BokehBeauty wrote:
In PP of my Q3 43 images there is no match between the SOOC Vivid jpg and the DNG with the Cobalt Leica SL3 Vivid profile and no other modifications, see below. To my eye the Fuji Vivid profile suits much better in color and gradation.
BokehBeauty wrote:
In PP of my Q3 43 images there is no match between the SOOC Vivid jpg and the DNG with the Cobalt Leica SL3 Vivid profile and no other modifications, see below. To my eye the Fuji Vivid profile suits much better in color and gradation.
BokehBeauty wrote:
As I said, I didn't change anything in the Vivid JPG files and in the DNG files, except for adjusting the profiles for the DNG files.
Understood that you didn't tweak anything. But, what was the WB temp being used?
Did you shoot the Q3 on AWB and just let it decide ? Then applied the profiles without establishing a correct(ed) WB, based on the WB set by the Q3?
Is that correct ... or, did you use a pre-set WB (Daylight, Cloudy, etc.) for the Q3 ... and that value carried through (non-adjusted), etc. ?
Or, did you have a reference (checker, etc.) that you determined the lighting color / WB and set the Q3 to match the lighting prior to capture ... and that value carried through (non-adjusted), etc. ?
Just trying to understand how the matter of WB was utilized in the workflow for these.
RustyBug wrote:
Understood that you didn't tweak anything. But, what was the WB temp being used?
Did you shoot the Q3 on AWB and just let it decide ? Then applied the profiles without establishing a correct(ed) WB, based on the WB set by the Q3?
Is that correct ... or, did you use a pre-set WB (Daylight, Cloudy, etc.) for the Q3 ... and that value carried through (non-adjusted), etc. ?
Or, did you have a reference (checker, etc.) that you determined the lighting color / WB and set the Q3 to match the lighting prior to capture ... and that value carried through (non-adjusted), etc. ?
Just trying to understand how the matter of WB was utilized in the workflow for these.
Yes, the Q3 43 was on AWB.
I have another example of an Italian vegetable stall, where I can't get the colours to match even with tweaking the WB for jpg and DNG individually.
This is exactly the problem I have seen. The profiles are nice but don't match the JPEGs at all. The issue I have is publishing them as profiles for specific cameras when they don't match the JPEG output. I am not swayed by the explanation that they match mathematical LUT curves, etc... The test for a user should be matching the JPEGs.
As a note I have also done with the Nikon Z7 and Z9 cameras (which I owned at the time) and see the same issue. They are close but not quite correct. If you sell them as matching they should match, or there should be a disclaimer about not quite matching. This is not to say they don't make good looking profiles but I find them lacking in actually duplicating an existing look.
bcaslis wrote:
This is exactly the problem I have seen. The profiles are nice but don't match the JPEGs at all. The issue I have is publishing them as profiles for specific cameras when they don't match the JPEG output. I am not swayed by the explanation that they match mathematical LUT curves, etc... The test for a user should be matching the JPEGs.
As a note I have also done with the Nikon Z7 and Z9 cameras (which I owned at the time) and see the same issue. They are close but not quite correct. If you sell them as matching they should match, or there should be a disclaimer about not quite matching. This is not to say they don't make good looking profiles but I find them lacking in actually duplicating an existing look. ...Show more →
Yes, I wouldn't have bought the Leica profiles as I had no real need but the desire to match the Leica jpgs. I'm very happy with the Cobalt Fuji and the old Fuji XPro2 profiles and the Cobalt CCD S5pro FujiChrome profiles, but not with these at all.
BTW, DXO Leica SL2 and Q3 43 profiles match the Q3 43 jpg for Std and Vivid with little tweaking.
Ulysseita wrote:
Before going further, I think a few foundational points need to be clarified.
Cobalt base profiles are not just an entry point for creative emulations. They are designed first and foremost as a more appropriate starting point for cameras that typically cost $2000 or more without a lens. Compared to the default base profiles provided by most editing software, they rely on true dual or triple illuminant profiling, with significantly higher colour separation and without the compression typical of canned profiles.
Complaints about “less greens” or “less blues” after applying an emulation are often a matter of mismatched expectations. It’s comparable to buying a seven-seat station wagon and complaining that it doesn’t handle curbs like a sports car, or buying a Lamborghini and complaining about trunk space. These are different tools designed for different purposes.
The colourimetry of our emulations is not a subjective choice and it’s not something we tweak by eye. We are not moving sliders to make cameras vaguely resemble each other. The process is based on calibrated physical captures and complex mathematical modelling. This is mathematics, not taste.
In the vast majority of cases where something “doesn’t look right”, the issue comes down to expectations of a one-click solution, or to errors in white balance evaluation. In the remaining small percentage of cases, we are always happy to investigate using real files and resolve genuine issues, as we have consistently done in the past.
