First Attempt at 35mm filmout with the HD-100

[Marty Baggen]

Steve... I think we may be asking the same question.

I suppose what I am curious about is how much "filmlook" does printing to film do?

If it is an optical process, it would seem to offer just as many opportunities for tweaking and adjusting as film does in the acquisition stage, thereby rendering the effort to create a filmlook with the video camera somewhat moot.

[Steve Mullen]

Not the same question.
When you print to film you are get
ting the look of the film you are printing too. For example, the film's grain.

Of course in post you can manipulate the look in any way you want.

So there are three points, at least, where you can alter the look. There's no single place where you get your look. You can even add grain in post.

[Marty Baggen]

Got it, I see what you're getting at.

So I guess my question still remains in light of your scenario, why add grain (or any other characteristic of your output film stock) to match the film you are printing to if the film already has the grain you are attempting to simulate?

It seem silly to go through the motions of film emulation for a filmout, if you get the real thing by doing so.

My question is irrelevant if the print-to-film process does NOT introduce all of those qualities that video tries to fake, such as the gamma curves, the grain, etc.

[Steve Mullen]

You've got it.

When shooting FOR film you do not do anything to create a FILM LOOK.

Of course, that doesn't mean you can't manipulate the image in all sorts of creative ways.

And, one could augment the look you'll get going to film. That's because often the stock used for printing is one that imprts very little look of its own. So you could, in post, create the look of a particular stock.

Or, you could set that look up in the camcorder.

[Andrew Young]

Originally Posted by Ben Syverson

I haven't played with an HD-100 yet, but if it does indeed offer a truly linear image with no gamma (it'll look extremely dark, with electric highlights), it would be wise to avoid it at all costs.

I was witholding judgement until shooting some tests, but I have to say that that's exactly what FILMOUT mode looks like to me and based on preliminary observations (hooking the camera up to a scope) I would agree with your assessment. It looks as though information is being lost rather than maximized. If that is the case, I'm not sure what the point is of this option. It appears to be not at all the same thing as Film Rec Mode in the Varicam. If anyone actually gets useful results from this setting I'd love to hear about it.

[Steve Mullen]

Originally Posted by Andrew Young

I was witholding judgement until shooting some tests, but I have to say that that's exactly what FILMOUT mode looks like to me and based on preliminary observations (hooking the camera up to a scope) I would agree with your assessment. It looks as though information is being lost rather than maximized.

Here's a quote from Panasonic found in my HDV Production Guide:

"Cine Gamma and Tele Gamma: Cine Gamma was developed on the assumption that the final content is released on 35mm film. We developed a gamma curve for the video system to closely match that of 35mm film. In addition, with the cooperation of a famous Japanese cinematographer and Toei Chemical Co, Ltd., we established a Look-Up-Table (LUT) to return the captured video signal back to 35mm film via Arri Laser Recorder. As a result, we were able to realize the appearance of an image almost equal to 35mm film recording.

Tele Gamma was developed on the assumption that the final content is viewed on a television monitor, for example, a drama, or documentary. It creates the look of film on video."

I'm not sure why FILMOUT doesn't look like Cine Gamma.

I note that in Tim's FilmOut he has:

MASTER BLK NORMAL
CINELIKE OFF
BLACK STRETCH3 <-----------------This realy alters the mid-range, I prefer to go no more than BLACK STRETCH2 which only alters the darker tones
COLOR MATRIX STANDARD <--- This could be set to CINE (whatever Cine Matrix is)
KNEE MANUAL
KNEE LEVEL 85%
GAMMA FILMOUT
LEVEL MAX <------------------------- I'm not sure what LEVEL Tim is talking about since in FILMOUT you can't adjust gamma level??????????????????????????????
WHITE CLIP 108%
DETAIL <----------------------------- I'm not sure what he uses.


Perhaps adjusting some of these would also make a difference:

COLOR GAIN NORMAL
R GAIN NORMAL
R ROT. NORMAL
G GAIN NORMAL
G ROT. NORMAL
B GAIN NORMAL
B ROT. NORMAL

In short, I'm not sure what Panasonic's adjustments are that you call FILM REC.

[Ben Syverson]

Originally Posted by Nate Weaver

Something implied by your page is that latitude in current cameras is at least limited as much by A/D bitdepth as the properties of the CCD itself...Just how much latitude is being thrown away by A/D compromises?

Well, it depends greatly on how much information the sensor itself is capturing; you can have a 48bit A/D, but if the signal-to-noise ratio of the CCD isn't very good, you'll basically have an extremely accurate representation of a noisy image. :)

I also assume complete linearity in my page, but most sensors are not entirely linear; they have their own response curves, just like a film stock might. The A/D may complicate that; I don't know.

