|
|||||||||
|
Thread Tools | Search this Thread |
January 1st, 2008, 11:24 PM | #1 |
Trustee
Join Date: Mar 2003
Location: Virginia Beach, VA
Posts: 1,095
|
New Film Looks available
Hi,
We've uploaded a bunch of new film-emulation looks that we have created for the SI-2K. These .look files will allow the SI-2K to emulate some common negative and reversal film-stocks. Please note that this is not the same as emulating the native SI-2K color-space to a film-print stock like a 2383 or 2393 print (Vision Premiere print stock) that you would use on a film-out . . . this is actually an emulation of the colorimetry and gamma that one would get capturing on an original film-negative and/or reversal film-stock. You can get them in the Silicon Imaging Look Library. Thanks, Jason |
January 2nd, 2008, 05:35 PM | #2 |
Regular Crew
Join Date: Feb 2006
Location: Los Angeles
Posts: 78
|
Hey Jason
could you please explain again the difference between these looks and film out looks? in what instance should we use these new ones then? Thanks Lior |
January 2nd, 2008, 09:14 PM | #3 |
Trustee
Join Date: Mar 2003
Location: Virginia Beach, VA
Posts: 1,095
|
"Film-out" .look files are like the "Neutral_2383_Film.look". They basically say, "I have an SI-2K aligned to sRGB or ITU-R709 color-space, and I want to know what the camera will look like when it goes out on a piece of print film at the movie theater."
It's basically seeing what the camera's default color-space will look like when you print it and want to see it at the theater. Rather than a film-print, it could also be a DCI-based digital projector, etc. The main point is that the color-space of the SI-2K will be it's default aligned color-space such as a sRGB or ITU-R709 color-space. These new looks take the SI-2K and basically say, "I want the SI-2K to mimic an image I would have captured if I shot on XYZ film-stock." There is a fundamental difference here. We're not talking about the camera's default color-space on a final release film-print at the theater, we're talking about wanting the scene to look like it was really shot with a certain shooting film-stock (In the case of Kodak these "shooting film stocks" would be the 52xx series versus a release-print stock that is in the 23xx series and in a normal film-chain would come after interpositive and internegative dupes). So with these new .looks, we're maping the color-space of the camera to the colorimetry of a shooting film-stock rather than a default color-space like ITU-R709. You will notice there is no 3x3 matrix entry for these .look files . . . that is because this is no longer a pure linear transformation like going from the native XYZ space of the camera to sRGB or ITU-R709. It is a non-linear transformation designed to have the camera mimic the non-linear colorimetry of color negative and reversal film-stocks. Thanks, Jason |
January 3rd, 2008, 08:03 AM | #4 |
Inner Circle
Join Date: Dec 2003
Location: PERTH. W.A. AUSTRALIA.
Posts: 4,477
|
Jason.
You won't forget the Agfa neg stocks and workprint stocks of the seventies-early eighties now will you? |
January 3rd, 2008, 01:58 PM | #5 |
Major Player
Join Date: May 2003
Location: Akron, Ohio
Posts: 496
|
This is awesome Jason! Thank you for listening to customer feedback (I asked about the Provia 100F look awhile back). I’m curious how you where able to ‘map’ the film stock. Did you do it by eye with a calibrated monitor, or did you use some type of calibration.
Forgot to add- My favorite stocks of all time- Kodachrome 25 or Ektachrome EPP 100 Natural, yet gorgeous color. Last edited by John DeLuca; January 3rd, 2008 at 02:11 PM. Reason: Favorite stocks added |
January 3rd, 2008, 01:58 PM | #6 | |
Trustee
Join Date: Mar 2003
Location: Virginia Beach, VA
Posts: 1,095
|
If we can get our hands on them, then we can do them . . . :)
At this point in time, we chose some based off Fuji Provia and Kodak 250D. The Fuji Provia-based versions are not exactly Provia, but are like them a bit . . . that's why I didn't label them as Provia. The Kodak 250D versions are a statistical "best-fit" with some hand-tweaking to balance between accuracy and noise, but overall are pretty accurate compared to scanned Kodak 250D film-stock (so i.e., matching against Kodak 250D may not be perfect if you use a different scanner or telecine compared to what we used for our source data). Quote:
Thanks, Jason |
|
January 4th, 2008, 10:45 AM | #7 |
Inner Circle
Join Date: Dec 2003
Location: PERTH. W.A. AUSTRALIA.
Posts: 4,477
|
Jason.
Does that mean that if you were given the colour curves from the particular film stock you could recreate the look from those? It is simply amazing you can do all this in a video camera. |
January 4th, 2008, 10:53 AM | #8 |
Trustee
Join Date: Mar 2003
Location: Virginia Beach, VA
Posts: 1,095
|
BTW, these types of transformations can't be done with curves (per-channel 1D LUTs), they can only effectively be done through 3D gamut mapping.