Different workflows have different needs, and that’s fine. These tools are built for users who want a precise, neutral and technically sound foundation, not a finished look baked in at the profile level....Show more →
This text was written by an LLM (as are the rest of your replies in this thread). I'm 100% sure because I've used Claude and ChatGPT a lot and this is exactly how those LLMs write.
Can you answer me (without using AI) why I should not expect my Sony A7CII pictures to look like they were taken with a Nikon Z9 after applying your presets? Because that's what I thought would happen. But you're saying I had the wrong expectations thinking that. Apparently my Sony is a station wagon and Nikon Z9 is a Lamborghini? Or I'm not entirely sure what you meant with that comparison. Or is there something else I have to adjust? There are no other steps mentioned on your website.
I will happily post some example pictures if anyone is interested. Maybe I just have something wrong with my eyes and the colors actually look exactly like Nikons. You'll be the judge. I just have to do a fresh Windows install first because I can no longer open Photoshop. Either way, the way Ulysseita writes does not inspire me with confidence. But I will happily apologize if I'm proven wrong or told what I'm doing wrong when processing the pictures.
jaygould wrote:
This text was written by an LLM (as are the rest of your replies in this thread). I'm 100% sure because I've used Claude and ChatGPT a lot and this is exactly how those LLMs write.
I've used AI ... and have had multiple instances where they have actually returned comments, that came directly from FM ... and, were even parroting threads that I myself had posted. I asked AI ... and it got its answer from me, trying to answer my question.
Not saying this one was / wasn't ... but I'm not 100% convinced that a "match" between a real person and AI means the person copied AI. It seems that the more esoteric the question / topic, the less pool of info AI has available to synthesize ... so, where the topic / question is specific enough to be a 1:1 relationship ... yeah, the AI return can be a match. Real Person = AI = Real Person = AI. Just saying, we need to be judicious about assertions of origins, when it comes to AI. At best, AI is a synthetic regurgitation of what exists from former origination(s).
Now, whether a person is copying such regurgitations vs. is AI copying the person ... that gets tricky.
Kinda reminds me of the old Memorex commercial ... "Is it live, or is it Memorex?"
RustyBug wrote:
I've used AI ... and have had multiple instances where they have actually returned comments, that came directly from FM ... and, were even parroting threads that I myself had posted. I asked AI ... and it got its answer from me, trying to answer my question.
Not saying this one was / wasn't ... but I'm not 100% convinced that a "match" between a real person and AI means the person copied AI. It seems that the more esoteric the question / topic, the less pool of info AI has available to synthesize ... so, where the topic / question is specific enough to be a 1:1 relationship ... yeah, the AI return can be a match. Real Person = AI = Real Person = AI. Just saying, we need to be judicious about assertions of origins, when it comes to AI. At best, AI is a synthetic regurgitation of what exists from former origination(s).
It's not the info but the writing style. The em dashes (—) are the clear giveaways (check the next reply after the one I quoted). I was forced to read multiple books written by Claude AI recently and this is exactly how it writes.
Using AI to help edit text and check grammar is fine but in this case it does not inspire confidence. If he knows what he's talking about he should be able to reply without help from an AI and in simple language that people will understand. Based on other replies here it seems I am not the only person who is confused by the verbose explanations and the overly technical language and strange analogies.
jaygould wrote:
The em dashes (—) are the clear giveaways (check the next reply after the one I quoted).
So, does that then infer that the reply by Raamiel was also AI generated. That reply also used two em dashes, just like the "next one" that you mention.
It strikes me that if an AI engine was using ... ellipses as part of its writing style, then someone could suggest that darn near everything I write is AI derived. Makes me wonder if the original author used the em dash, and as such the assessment that AI matches the author's style is indication that it was AI, rather than original. Kind of a "chicken and egg" circular reasoning, if you will.
That said ... I'll be the first to acknowledge that I'm not a forensic expert in AI, so I clearly can't provide any authoritative context on whether it is or isn't. But, I'm not certain that where a circular response (mind you, AI circled me personally multiple times) potential exists, that I can immediately jump on the "must be AI". Maybe it is, maybe it isn't.
Personally, it strikes me that folks might be wanting a matchy-matchy without recognizing the technical impact of the predicate WB issue underlying the differences, and may be looking to discount the technical explanation. Multi-cultural communications can be a tough gig sometimes, so I'm prone to afford a little latitude in that regard, if something is readily clear. But, in this case, the interaction to request specific clarification (imo) has been presented in a manner that didn't trigger my "must be AI" radar.
Short answer, I can't confirm either way, so I'm not as 100% confident as you present. I'd like to think that the responses have been genuine, not AI ... but, reciprocally ... I'll give similar credence for it to be proven one way or the other.
