 codec is used to decode imported AVCHD files. If desired, you can override this by installing an external decoder plug-in for H.264 videos (or any other type of video format). But as far as I can tell, there are no ffmpeg or x264 plug-ins available to use for testing this feature in CS5.
Thomas Worth
Active member
This isn't a problem. Six 7200 RPM one terabyte disks in a RAID5 would easily do this.
And I agree, it is not a good idea to use rotating disks in RAID0 for anything other than testing.
Thomas Worth
Active member
Someone can correct me if I'm wrong, but as long as the MOV file is already compressed DNxHD, doesn't Avid simply re-wrap the QuickTime MOV into MXF upon import (fast import)? If so, the footage would be unaffected as I suspect Media Composer ignores all gamma flags in QuickTime files (as it should).
It's not complicated. Paying work just takes precedence.
roei z
Veteran
No, mts files don't get "fast import". They're transcoded
To DNxHD. that's why I asked about this.
I can choose between 701 or RGB import.
can you check this somehow?
Need me to test it and send it to you?
PM me if you want to, thanks
Ivo Yurich
Member
No, mts files don't get "fast import". They're transcoded
To DNxHD. that's why I asked about this.
I can choose between 701 or RGB import.
can you check this somehow?
Need me to test it and send it to you?
PM me if you want to, thanks
Did you try AMA? You can avoid transcoding all together...i suppose you need to keep copy of original folder structure
With Media Composer 5, AMA (Avid Media Access) you can work natively directly accessing MTS files (all footage is scaled to HD frame size). Plus, you have access to its full metadata and color management data too.
JamesRitson
Member
Unless there's an update I haven't tried yet, AVCHD files from the GH2 won't import into Avid full stop. AMA doesn't work either. I couldn't say for sure but I have a feeling it's because the GH2 uses a consumer model of AVCHD, whereas a camcorder like the AF100 (whose .mts files import with no issues) uses Panasonic's AVCCAM implementation. Even with AMA, Avid is much stricter about its media standards than other NLE systems.
If you want to use GH2 clips with Avid your best bet is to transcode them to DNxHD then fast import them. Once transcoded to DNxHD you would use REC.709 for the pixel mapping.
If someone's managed to import GH2 .mts clips directly into Avid without transcoding externally please let me know!
If you want to use GH2 clips with Avid your best bet is to transcode them to DNxHD then fast import them. Once transcoded to DNxHD you would use REC.709 for the pixel mapping.
If someone's managed to import GH2 .mts clips directly into Avid without transcoding externally please let me know!
roei z
Veteran
AMA is still not supported for AVCHD GH2 codec.
BUT- i just import the files (not a fast import. regular import) to my choice of DNxHD.
MAKE SURE you have the latest MC5 though. i had some trouble at first, than i put a 5.03.2 patch and it was perfect.
i would VERY MUCH like you to do some import experiments on color space so we could finally decide.
PM so we could share more details on this.
so, if i change things on this "image import" tab, is this relevant to the AVCHD import ?
i just followed an avid forum thread and they are still unsure of the best workflow in our case.
it seems like a good option is to open the files on CS5 or VEGAS, and then finding the best appropriate export,
so you could fast import or even use AMA on MC5.
any thoughts on that?
Last edited:
I'm on version 5.0.3. Your import settings look good. Compared to canon's h.264 files, the avchd files from my gh13 seems a lot smaller, and import directly into avid much quicker. Checked the footage on a sony hd crt in our symphony suite today and was blown away by the quality. You may be best just doing a simple import to DNxHD.
Not really enjoying the ama experience on h.264 stuff myself - sticky and seems to drop frames
Not really enjoying the ama experience on h.264 stuff myself - sticky and seems to drop frames