Primers colour management: transformation, optimisation, unification

[Hrvoje Simic^]

How to develop ACES to avoid/minimise processing issues and achieve fully natural & utmost quality imagery from digital cameras within cca 1 year :

1) Face the possibility of having to alter the paradigm and pipeline

2) Contact Omeneo for next steps

 

[Hrvoje Simic^]

Human desire for emulating film on digital is in great part an attempt of escaping digital processing image degradation and unnatural looking imagery.

As film look was the only natural looking reference Man has seen ever on the screen...



Primers for Canon Gen10

 

[Hrvoje Simic^]

Hmmm...

 

[Hrvoje Simic^]

 

[Anton Meleshkevich]

So it took 5 pages to answer the basic question, what the inputs and outputs of the LUTs are. If the developer can't even communicate with colorists using common (for colorists) terms, how can we expect that this product was created by a person who has at least some experience in color grading?

What about SDR HDR match? Do you rely on per-channel DaVinci tonemapping?
Do you use per-channel tone-mapping?
Any pics of out-of-gamut colors handling? Both real word footage and charts?
What happens with out-of-gamut colors in shadows? Are they clipped like in traditional per-channel DRTs? Or soft-clipped? Any good examples?
What is considered to be ground truth? It's impossible with current display technologies (screen size, screen curvature, surround, gamut, dynamic range, mono).
So it's some algorithm that predicts human perseption of colors? If your LUTs are DRT.
Or is it a log-to-log sensor optimisation that brings virtual out-of-gamut colors of a sensor back into spectral locus? Sort of a gamut mapper based on real measurements? What it has to do with going from log to SDR then?

For now it's a ton of text that means nothing and just spamming with images.

 

[Hrvoje Simic^]

For now it's a ton of text that means nothing and just spamming with images.

Ever heard "picture is worth a thousand words" ?

The approach you have chosen doesn't lead to answers you are looking for.

 

[Luis Carnevale]

After taking a closer look at the content within this thread, it's clear that this discussion feels like it's going nowhere. The developer has flooded the conversation with an excessive amount of images and lengthy, confusing responses that lack any real proof from hands-on testing by experienced colorists. I hope that changes. When I examine the so-called "Superior imaging performance" stills, I struggle to find any notable improvements in image quality that couldn't be achieved by using Gamut Mapping OFX and making some targeted adjustments.
And then there's a lack of transparency about the potential costs associated with purchasing and implementing Primers, which is a big piece of information for anyone considering adopting this technology. Speaking for myself... it's my honest opinion that the current direction of this discussion is unproductive and unlikely to provide any meaningful insights or practical conclusions for colorists who are looking for genuine advancements

 

[Hrvoje Simic^]

Writing between work with little spare time, quotes separated for clarity:

After taking a closer look at the content within this thread, it's clear that this discussion feels like it's going nowhere.

Discussion consists of participants. Otherwise monologue.

The developer has flooded the conversation with an excessive amount of images and lengthy, confusing responses that lack any real proof from hands-on testing by experienced colorists.

- The developer has developed, is further developing, has presented
- Images are used for presenting outcomes of image processing (with some stages of development)
- Real proof is in images and direct experience, it doesn't manifest through word exchange
- Experienced colourists already use these

I hope that changes.

Good news, it's already different.

When I examine the so-called "Superior imaging performance" stills, I struggle to find any notable improvements in image quality that couldn't be achieved by using Gamut Mapping OFX and making some targeted adjustments.

- Image processing errors and distortions/deviations cannot be undone, not the best idea ever to attempt if avoidable in the first place
- Image engineering for proper non linear transformation for display is far too complex and specific and requires tools and multiple processes outside post, is misplaced in post and user tampering for a bunch of reasons
- Oversimplification is a tendency, underestimation also
- Tackling preconceptions externally is a futile endeavor, it only works internallly if addressed by their author

And then there's a lack of transparency about the potential costs associated with purchasing and implementing Primers,

- Pricing for sets is on every camera model page
- There is no additional cost in implementing sets of 3DLUTs
- There is reduction of time/cost by avoding fiddling with speculative simplistic image engineering in post

it's my honest opinion that the current direction of this discussion is unproductive and unlikely to provide any meaningful insights or practical conclusions for colorists who are looking for genuine advancements

Reconsidering helps.

