@_gl, as I said before, you can lower the quantizer setting of Cake to bring it up to the quality level of CBrandin. Cake deliberately uses a higher quantizer to be more efficient. Incidentally, Flow Motion now uses the same quantizer setting that I use in Cake.
Lpowell, again, the GH2 has a proper implementation of rec.709 studio levels. Your test is flawed.
@balazer I'm not sure what your point is. Regardless of whether it's "proper", the 7.5 IRE pedestal shift needs to be corrected before you can do a good color grade. For gamma adjustments to work properly, you need black to be at 0 IRE.
Also, the QP quantizer setting is only half of the quantization story. The other half is the contents of the Scaling Tables, which in Flow Motion v1.11 are customized to produce P and B-frames of comparable quality to I-frames.
@jshzr Splash Lite works very well and is free.
The pedestal correction you're talking about has everything to do with the fact that the software you're using doesn't comply to the rec.709 standard, and nothing to do with any fault or difference in the GH2. Other cameras produce rec.709 video just like the GH2 does, and would require the same correction using your software. Other software is compliant, and requires no special correction.
And pleeeease stop talking about IRE. IRE is a scale of voltages in analog video, completely meaningless for digital video.
@balazer - The disconnect here is that you're assuming that because the Rec. 709 standard is a widespread broadcast content delivery standard, it should therefore be universally applied back up the post-production chain. As a motion capture format for post-production, Rec. 709 is usable, but far from ideal, and requires corrective handling. The GH2 uses it because it's designed to be a consumer-oriented camera that produces a finished product, ready to view straight out of the camera.
While consumer-friendly editing apps may automatically do the pedestal correction for you under the hood, I wouldn't want a professional tool to attempt to conceal things like this. I want it to show me exactly what's in the video stream so I can decide how to handle it appropriately.
Thats not really true, most pro software can and does apply a correction, but it can also be turned off. In most cases software shifts (as in expands) the footage into sRGB (or linear space, or whatever it normally works in) or even 709 extended (as opposed to legal) for preview and for internal processing. Edit : Although the preview/internal processing will be different conversions I mean, the conversion for internal processing will probably be putting the information into something much wider than the sRGB preview. When you write out your processed footage, then you select what you're delivering to and let the software do the conversion on the whole. But of course, you can with good software force this to be done, delivered, in a specific way. There are differences software to software of course, for example Nuke assumes when it reads 709 footage that its 709 extended and not 709 legal, so you read it in set to 'raw' and then a grade node is needed to set the blackpoint/whitepoint (to 64/1023 and 940/1023).
The proper solution is to expand into the full RGB colourspace, which adjusting the pedestal isn't doing I believe. Thats just a clamp I think.
lpowell's suggestion seems to help, so I tried it. But it isn't required--I mean, you can grade your video anyway you want. The GH2 has a fraking green tint, so I correct it. Even if I measured it, and it was perfect, I would still correct it.
@Stray - "...for example Nuke assumes when it reads 709 footage that its 709 extended and not 709 legal..."
And exactly how is the NLE supposed to know that it is in fact reading "709 footage"? After all, there is no universal metadata tag that brands a video stream as "Rec. 709". Is each NLE then expected to maintain a complete database of codec file formats (including camera manufacturer-specific variants) and automatically and unerringly select the appropriate color-space correction for each video it imports?
Okay.. fair point, but yeah actually it is available information in some codecs. Yep, and some apps assume too (mostly wrongly). However, most apps allow you to interpet the footage, so that it gets it right. I don't know color finesse, or what you can tell it about interpreting the footage, but if you can't tell it its working with 709 legal (so it gets its blackpoint right) then I dunno what to say. I still say that a pedestal adjustment, while acceptable and obviously works, is the wrong approach to me.
@Stray - The reason I recommend doing a pedestal adjustment first is because it corrects the 7.5 IRE black-point offset of Rec. 709 without affecting any other aspect of the video's color balance. Shifting the entire waveform down to set black to 0 IRE simplifies all of the mathematical interactions of the various grading adjustments, and gives you a bit more headroom on top as well. After the pedestal correction, my next step is usually to adjust the gamma and/or gain, and these adjustments will not produce the intended effects if black is set to a dark shade of gray instead of at zero.
Cool, yeah, I didn't think about the headroom, of course. I just worried that as it wasn't being expanded/converted that the midtones wouldn't be as great (as would hold up) as they could be, but then I hadn't considered your gamma adjustment which would sort that out. The headroom will let you, so you're basically expanding your range out anyway. I get it now, thanks.
@LPowell regarding the recording time with different settings, since there are coarser quantizers on 24L, am I better off selecting settings that give me the amount of recording time on 24H that I need? I was hoping to have the length I need on 24L, with the ability to shoot 24H when recording time wasn't a factor.
Also, what is the best way to make a pedestal adjustment using Vegas Pro 10?
@Stray, the big problem is that nobody sticks to the standard. The Canon HV20 for example has superbrights (>235) which you wouldn't want clipped off by default. I bet some cameras output super-blacks that you also wouldn't want to loose. So saying that raw footage is 709 really doesn't mean anything reliable - I agree that it should, but there are way too many exceptions now, so it unfortunately makes sense to show you the full range and let you intelligently massage it.
