Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

FilmConvert's IPP2 profile looks terrible - what an I missing?

Kevin Marshall

Kevin Marshall

Well-known member
Joined
Jul 17, 2009
Messages
429
Reaction score
0
Points
16
Age
37
Location
Balitmore, MD
Anybody using FilmConvert with their IPP2 preset? I've been testing it out, and while the DC2/RLF profiles are a bit dark (except the Epic-W one which is brighter for some reason), changing the footage to RWG/Log3g10 and FilmConvert's IPP2 profile makes the image vastly under exposed (roughly 1 1/3 stop), and heavily desaturated. I need to either boost the ISO (i.e. from 800 to 2000), or up the exposure in FilmConvert to around 16 to get the base exposure the same as in-camera.

Does anyone know why there's such a disparity between the FilmConvert options?

EDIT: Here's a quick comparison:
hzx6E0X.jpg
 
Last edited:
FilmConvert is hard to trust, if you test all the different color space/gamma combos with their respective FilmConvert profiles, even while even staying on the same camera, you'll notice every single one looks different. So it's hard to say what's what.

I would imagine there's a 1 stop discrepancy thrown in there somewhere from them either matching old RED profiles to Alexa brightness, or compensating for the IPP2 ISO Calibration 2, or something like that.
 
Yeah, I wondered if the recalibration played some factor in the difference between the Weapon Helium preset (darker lime the Dragon) and the Epic-W Helium preset (brighter as seen). But no clue why the IPP2 preset looks so bad. It actually looks more correct with the RAW settings on Rec.709 instead of RWG. It's all just very odd.
 
Yeah - I think FilmConvert just has a very aggressive toe with IPP2 for some reason. Here's another example (pulling from RED's samples). Using the RLF preset is a little under (but fine), whereas the IPP2 preset is a further 1 stop below the RLF preset:

FjkmAVp.jpg


Ultimately it's all workable - I can just use the RLF preset, or add a push+sat to the IPP2 preset. I just don't know why they don't match to begin with - the various RLF presets match pretty well across sensors (except the Epic-W one). I want to mess around with creating camera LUTs, but I was hoping RWG/3g10 would've been less fiddly on FilmConvert's part.
 
Last edited:
3g10 maps the mid gray lower than Redlogfilm, so if the filmconvert transforms are not compensating for that, then the image will be quite a bit darker.
 
My advice would be not to use this and instead color-correct the image in a real color-correction program. Every lighting situation is different, every director and DP want a different kind of image, so you need something with more variability and more range than just a set of presets. Having the ability to change brightness in different parts of the image at the same time is important, as is having the ability to gradually make an exposure change when the actor walks into light or into shadow. You can't do this with a program like this.

I don't dispute there are people who need fast and simple tools to just get out a very quick image, but I think you're better off just turning the knobs yourself and doing it by hand.
 
Marc Wielage said:
My advice would be not to use this and instead color-correct the image in a real color-correction program. Every lighting situation is different, every director and DP want a different kind of image, so you need something with more variability and more range than just a set of presets. Having the ability to change brightness in different parts of the image at the same time is important, as is having the ability to gradually make an exposure change when the actor walks into light or into shadow. You can't do this with a program like this.

I don't dispute there are people who need fast and simple tools to just get out a very quick image, but I think you're better off just turning the knobs yourself and doing it by hand.
Click to expand...

Absolutely. I've been using Resolve for about five years now, and Apple Color before that when I was in school. I know a few people that are into the film emulation LUTs like Impulz, but I avoid them as I feel like I'm working against the LUTs to get a sensible image back, adding unnecessary time and complexity. The extra controls available in FilmConvert as an OFX plugin I feel are more helpful in essentially customizing the LUT - specifically getting away from "presets", and making the rest of grading more like normal.

I also wanted the ability to export those into a LUT for camera/monitor usage (e.g. creating a show LUT), and wanted to avoid essentially concatenating LUTs, so again making those adjustments within the FilmConvert framework before exporting would be simpler and less prone to artifacts (based on my understanding).