 

[Hrvoje Simic^]

Upper post:

- Log part clearly covered in the first post of the thread, visible on every camera model page
- Gamut handling addressed multiple times in the thread, also shown in multiple examples, compared to camera maker's LUTs, ACES and RCM, also on the web
- 709 out talk postponed due to misconceptions and beliefs about display out conversion, preconceptions first, understanding later, imaging signal chain is not the only thing which requires proper order for proper outcomes
- SDR/HDR relation already covered, also clearly stated on the web
- Belief, debating belief
- Specifics on engineering procedures cannot be disclosed
- Asking to see, "spamming with images" = contradiction

 

[Luis Carnevale]

Hrovje, I won't reply to your responses as this thread is now going 10 pages and any further discussion is unfortunately unproductive and I can't see any member of this appreciated coloris/tDI community is in favor of your technology.

One thing I'm still interested is the price tag: you have stated "Pricing for sets is on every camera model page". For the Alexa I can't find on the page. Can you help me on this?

 

[Hrvoje Simic^]

Hrovje, I won't reply to your responses as this thread is now going 10 pages and any further discussion is unfortunately unproductive and I can't see any member of this appreciated coloris/tDI community is in favor of your technology.

One thing I'm still interested is the price tag: you have stated "Pricing for sets is on every camera model page". For the Alexa I can't find on the page. Can you help me on this?

Have a nice day, Luis.

 

[Michael Weinheim]

One thing I'm still interested is the price tag: you have stated "Pricing for sets is on every camera model page". For the Alexa I can't find on the page.

Simple answer to your simple question:

$55 für Alexa 35.
$54 for Alev3 sensors.

It's not very easy or intuitive to find: chose the manufacturer, then click on the image of the camera model. There are instructions, buried in a block of text.

This overly complicated - and in this thread capricious - communication is not beneficial when trying to get people interested in your product.

 

[Hrvoje Simic^]

- Alexa 35 set goes up soon, examples online are from previous generation
- Alexas ALEV3 set goes up after next update aligned to last A35 generation

also:

- Luis didn't get an answer because disqualified himself from it, because upper mess (strike #1) & see bolded (strike #2)
- There is a reason prices are where they are and info
- Easy comes after the harder part

also:

- Complaint department seeking playing buddies for battling sprouting imaginary windmills + people ignoring/devaluing/rejecting written and shown are likely to face dissatisfaction in this thread


Anyone genuinely interested in the project, feel free to ask questions or get in touch.

 

[Juan Pablo Zambrano]

- Alexa 35 set goes up soon, examples online are from previous generation
- Alexas ALEV3 set goes up after next update aligned to last A35 generation

also:

- Luis didn't get an answer because disqualified himself from it, because upper mess (strike #1) & see bolded (strike #2)
- There is a reason prices are where they are and info
- Easy comes after the harder part

also:

- Complaint department seeking playing buddies for battling sprouting imaginary windmills + people ignoring/devaluing/rejecting written and shown are likely to face dissatisfaction in this thread


Anyone genuinely interested in the project, feel free to ask questions or get in touch.

One way you could help colorists feel at ease could be by giving away a sample? if you are worried about it been shared around you could save it into an encrypted DCTL with a copyright pattern so they can't make a copy.

I have created my own picture formation method as well and making as easy to use its a must. Changing the way people see these transformation is not easy (even when its free) so you need good, simple and different kinds of examples. When the user gets the basics, you can further expand to more complex/hard to see explanations/examples. I think you have tried to do this but most of your examples have not worked that well, I am sorry to say.
"Picture is worth a thousand words", but when the images look almost identical it's hard to parse what the viewer should be comparing, even if the transformation is great.

Some of the most open to conversation Colorists that I have met have tried to understand your approach, asking you questions, but they have met with mean responses or half answers. This community even when "fighting"/disagreeing has tried to have the most open conversation possible in many other threads, and this one is no different. You will find many free tools, and paid ones as well, and almost all of them have been welcomed as long as they are explained to satisfaction.

 

[Michael Weinheim]

Showing parts of the workflow we developed and how we get superior results than billion dollar corporations would not be very smart.

Primers function is digital camera image optimisation.
What Primers are essentially would require a more philosophical, poetic approach.

A better way to demonstrate your product's 'superiority' than before/after images would be to allow users to test it themselves.