But yes, this is pretty cool technology, and one of the great powers that having a real-time 64-point cube 3D LUT system running in real-time on the camera can provide. |
January 8th, 2008, 11:01 PM | #9 |
Major Player
Join Date: Apr 2006
Location: paris
Posts: 289
|
Congratulations. The competition should follow your finest example.
|
January 9th, 2008, 05:24 AM | #10 |
Trustee
Join Date: Nov 2005
Location: Sydney Australia
Posts: 1,570
|
I tried loading some of these Looks a few days ago, wasn't paying much attention to which ones sorry. The thing I did notice was they alter the exposure and readout on the meters, so presumably the false colour and histogram readings are taken after the Look is applied.
This raises the question of how should we actually be using the metering in the camera and what significance can we attach if the Look we'd like to use indicates we're clipping and the default Look shows we're not. I think I can understand the logic of this but I could use some clarification to be certain. |
January 9th, 2008, 10:19 PM | #11 |
Trustee
Join Date: Mar 2003
Location: Virginia Beach, VA
Posts: 1,095
|
Both Histogram and False color are taken pre-look. This was done so that no matter what the .look is, you'll know if you're screwing yourself during a shoot (i.e., if the .look is clipping but you're not clipping the RAW data, then no big deal . . . but you don't want to have a .look that doesn't show clipping, or somehow changes the contrast curve and doesn't give you an adequate indication of what the "true" RAW exposure is at, then you can do silly things like underexpose too far, massively over-expose, etc.).
The ONLY meters that should change relative with the .look file are those inside the IRIDAS control panel. This is so you can see the effects on the visual exposure of your changes when you're making a new .look file. Now one thing to keep in mind is that a different .look file will change how you expose to get a "proper" looking image on the screen . . . i.e., Default.look uses a standard REC709 curve, and that crushes the shadows quite a bit and maximizes the effective SNR (i.e., middle-grey is "pushed higher" in the exposure scale). This forces one to expose a stop or two more open than another .look that would open up the shadows, and push the middle-grey point down in the exposure scale, so with a different .look you would be stopping down a little bit to get the same "visual" exposure. That is not a bug, that is the nature of different .look files having different digital gain and contrast mapping features. So for instance, if you set the meters, and start flipping through .looks (i.e., set the main screen to false color meter mode, and just flip through the .looks), you should not see any switch in the false color meter. The same with the histogram. Now with the meters off, you will *visually* see a difference, but the underlying RAW data is all the same. The reason I'm adding this note is because I can't reproduce a meter variation in the main GUI with the switching of .looks, so it seems to be operating normally in the latest builds (this build is not ready for release though). If you want some more information on what I'm talking about with contrast mapping using a .look file, you can view my power-point presentation on exposing for RAW from NAB07 here. If your meters are really changing as you swap though .looks (they shouldn't have), that would be a bug. Please let me know which version of software you are using if you think you've encountered a bug. Thanks, Jason |
January 10th, 2008, 08:19 AM | #12 |
Trustee
Join Date: Nov 2005
Location: Sydney Australia
Posts: 1,570
|
Thanks for the detailed reply. I'll check this again just to be 100% certain.
|
January 11th, 2008, 03:56 AM | #13 |
Trustee
Join Date: Nov 2005
Location: Sydney Australia
Posts: 1,570
|
Went back and checked again today. All is good, the meters aren't changing. What threw me was as you said was the shift in visual exposure, I think somehow because the colors of the subject were very close to the colors of the false color metering I got confussed.
However your detailed post has explained a lot for me, the pieces are coming together. |
January 11th, 2008, 11:11 AM | #14 |
Trustee
Join Date: Mar 2003
Location: Virginia Beach, VA
Posts: 1,095
|
Great, thanks for the feedback though, this was a very good question (and I should probably post it in the FAQ).
|
January 11th, 2008, 06:31 PM | #15 |
Trustee
Join Date: Nov 2005
Location: Sydney Australia
Posts: 1,570
|
I had one thought later. Would there be any advantage to having the DVR meters switchable pre and post Look?
One other issue I struck yesterday. Please, please, can we have the DVR software accept keyboard as well as touch screen input. Trying to enter project names by clicking a spot on the screen using a dodgy laptop touchpad drove me nuts. Partially my own fault, I'd left my USB mouse behind. For some reason on the Dell 6300 the DVR software seems way more sensitive to the touchpad then Windows. I had to barely touch the pad or as I moved the pointer around every time it passed over a charcter on the onscreen keyboard DVR would input the character. Not a big thing but on a live shoot I can see this creating a lot of frustration. |
| ||||||
|
|