All this to say that I just wanted to make sure the problem was on their end, and not a glitch or user error - like if their IPP2 profile was expecting something other than RWG/3g10 as input. But it looks like their IPP2 transform is worse than their legacy one, so I'll either avoid it or work within its constraints.
 
In case anyone else cares about this - I think I've figured it out. Since the profile is labeled IPP2, I thought it might be meant to actually work in IPP2 as a Creative LUT, i.e. with no actual output transform. So I lowered the Curve setting to zero, leaving a Log3g10 image, and put one of the IPP2 LUTs in the next node. The curve looked fine, but the image was oversaturated, so I dropped the saturation control from 100 to aruond 40-50 depending on the selected stock. I personally felt it consistently pushed the image too magenta, so I put a bit of green either in the RAW tint or the Filmcovert Midtones.

I'm happy because this means I can use FilmConvert as I had hoped - as an in-camera LUT as well as in post, it can work well within RED's IPP2 workflow, and I can still grade normally.
 
It does seem kinda weird; you would expect that all the different gammas/colourspaces (regardless of the camera even) should match or be relatively close to the film stock they're emulating when you select the correct source. If not, than why bother having the different profiles at all?

Hell, ideally, if I had a 5D3/Standard "converted" to 5213, it should look relatively close to RED/IPP2 converted to 5213... And both should look close to 5213, otherwise what the hell is the point of FC?
 
Mike P. said:
It does seem kinda weird; you would expect that all the different gammas/colourspaces (regardless of the camera even) should match or be relatively close to the film stock they're emulating when you select the correct source. If not, than why bother having the different profiles at all?

Hell, ideally, if I had a 5D3/Standard "converted" to 5213, it should look relatively close to RED/IPP2 converted to 5213... And both should look close to 5213, otherwise what the hell is the point of FC?
Click to expand...

Yeah, that's kind of their pitch, but it doesn't really pan out. I also got in touch with FilmConvert, and they told me that you need to double the ISO when using their IPP2 profile because "IPP2 was designed for better sensitivity." Seems to be better, but why they essentially built in a 1-stop pull in the IPP2 profile is beyond me. The Epic-W Helium DC2/RLF profile is also about a stop brighter than any other profile (including the Weapon Helium DC2/RLF), so there's weirdness all around.

It seems to work fairly decent with other cameras' footage, so I don't know why their Red profiles are such a mess.
 
Kevin Marshall said:
Yeah, that's kind of their pitch, but it doesn't really pan out. I also got in touch with FilmConvert, and they told me that you need to double the ISO when using their IPP2 profile because "IPP2 was designed for better sensitivity." Seems to be better, but why they essentially built in a 1-stop pull in the IPP2 profile is beyond me. The Epic-W Helium DC2/RLF profile is also about a stop brighter than any other profile (including the Weapon Helium DC2/RLF), so there's weirdness all around.

It seems to work fairly decent with other cameras' footage, so I don't know why their Red profiles are such a mess.
Click to expand...

It probably has to do with what you mentioned; the fact that there are a few different RED sensors now and they all use the same IPP2 and some (helium) are a stop brighter having remapped ISO (so 800 actually behaves like 800 and not 400 like on all the other sensors). Still bothersome though because they have different profiles for all the different sensors so you'd think it would've been taken care of during testing/profiling...
 
Last edited:
Marc Wielage said:
My advice would be not to use this and instead color-correct the image in a real color-correction program.
Click to expand...

And DPs for the last 100 years haven't had preferences on what film stock they wanted to start from? People shoot with Alexa because "it looks like film". Almost nobody wants to start from a perfectly scientific blank slate.
 
Kevin Marshall said:
In case anyone else cares about this - I think I've figured it out. Since the profile is labeled IPP2, I thought it might be meant to actually work in IPP2 as a Creative LUT, i.e. with no actual output transform. So I lowered the Curve setting to zero
Click to expand...