I think part of the problem is that you make very bold, even scientific claims and then expect people to take your word and a number of before/after pictures as evidence.

Anyone genuinely interested in the project, feel free to ask questions or get in touch.

It seems to me that everyone who asked questions was genuinely interested. I'd still be if I could test it. You said this is the 8th year you work on this product and that it's not part of your work to engage with doubters. If you don't want to engage it's fine, not everyone likes that part.
A good idea then would be to let people test the product.

 

[Hrvoje Simic^]

I think part of the problem is that you make very bold, even scientific claims and then expect people to take your word and a number of before/after pictures as evidence.

Problems are in the heads and nowhere else.
This project doesn't deal with these types of problems.

Case examples of problem sprouting over here for which someone else is the problem of course:
# 1 - You imagining there are some expectations over here and adressing me as if your imagination is reality
# 2 - See above wishful thinking
# 3 - See above projecting blame for missing elements while lacking the will to read and check the page
# 4 - See above stated preconceptions combined with projecting blame for the lack of understanding

Mor useful info coming up, no problem there.

 

[Juan Pablo Zambrano]

Okay, I wanted to test primer, as I really couldn't understand from the post in here how they are supposed to work and I wanted to share my findings.

I bought the GH5 primers.
- First thing I noticed is that nowhere in the manual or the LUTs themselves did they specify what kind of color space they output, the input its assumes to be Vlog , V-Gamut.
- Second, it comes with 5 different LUTs. They seem to mainly differ by their curve output, from low contrast to high contrast
- Third, The LUTs seem to be very smooth,and the transformation seems very well encoded into the them
- Fourth, the look is not my favorite but this could go back to the two first points. It's hard to judge them fairly as I don't know what they are suppose to output.
- Forth, the LUTs I received are dated from february of 2024, from this I can assume that they are relatively new, and should reflect the intented result.

Analysis of the 4 previews points:
- Baking these kinds of transformations is not inherently a bad thing .
- The "base" LUT is referred as "log-like" in the manual , and "as a good starting point for grading", so I am guessing this means that they are supposed to be seen as log-to-log transformations,but if this is the case it doesn't say for what "DRT" they are made for.
- The "DMax" and "Full" are the higher contrast ones, and are referred as " typically closest in contrast to standard log conversions", this to me (might be wrong) means that they are V-gamut in and out, but already "compressed" to display, so they would need an extra CST to convert to Rec709 or what ever else, in this case it doesn't say what display type they are, 2.4? 2.2? HLG? .
- Using all the transformations they look very wrong if assumed to be directly converting to rec709 , this could corroborate my hunch that this are not to be seen as Vlog to Display convertions.
- Last point, this transformations are not well explained or at least don't adhere to any common knowledge process by any other manufacturer, this is potentially okay if explained in depth but this is not the case. In this thread and on the website the "color managment" settings used in Davinci Resolve don't make any sense, and seem to show a lack of understanding of how Resolve deals with transformations.

Hopefully this information helps the people here understand the product better

 

[Hrvoje Simic^]

Hey Juan,
Thank you for both posts, response on both coming up, slammed with work and exhausted...

 

[Hrvoje Simic^]

One way you could help colorists feel at ease could be by giving away a sample? if you are worried about it been shared around you could save it into an encrypted DCTL with a copyright pattern so they can't make a copy.

I have created my own picture formation method as well and making as easy to use its a must. Changing the way people see these transformation is not easy (even when its free) so you need good, simple and different kinds of examples. When the user gets the basics, you can further expand to more complex/hard to see explanations/examples. I think you have tried to do this but most of your examples have not worked that well, I am sorry to say.
"Picture is worth a thousand words", but when the images look almost identical it's hard to parse what the viewer should be comparing, even if the transformation is great.

Some of the most open to conversation Colorists that I have met have tried to understand your approach, asking you questions, but they have met with mean responses or half answers. This community even when "fighting"/disagreeing has tried to have the most open conversation possible in many other threads, and this one is no different. You will find many free tools, and paid ones as well, and almost all of them have been welcomed as long as they are explained to satisfaction.

Thank you for all the suggestions, Juan.

On the more info - yes of course. It was already intended for earlier this year but work on multiple fronts stepped in.
Also, more structured info coming up on this thread covering many aspects. Some approaches will differ from the expected.
Also, more info on the online examples and what to look for.