For my money, studio levels, like interlacing, should just go away completely - not only are they stupid/totally unintuitive and error-prone, but you're also wasting precious level resolution on those (supposedly) unused super black/white regions, and introduce some banding when the range is expanded to full 8bit levels (which almost all displays are anyway).
That's the pain of backwards compatiblity...
Strange I loaded the last patch v1.11. Is it really 75mb in L mode cause when I open Ptool the modified bitrate is under the original one: 12500000 instead of 1600000 A I wrong? And my remaining time on my SD card seems so high in L mode: 2h34 (16gb sd card) How can I change this: One hour in L mode would be enough.
@bricolo in my tests I never got close to the remaining time it initially displayed. I think this has to do with the VBR, but other people on here could give you a more technical answer.
@LPowell Thanks for this discussion about pedestal. I have a question - maybe you can help. I open a raw GH2 MTS file in Premiere CS5.5, and display it on the waveform monitor. There are no blacks below 7.5 IRE, but that's because there's a little box that says 7.5 IRE and it's checked. If I uncheck the box, the blacks now reach down to zero. So my question is, which is correct? What is actually on the file?
@Ralph-B
"What is actually on the file"
The GH2 video is 16-235 (Y').
"After the pedestal correction, my next step is usually to adjust the gamma and/or gain, and these adjustments will not produce the intended effects if black is set to a dark shade of gray instead of at zero"
this is due to Premiere's design (which is actually a Photoshop-Style application for moving images). For instance in Avid curves, gamma, gain and you name it work as supposed to in a rec709 workflow (also as supposed to in a fullswing workflow, if you want to).
"Is each NLE then expected to maintain a complete database of codec file formats (including camera manufacturer-specific variants) and automatically and unerringly select the appropriate color-space correction for each video it imports?"
In my view this question is not relevant here. The "rec709"/broadcast levels and "RGB"/computer levels settings don't work like an "auto adjustment". Mostly they really only preserve levels or they expand (or compress) levels. For instance on Avid's import dialogue you can choose "rec709" and "computer RGB". Avid (as a professional tool based on broadcast standards) expects footage to be 16-235 Y'. Checking the "rec709" import option means: the original file is already at rec709 - do not modify levels on import. The "computer RGB" import option means: the original file is fullswing - push the blacks by 16 Y' levels and decrease the whites by 20 Y' levels on import. On export the "RGB" option means: the movie is studio swing - expand levels to fullswing on export. That's it. The settings do NOT analyse the file and adjust the blacks to an absolute value of 16 (or 0) Y' and the whites to an absolute value of 235 (or 255) Y'. The converter 5DtoRGB works exactly the same way (settings called "full range" and "broadcast range" here)... and many other tools also work exactly this way. So there is really no database of codecs or so needed. It's just about preserving or compressing/expanding levels by a determined value.
@royadkins @bricolo - I designed 24L mode (as well as FH and L) to serve two main uses: ETC zoom mode and 4GB file-spanning on Class 10 SD cards (95MB/sec UHS-1 cards will span in all modes). To support file-spanning at over 50Mbps, I used the GH2's "Trick Mode", which drives the file transport stream at a lower bitrate than the video stream. A minor side-effect of Trick Mode is that it confuses the camera's recording-time estimate and the time-duration listed for the file on the Window's desktop. NLE's however, detect the correct duration of the video stream and have no difficulty handling it correctly.
At 100Mbps, 24H mode will record for at least 5 minutes without needing to span files. If you need a longer single take than this, I recommend using either a 95MB/sec UHS-1 SD card in 24H mode, or a Class 10 SD card in 24L mode.
So you mean that even if the bitrate is 12500000 it is still a good quality? better than the stock settings. Sorry for my stupide question... Just trying to understand. And you mean it's not really 2h30 in L but 20 mm in H as it is said on the camera? Do you have an idea on how much time do I have on L with that 16 GB 45mb/s sandisk extreme pro?
@bricolo - the numbers displayed in PTool patches should not always be interpreted as actual bitrates. They are interdependent parameters that work in combination to produce the overall behavior of the encoder in various modes. In Flow Motion v1.11, 24L, FH, and L record at a max bitrate of around 50-60Mbps, but in dimly-lit situations bitrate can drop lower than this. At 50Mbps, you would get at least 40 minutes of recording time on a 16GB SD card.
Thanks a lot to take time to answer.
@towi "The GH2 video is 16-235 (Y')"
Yes, but my question concerns the 7.5 IRE checkbox in Premiere's waveform monitor. Do I assume this is just how the picture is going to be mapped in the waveform monitor, and has no bearing on the underlying data? If this is so, why does LPowell feel it's necessary to do a correction for 7.5 pedestal?
The check box just scales the monitor, it does not affect the render unless you make adjustments. Same thing in Sony Vegas Pro.
For those of us using Premiere, is setting the black input level in Color Corrector to 16 the same as a Pedestal adjustment, or should I set the pedestal in Luma or RGB corrector? (trying to use fewer filters) and what is the equivalent in Pedestal in the Luma/RGB setting (obviously not 16)? Or, simply, what are the ideal numbers to plug in using Premiere?
Hello all,
I do not know if this has been asked but is it possible to combine the flowmotion 1.11 with the "B-frame" in 720/50p? I read that 720p50 can be configured to record with B-frames with Encoder Setting 1 720p = 3
So is the combination possible and does it make sense at all?
Thanks for the answer.
It looks like you're new here. If you want to get involved, click one of these buttons!