Yes, if the LUT is designed for base colour science, i.e. Log3G10 and RWG, the output curve should be set to none. Otherwise all bets are off.


Kevin Marshall said:
Ultimately it's all workable - I can just use the RLF preset, or add a push+sat to the IPP2 preset.
Click to expand...

Pushing Sat into a complex LUT designed for base colour science is not the best idea ever.

RLF and Log3G10 are completely different gammas affecting everything else and same LUT won't work the same, regardless of the manual tweaking. Some aspects of a complex LUT cannot be adequately compensated with post tools.

Gavin Greenwalt said:
And DPs for the last 100 years haven't had preferences on what film stock they wanted to start from? People shoot with Alexa because "it looks like film". Almost nobody wants to start from a perfectly scientific blank slate.
Click to expand...

In that case same LUT should be used for monitoring, affecting exposure and lighting decisions.
Otherwise, head-scratching in post is a likely scenario.
 
In the newest version of Resolve, if you select RWG and a standard gamma curve (2.2, BT.1886, REDgamma4, etc.) in the IPP2 raw settings and then drag the contrast slider to -1.00, you get a properly exposed result with the FilmConvert IPP2 setting. Not sure if this is better or worse than raising the ISO, but they seem nearly identical.
 
Last edited:
Mike P. said:
It does seem kinda weird; you would expect that all the different gammas/colourspaces (regardless of the camera even) should match or be relatively close to the film stock they're emulating when you select the correct source. If not, than why bother having the different profiles at all?

Hell, ideally, if I had a 5D3/Standard "converted" to 5213, it should look relatively close to RED/IPP2 converted to 5213... And both should look close to 5213, otherwise what the hell is the point of FC?
Click to expand...
It's for reasons like this that I'm not a fan. At best, I think all Film Convert does is give you some interesting and/or different looks, but I'm skeptical as to whether it really looks like a films look per se. And what I see sometimes damages the image. But I own it just on the possibility that a client will specifically ask for it. So far, they haven't.
 
Evan B said:
In the newest version of Resolve, if you select RWG and a standard gamma curve (2.2, BT.1886, REDgamma4, etc.) in the IPP2 raw settings and then drag the contrast slider to -1.00, you get a properly exposed result with the FilmConvert IPP2 setting. Not sure if this is better or worse than raising the ISO, but they seem nearly identical.
Click to expand...

Properly exposed/matching the DC2/RG4 stuff or properly looking film emulation? Still kinda frustrating, and uncranking the contrast so much makes me apprehensive.

Marc Wielage said:
It's for reasons like this that I'm not a fan. At best, I think all Film Convert does is give you some interesting and/or different looks, but I'm skeptical as to whether it really looks like a films tock per se. And what I see sometimes damages the image. But I own it just on the possibility that a client will specifically ask for it. So far, they haven't.
Click to expand...

I think they do some decent things with grain, and it can often look good... But, yeah, I definitely don't have a ton of faith that it's an accurate emulation.

Originally they had comparison frames -- the filmstock shot, next to a simultaneous redmx shot, and then the redmx shot with FC on, and they looked quite close (like pixel-peep close). But I think they should've had a shoot-guide basically saying, 'you have to use these settings (proper exposure at this ISO, WB, gamma, light sources, etc.) for the most accurate results with camera xxxxxx. Outside of that your mileage my vary.' It's just tough because content captured on film is still colour corrected/graded, but FC is applied more towards the end of the chain (because of the grain)... Or maybe I've used it wrong (which, again, a guide to maximize results would be helpful).
 
Last edited:
The issue is as follows.

EpicW’s (mine included) sometimes have a bug in the SDI/Prores out, that makes it look like its a stop to hot. So what happens, is the PRORES/SDI look one way, the RAW is recorded as correct to the light meter.

This makes the LUTS you apply on RCX or DAvinci seem dark if you expose to the monitor.
 
You must log in or register to reply here.
Back
Top