On the freebies - there is no intent in luring people nor attracting anyone through freebies. Neither by words like "easy" or "quick and easy" or "great results with one click"...although they would be true.
You can deduce why.

Software package is being considered.

"Mean" - likely misinterpreted or taken out of context
"Half answers" are not there by omission. Some things require active participation.

On the testing...well...this project is dedicated to DoPs, DITs and Colourists. Also...post 193, last line. Some things do not have to be underlined.

 

[Hrvoje Simic^]

Okay, I wanted to test primer, as I really couldn't understand from the post in here how they are supposed to work and I wanted to share my findings.

I bought the GH5 primers.
- First thing I noticed is that nowhere in the manual or the LUTs themselves did they specify what kind of color space they output, the input its assumes to be Vlog , V-Gamut.
- Second, it comes with 5 different LUTs. They seem to mainly differ by their curve output, from low contrast to high contrast
- Third, The LUTs seem to be very smooth,and the transformation seems very well encoded into the them
- Fourth, the look is not my favorite but this could go back to the two first points. It's hard to judge them fairly as I don't know what they are suppose to output.
- Forth, the LUTs I received are dated from february of 2024, from this I can assume that they are relatively new, and should reflect the intented result.

Analysis of the 4 previews points:
- Baking these kinds of transformations is not inherently a bad thing .
- The "base" LUT is referred as "log-like" in the manual , and "as a good starting point for grading", so I am guessing this means that they are supposed to be seen as log-to-log transformations,but if this is the case it doesn't say for what "DRT" they are made for.
- The "DMax" and "Full" are the higher contrast ones, and are referred as " typically closest in contrast to standard log conversions", this to me (might be wrong) means that they are V-gamut in and out, but already "compressed" to display, so they would need an extra CST to convert to Rec709 or what ever else, in this case it doesn't say what display type they are, 2.4? 2.2? HLG? .
- Using all the transformations they look very wrong if assumed to be directly converting to rec709 , this could corroborate my hunch that this are not to be seen as Vlog to Display convertions.
- Last point, this transformations are not well explained or at least don't adhere to any common knowledge process by any other manufacturer, this is potentially okay if explained in depth but this is not the case. In this thread and on the website the "color managment" settings used in Davinci Resolve don't make any sense, and seem to show a lack of understanding of how Resolve deals with transformations.

Hopefully this information helps the people here understand the product better

Thank you for the purchase Juan.

- Primers will optimise the image data to land properly in 709/2.4, no other colour management needed for 709 whatsoever. These are supposed to simplify your life, not complicate it. If you use these you don't use any other conversions (for 709).
- Once the image data is optimised for displaying (which is chosen to be 709/2.4 for obvious reasons) from there you go further if you need by choosing appropriate CST using 709/2.4 as input and that CST will scale a fully formed and properly normalised non-linear image data to something else.
- The look is most neutral for the sensor response & camera maker's scene referred coding colour science. Take this as "no creative look".
- Make sure to check many different scenarios, preferably with tonally rich high contrast scenes and compare to Panasonic LUT, RCM and ACES. Not in the grading session and in the way you evaluate looks, export stills and toggle between A/B/C routes many times back and forth.
- GH5 is older sensor/processing tech, with newer cameras more things pop out as they can capture more and capture it better. If you are using some newer camera supported, get in touch.
- Make sure you use these exactly on the camera they are designed for. If applied on other camera material you will not have proper tonal distribution and the image may feel "off" even if you cannot pinpoint the exact thing in the image.
- If you use GH5 camera you will get best results by exposing based on previewing these. In this case you will need external monitor with Primers loaded, internal LUT processing in these cameras is...limited.
- If by "baking" these you mean in recording, I'd avoid this. These cameras use 17^3 internal LUT processing, insufficient for proper log-to-display conversion.
- Each Primer works applied separately, not combined. More flat give more room for manual contrast design, 01 Base is most flat, 02 NTR is mid contrast with very gentle feel and very valuable in high contrast scenes.
- 01 Base was in previous iterations closer to log, after more usage, testing and tweaking it is now more flat and will make manual contrast shaping more intuitive
- 04 Full is your reference for full contrast image landing on 709. 05 DMax is about half of stop pushed down.
- Log to log route was considered, tested and discarded, no need.


(This post is not the "more info coming up" from previous thread)