[Andrew Kimery]

[Robin S. Kurz] “DV is DV is DV.”

DV all adheres to DV specs, but the quality of the image that ends up on tape depends, in part, on the quality of the encoding done in camera. And once you get out of camera there are different flavors of the DV codec depending on what you are doing in post. Going back to an old post from Adam Wilt, “Some DV codecs, like Avid’s, filter or smooth the chroma on decompression (details here). Others, like Matrox’s, let you turn chroma filtering and interpolation on or off as you see fit. But the Apple codec used in Final Cut Pro doesn’t give you any choice: it’s steppy edges, all the way. ”

https://www.adamwilt.com/SteppyEdges.html

[Andrew Kimery]

[Robin S. Kurz] “Wait… so you’re saying the quality of the 1’s and 0’s being moved was higher than the ones coming through Firewire? :-D”

Can’t speak to what Mitch saw but Graeme Nattress (plugin maker, worked at RED developing their codecs, etc.,) wrote an extensive article a long time ago about chroma sampling, and did find benefits to capturing DV via SDI into an uncompressed codec (this predates ProRes). It obviously can’t add any more image detail than what exists on the DV tape, but it seemingly can help smooth out some of the jaggies in the chroma caused by the compression.

https://www.nattress.com/Chroma_Investigation/chromasampling.htm

[Andrew Kimery]

[Walter Soyka] ” You’re still trapped in your NLE, so workflow and grade management is a headache.”

Workflow and grade management isn’t sexy, but after grading nearly full time for a couple of year it’s possibly what I miss the most when I’m grading inside of an NLE.

[Andrew Kimery]

[Bill Davis] “It Addresses best practices when you ONLY have access to a VHS or DV dub. Nothing better. If that is your inescapable starting point – and you need to use that AS your master source – you gain little to nothing if you capture it as ProRes 444 for your mezzanine codec – rather than just using, say, h-264 before you bring it into your NLE. Adam explained that AFTER the hit to analog VHS or DV25 – digital encoding has little preservable data left that warrants using a beefier codec for storage.”

I totally, 100% understand what you are saying. My remaining question doesn’t have to do with the amount of detail left in the image though. My question has to do with the amount of the image that changes from one frame to the next because big changes from one frame to the next (say flash bulbs at an event or a roaring fire in the foreground) can break inter-frame codecs. Those codecs get their small file size by basically assuming that not a whole lot in the image changes from one frame to the next.

Intra-frame codecs (like ProRes) compress each frame as a stand alone image (for lack of a better term), but inter-frame codecs will basically (and this is a really superficial description) compress a whole frame, then interpolate the next 12-15 frames (typically whatever makes up a 1/2 second), then compress another whole frame, then interpolate the next 12-15, etc.,. If a lot of change happens during those interpolated frames (such as a camera flash) there is a good chance the image will turn into macro blocks because the codec doesn’t retain enough image information to recreate such a big change. You can crank up the bit rate to help alleviate those problems, or add more key frames (which reduces the number of interpolated frames) but then the file size goes up which starts eliminating the smaller file size advantage of going with an inter-frame codec to begin with.

I’m also not saying to go to ProRes 4444 (I actually specifically mentioned thinking even HQ was overkill), but just going with a quality, inter-frame codec (which is why I gave ProRes LT as an example) or conducting tests on potential problems areas in the footage to see if they will break whatever ‘flavor’ of H.264 one is using.

So, in a nutshell, inter-frame codecs have potential problems that intra-frame codecs do not so do some tests with the specific inter-frame codec you intend to use (settings, encoder, etc.,) before converting a ton of footage and getting neck deep into the edit.

[Andrew Kimery]

I don’t know if it’s true or not but merged clips used to cause big workflow headaches in PPro. I work with multicam clips even if the shoot only had one camera plus second system sound.

[Andrew Kimery]

Thanks Bill.

Lower res formats inherently lacking fine detail (which is hard on both inter and intra frame compression schemes) makes sense. My concern was more about the changes between frames as opposed to the amount of detail in a single frame (so temporal vs spacial I guess). Things like camera flashes going off or a campfire fire dancing in the foreground are much tougher on inter-frame codecs than intra-frame codecs since the space-saving aspect of inter-frame codecs is based on the assumption that not much is changing on a frame-by-frame basis. I’ve seen red carpet footage from DV cameras that doesn’t show any artifacting beside clipping when the flashbulbs go off but red carpets shot with Sony’s EX1 (which uses the XDCAM EX codec) can turn into a pixelated mess for a frame or two if a flash lights up what the camera is pointed out.