I have to admit, the accusation itself makes me smile a bit.
If the implication were that our work is just “AI-generated claims”, then it would also imply that we’re selling empty promises generated by a language model — which clearly makes no sense given that the products are based on measured data, real cameras, and reproducible results used daily by thousands of photographers.
The reality is much simpler: English is not my native language. When writing longer or more technical explanations, I sometimes use tools like ChatGPT to help polish the language, not to invent the technical content. The ideas, the testing, and the conclusions are mine.
I also find it a bit odd that using a language tool is framed as suspicious, as if everyone in the world were expected to be a native English speaker. Photography is global, and so are the people discussing it here.
Or, if that makes things easier, I can simply write in Italian and let others take on the task of translating technical language that isn’t their own.
Jokes aside, this is exactly why language tools exist: to help non-native speakers communicate complex technical topics clearly in English.
I’d much rather keep the discussion focused on color science and workflow than on language mechanics.
BokehBeauty wrote:
In PP of my Q3 43 images there is no match between the SOOC Vivid jpg and the DNG with the Cobalt Leica SL3 Vivid profile and no other modifications, see below. To my eye the Fuji Vivid profile suits much better in color and gradation.
Ulysseita wrote:
I have to admit, the accusation itself makes me smile a bit.
If the implication were that our work is just “AI-generated claims”, then it would also imply that we’re selling empty promises generated by a language model — which clearly makes no sense given that the products are based on measured data, real cameras, and reproducible results used daily by thousands of photographers.
The reality is much simpler: English is not my native language. When writing longer or more technical explanations, I sometimes use tools like ChatGPT to help polish the language, not to invent the technical content. The ideas, the testing, and the conclusions are mine.
I also find it a bit odd that using a language tool is framed as suspicious, as if everyone in the world were expected to be a native English speaker. Photography is global, and so are the people discussing it here.
Or, if that makes things easier, I can simply write in Italian and let others take on the task of translating technical language that isn’t their own.
Jokes aside, this is exactly why language tools exist: to help non-native speakers communicate complex technical topics clearly in English.
I’d much rather keep the discussion focused on color science and workflow than on language mechanics....Show more →
AI polished ... not, AI generated.
That explains why I felt the content was not AI generated, but had originated from the OP ... because, indeed it had. OP > AI > OP.
Ulysseita wrote:
I have to admit, the accusation itself makes me smile a bit.
If the implication were that our work is just “AI-generated claims”, then it would also imply that we’re selling empty promises generated by a language model — which clearly makes no sense given that the products are based on measured data, real cameras, and reproducible results used daily by thousands of photographers.
The reality is much simpler: English is not my native language. When writing longer or more technical explanations, I sometimes use tools like ChatGPT to help polish the language, not to invent the technical content. The ideas, the testing, and the conclusions are mine.
I also find it a bit odd that using a language tool is framed as suspicious, as if everyone in the world were expected to be a native English speaker. Photography is global, and so are the people discussing it here.
Or, if that makes things easier, I can simply write in Italian and let others take on the task of translating technical language that isn’t their own.
Jokes aside, this is exactly why language tools exist: to help non-native speakers communicate complex technical topics clearly in English.
I’d much rather keep the discussion focused on color science and workflow than on language mechanics....Show more →
After several tests, I can confirm that the Core Looks of the Q3, SL, and other cameras — including Standard, Vivid, Natural, BW Natural, and BW High Contrast — differ from model to model, and in the case of our packages, as expected, they are an emulation of the SL3. As a result, it makes no sense to compare them with the SOOC JPEGs produced by Q, M, etc.
We are not yet certain whether the downloadable Looks, namely Essential and Artistic, also differ, but it is reasonable to assume that if the base JPEG engine changes from one model to another, the resulting Looks will change as well.
Ulysseita wrote:
After several tests, I can confirm that the Core Looks of the Q3, SL, and other cameras — including Standard, Vivid, Natural, BW Natural, and BW High Contrast — differ from model to model, and in the case of our packages, as expected, they are an emulation of the SL3. As a result, it makes no sense to compare them with the SOOC JPEGs produced by Q, M, etc.
We are not yet certain whether the downloadable Looks, namely Essential and Artistic, also differ, but it is reasonable to assume that if the base JPEG engine changes from one model to another, the resulting Looks will change as well....Show more →
Thanks to the Cobalt team for all the hard work on these profiles, and for stopping by to support everyone in this thread. I'm starting to warm up to some of the Leica Looks, especially 'Contemporary'. Do you plan to keep updating the Leica Looks pack as Leica releases new ones? I just noticed they have added a new 'Silver' look.
That reply also used two em dashes, just like the "next one" that you mention.