Furthermore, from what I understand, CMOS chips don't need separate A/D at all; they simply use per-pixel transistors to read the photosites directly. So with CMOS, you can truly say the sensor itself has a bitdepth of X, because the chip itself has a max bitdepth. Even Juan's magic won't squeeze more information out. :)

The area where the most information is lost is presumably not the A/D, but the image processing that takes raw data off the sensor or A/D and outputs a JPEG, an MPEG2 stream, or whatever. The deepest shadows are usually chopped off due to noise, the brightest highlights are clipped, the image gets a gamma of ~2.2, the brightest highlights are clipped again when the white balance is applied, the image is de-Bayer'd (which can be a lossless process, but is usually lossy in-camera), the image is sharpened, and then compressed. There are also countless "image enhancements" made during this process, which give an image "that Canon look" or "that Sony feel."

The end result is that latitude is not the only thing you've lost. ;) Unfortunately, it's not very convenient to record raw data all the time. Some cameras have extremely good engines that really maximize the data coming off of the sensor. From what I've seen, the HD100 is generating nice, fairly flat, clean images. With some care, they should be perfectly usable for filmouts, compositing and chroma key.

Actually, not to plug my own stuff, but if you are going to key DV or HDV, you should give dvmatte a try. We've shot a lot of HDV, and with my dvmatte blast, we've been getting really good results from HDV -- oh, and it renders in realtime. :) I love GPUs.

[Ben Syverson]

Originally Posted by Stephen L. Noe

The BMP (above) is an uncompressed frame right out of the shot footage. Can you download it and give an assessment?

Honestly, it's encouraging to see this footage, because it shows me what the HD100 is capable of. I mean, look at that blue channel! Beautiful!

But look at that red channel...

Blown out. In my opinion, the image is overexposed by around a stop--possibly more. I think the image fidelity is holding up enough after the MPEG2 compression that you could increase the exposure in post with no problems (I do it with DV that's nowhere near this clean)... The image also has more sharpening than I'd be comfortable with, but that may be a personal issue.

But again, nice to see the level of quality that the HD100 can produce!

[Jiri Bakala]

Originally Posted by Ben Syverson

Actually, not to plug my own stuff, but if you are going to key DV or HDV, you should give dvmatte a try. We've shot a lot of HDV, and with my dvmatte blast, we've been getting really good results from HDV -- oh, and it renders in realtime. :) I love GPUs.

Are you going to make it available as a plugin for Avid? (AVX)

[Tim Dashwood]

Originally Posted by Steve Mullen

I note that in Tim's FilmOut he has:

MASTER BLK NORMAL
CINELIKE OFF
BLACK STRETCH3 <-----------------This realy alters the mid-range, I prefer to go no more than BLACK STRETCH2 which only alters the darker tones

I determined these settings by attempting to "straighten" the curve to be as linear as possible. No one black stretch setting alters the mids any more than another. They only seem to affect 0-25IRE range - no matter which setting.
When you put the default FILMOUT gamma on the scope without any other adjustments, you can see that it is nowhere near perfectly linear.

You can check out my scope outputs here:
http://homepage.mac.com/timdashwood/...X-Sorenson.mov

Originally Posted by Steve Mullen

COLOR MATRIX STANDARD <--- This could be set to CINE (whatever Cine Matrix is)

I logic here was that the cine matrix attempts to emulate the more muted tones of film. I figured it would make more sense to capture a bit more colour information so there is more to work with in traditional lab colour timing.
Interestingly, when you compare the two matrixes, you can see that the standard matrix increases gain on the skin tone line. This says to me that STANDARD is more like a Kodak stock, and "CINE MATRIX" with some colour gain is closer to a FUJI stock.

Originally Posted by Steve Mullen

KNEE MANUAL
KNEE LEVEL 85%
GAMMA FILMOUT
LEVEL MAX <------------------------- I'm not sure what LEVEL Tim is talking about since in FILMOUT you can't adjust gamma level??????????????????????????????

That is a typo - and irrelevant anyway. However the last setting stays - even though it is unselectable. I hope that it actually has no effect.

Originally Posted by Steve Mullen

WHITE CLIP 108%
DETAIL <----------------------------- I'm not sure what he uses.

I'm still determining the effects of the detail setting. In my resolution tests, negative detail definitely decreased resolution (bad), but detail settings over +3 didn't seem to increase detail to any degree. My biggest question is whether this "detail" setting is actually a traditional "sharpness" setting - in which case leaving it at 0 or slightly negative would be ideal.

Originally Posted by Steve Mullen

Perhaps adjusting some of these would also make a difference:

COLOR GAIN NORMAL
R GAIN NORMAL
R ROT. NORMAL
G GAIN NORMAL
G ROT. NORMAL
B GAIN NORMAL
B ROT. NORMAL

Of course, but would require testing.

[Stephen L. Noe]

Originally Posted by Ben Syverson

Honestly, it's encouraging to see this footage, because it shows me what the HD100 is capable of. I mean, look at that blue channel! Beautiful!

But look at that red channel...

Blown out. In my opinion, the image is overexposed by around a stop--possibly more. I think the image fidelity is holding up enough after the MPEG2 compression that you could increase the exposure in post with no problems (I do it with DV that's nowhere near this clean)... The image also has more sharpening than I'd be comfortable with, but that may be a personal issue.

But again, nice to see the level of quality that the HD100 can produce!

I've adjusted the image to where I think it should be in the digital realm and separated the channels. Red is brought back in line as well as luminance.