Obviously the bit rate and the encoder have to be taken into consider as well as the more bits you have, and the better encoder you are using, the more robust the process is to handle rapidly changing images.

So, ultimately I guess my advice would be to prescreen footage and test sections that look like they might be problematic for H.264. Or if one has the storage space just capture everything in a codec like regular ProRes 422 or it’s DNx counterpart. It might just be paranoia, but I do share Michael Gissing’s concern of if you are essentially making new master copies of the footage why not err on the side of caution?

FWIW I grabbed some high res fire stock footage I have (1080p image sequence), turned it into DV using Adobe Media Encoder and then converted that DV file into H.264 using AME’s “Vimeo 1080p” preset and the results were surprisingly good. It was soft, obviously coming from DV, but it looked fine all things considered (no obvious or distractingly bad macro blocking from my quick look at it).

[Andrew Kimery]

[Eric Santiago] “That one I know about but doesn’t work after the fact.
“

I came across this web page last night talking about the Markers and XMP and one of the commenters might have a solution (assuming you haven’t tried it already).

“Tw 2015/03/22 at 17:45
.
.
.
We have a similar problem – Started logging with markers writing to XMP before we realized the markers wouldn’t be in the Project file. Best workaround we have so far is this:
1.) Uncheck Write clip markers to XMP in Preferences >> Media
2.) open markers panel with the clip
3.) double click on each marker and then simply make an edit to marker, e.g., change color of marker

This has the effect of moving marker data from XMP to the project file.

Yes it’s very tedious, but at least it doesn’t require a lot of brain work.”

https://www.creativeimpatience.com/premiere-pro-clip-markers-solved/

[Andrew Kimery]

[Bill Davis] “Adam is a friend and I reached out to him for his thoughts on this specific circumstance.

When he gets back to me, I’ll post his thoughts.
“

Cool, I look forward to hearing his 2 cents.

[Andrew Kimery]

[Bill Davis] “What I want to know is if the conventional wisdom that suggests that when you start with a relatively poor resolution analog master, applying the same codec if it makes things AS MUCH worse as the drop from high rez to compressed? And how much worse do things get. “

How much depends on the the footage. An INTV with a guy in his living room will most likely be less taxing from a compression standpoint than footage of a campfire (the constantly changing flames) or a red carpet event (flashbulbs going off). The more that changes between frames the harder it is for compression to keep from macro blocking. Footage of running water, leafy trees or tall grass in swaying in the breeze, color gradients (like a wall lit unevenly by a diffuse light source), etc.,. are also good things to look at if you are stress testing a codec.

On a related note, a long time ago I remember reading a very detailed, objective comparison between BetaSP and DV done by Adam Wilt and, long story short, DV had higher image quality than BetaSP (all other things being equal of course). One thing Adam noticed though is that when both formats reached their limits in terms of resolving image detail, the BSP might be perceived as looking better (in a subjective sense) because the fine detail starts going soft as opposed to the blockiness you see with digital compression.

So even when coming from relatively low-res analog sources you might have to aim higher than you think you need to during the digital conversion to avoid the digital artifacting which people find more bothersome than it’s analog equivalent. I always try to stay with an intra-frame (ProRes, Cineform, DNxHD, etc.,) since the compression happens on a frame-by-frame basis and is less likely to fall apart as opposed to an inter-frame codec where groups of 12 to 15 frames are all compressed together. I worked at a place once where, due to the extremely high volume of footage they created on a weekly basis, ProRes LT was the default capture codec (prior to that it was DVCPro HD) and our image quality always passed QC for broadcast. I’m not one to automatically suggest taking VHS to ProRes HQ because, at some point, all you are doing is pouring a 12oz can of Coke into a 2 liter bottle.

-Andrew

[Andrew Kimery]

[Bill Davis] “Send lawyers, guns, money and creative writers for our new terms of service. The s*** has hit the fan.
; )”

haha!