Click here for adjustments screenshot

These adjustments resulted in this frame: Click here for single frame after adjustment

My questions to you are: Should I adjust the video to my satisfaction in the video realm in order to send to film print (using the frame above compaired to the original as an example)?

The stop too light plays in my favor compared to a stop too dark in the video realm. Does that hold true in the video to film trasfer realm as well?

I agree to the fact of JVC quality in their compression. It's been a joy to work with images that don't fall apart when you adjust them (particularly the blacks).

[Ben Syverson]

Originally Posted by Stephen L. Noe

I've adjusted the image to where I think it should be in the digital realm and separated the channels. Red is brought back in line as well as luminance.

This definitely looks subjectively better to my eye, but the problem is that since the red channel was blown out, adjusting the exposure in that channel down will just fill in those highlights with gray, rather than maintaining that detail. In this particular shot, it doesn't seem to be a very big deal, and it's perfectly usable. However, if we were looking at a face, I have a feeling it would be more apparent.

In terms of what you should do in order to send your stuff to filmout, that will depend greatly on how and where you're doing the filmout. Some houses will probably want your stuff adjusted to "look good" as video, and then they'll apply their special sauce to get it to look the same on film. Personally, if I were doing a filmout, I'd find a place that would take Cineon/DPX files, and do all my color in a floating point environment. That way, you don't lose any overbrights that you might gain by shooting under and pushing the exposure in post.

For a lot of good information about working in Linear Floating Point, check out the blog of Stu Maschwitz from The Orphanage, prolost. The eye-opener is his eLin demo movie, "LA Day," which you can see here.

[Ben Syverson]

Originally Posted by Jiri Bakala

Are you going to make it available as a plugin for Avid? (AVX)

I only get this question about once a year, which means "probably not." :(

FCP will likely support Motion plugins in the next version, but Avid is sadly a long way from supporting GPU-native plugins...

[Stephen L. Noe]

Originally Posted by Ben Syverson

... adjusting the exposure in that channel down will just fill in those highlights with gray, rather than maintaining that detail.

I totally agree. In this case, I can make the scene pop without touching saturation just by adjusting the red channel and luminance. I didn't want to do any color correction, merely because I'm not sure how the transfer affects the colors and whether there is a color shift. If there is a color shift then in what direction? Lighter overall or darker?

In the case of skin tones, you can not get away with adjusting individual channels most of the time without throwing skin to an unusual color. In the case of the supplied frame, I'd normally pull the blacks down and leave the gamma alone. In a full sun scenrio the ND filters mounted on the camera are not strong enough and added ND filters are required (in my estimation) to get the image to land where I think it should be, in camera. In the end, we'll see how it all pans out.

[Steve Mullen]

Tim Dashwood -- "No one black stretch setting alters the mids any more than another. They only seem to affect 0-25IRE range - no matter which setting."

That's interesting because I see a definite visible change and a histogram also shows that #3 definitely alters up to about 50IRE whereas #1 and #2 only alter the under 25 IRE level. Now using #3 may make the FILMOUT more linear -- but I was cautioning those who are using Black Stretch for any other reason that I find #3 is too much for my tastes.


"Interestingly, when you compare the two matrixes, you can see that the standard matrix increases gain on the skin tone line. This says to me that STANDARD is more like a Kodak stock, and "CINE MATRIX" with some colour gain is closer to a FUJI stock."

I asked JVC what Color Matrix OFF was. I assumed that STANDARD was an adjustable Video while OFF was a non-adjustable Video. Their response was that OFF is the 709 standard. Which rasied the question -- did that mean that adjusting the Color Matrices could lead to non-709 color? I don't think so, but maybe I'm wrong. Or, maybe they are.

What I can't understand is since it's an HD camcorder how STANDARD could not be 709. And, if it's not, then what does STANDARD mean? It's OK that it looks like Kodak stock, but why did JVC call it STANDARD? And, why label 709 as OFF.

I think -- JVC means that OFF is the camera's built-in JVC (and/or Panasonic) HD colorimetry. Then there is CINELIKE which is simply a muted version of this (OFF). How do you find CINELIKE compared to OFF?

Then, there is STANDARD, which I take to be more like JVC's SD camcorders, i.e., JVC's "standard" SD colorimetry. How do you find STANDARD compared to OFF?


"That is a typo - and irrelevant anyway. However the last setting stays -- even though it is unselectable. I hope that it actually has no effect."

Wow -- this could cause a huge problem. Are you saying you choose CINELIKE and set HIGH and then choose FILMOUT? And, you would get a different result if the last gamma you chose was STANDARD and/or LOW????


"I'm still determining the effects of the detail setting. In my resolution tests, negative detail definitely decreased resolution (bad), but detail settings over +3 didn't seem to increase detail to any degree. My biggest question is whether this "detail" setting is actually a traditional "sharpness" setting - in which case leaving it at 0 or slightly negative would be ideal."

I think they key is the H/V balance and the Contour setting -- not DETAIL. I would assumed Film to be equal H/V ( and since the camera has Square pixels and thus near equal H and V resolution), but FINE (hi frequency) contour for film.