p.5 #1 · New Cobalt Kodachrome 25 and 64 film profiles
Nifty Fifty wrote:
I wonder how a preset can possibly achieve the look of a particular film, given that film has a defined and consistent white balance under all conditions, whereas digital white balance is not standardized, neither during the photographic capture (from the sensor) nor during processing (from the editing software). In my opinion, these "scientifically created" film simulations mislead the unsuspecting buyer. However, I'm happy to be proven wrong if someone is capable of doing so.
Or is an integrated white balance fixed and binding in the preset?
The best way to build a profile scientifically is to build it with white balance set to daylight (or tungsten, if that's what the film is balanced for). It is then also crucial to have training data which shows the color response for a variety of lighting scenarios in terms of color and intensity as film does not react linearly the same way digital does - you get color shifts depending on light intensity, e.g. your red could lean orange in the highlights and magenta in the shadows.
I have never seen a digital preset that went that far. It seems this one is no exception.
I'm not even sure if it's really possible with just a single third-party profile. Maybe a better approach would be creating a neutral pre-set that you apply to the RAW file, then export as tiff and a separate software runs a scene recognition algorithm first (e.g. daylight, artifical light, blue hour, sunset, etc) and then applies a matching film profile that suits the light in the scene.
p.5 #2 · New Cobalt Kodachrome 25 and 64 film profiles
fjablo wrote:
The best way to build a profile scientifically is to build it with white balance set to daylight (or tungsten, if that's what the film is balanced for). It is then also crucial to have training data which shows the color response for a variety of lighting scenarios in terms of color and intensity as film does not react linearly the same way digital does - you get color shifts depending on light intensity, e.g. your red could lean orange in the highlights and magenta in the shadows.
I have never seen a digital preset that went that far. It seems this one is no exception.
I'm not even sure if it's really possible with just a single third-party profile. Maybe a better approach would be creating a neutral pre-set that you apply to the RAW file, then export as tiff and a separate software runs a scene recognition algorithm first (e.g. daylight, artifical light, blue hour, sunset, etc) and then applies a matching film profile that suits the light in the scene. ...Show more →
Here's a somewhat more detailed and technical description of what I mean, courtesy of Gemini:
Why Standard Presets and Profiles Fail
The core issue is that digital sensors and photochemical film capture light in fundamentally different ways. A digital sensor is a linear photoelectric device; it simply counts photons. Film has a non-linear, logarithmic response driven by complex chemistry. Standard presets fail because they rely on static mathematical transformations (like 3x3 color matrices or basic tone curves) that cannot adapt to changing variables.
Specifically, standard profiles ignore "chemical crosstalk." Real film uses Development Inhibitor Releasing (DIR) couplers. When you heavily overexpose a specific color on film, these couplers release chemicals that physically suppress the development of adjacent color layers. This means a color's hue and saturation shift dynamically based purely on the intensity of the light hitting it. A static digital preset only maps "Color A" to "Color B"—it cannot dynamically bend hues based on varying exposure levels or adapt to how film dyes react differently to shifting color temperatures.
What we'd need to have instead of a 3D LUT:
1. Scene-Dependent Neural Networks
Researchers have developed deep learning frameworks that analyze the global and local context of a scene (such as evaluating the image's luminance and color histograms) to understand the specific lighting and exposure conditions before processing,. Instead of relying on a deterministic, static imaging model, the neural network dynamically generates a scene-dependent color transformation that accurately mimics how film chemically responds to that specific environment.
2. Spatially Adaptive 4D LUTs (SA-LUTs)
This concept works almost exactly like picking from an array of LUTs. Instead of applying one global 3D LUT, an AI "Context Generator" analyzes the frame to detect different regions, luminance ranges, and lighting contexts. It then dynamically selects, weights, and blends different tailored LUT transformations (forming a 4D LUT) for different areas of the image.
By using this secondary analysis step, the system becomes context-aware. It ensures that a heavily overexposed red object in daylight receives a completely different mathematical adjustment than a normally exposed red object in a dark room, perfectly replicating the complex, non-linear chemical crosstalk of real film.
p.5 #3 · New Cobalt Kodachrome 25 and 64 film profiles
Basically, all we need is the acceptance that digital is not analog and the realization that you should put in the work to expose film if you want film results.
Film presets are like a Nikon Zf. They evoke analog photography, but it's just folklore.
p.5 #4 · New Cobalt Kodachrome 25 and 64 film profiles
Nifty Fifty wrote:
Basically, all we need is the acceptance that digital is not analog and the realization that you should put in the work to expose film if you want film results.
Film presets are like a Nikon Zf. They evoke analog photography, but it's just folklore.
I mostly agree.
However it is theoretically possible to emulate the analog imaging process with digital. It's just not easy and all those emulations fail.
p.5 #5 · New Cobalt Kodachrome 25 and 64 film profiles
Nifty Fifty wrote:
Basically, all we need is the acceptance that digital is not analog and the realization that you should put in the work to expose film if you want film results.
Film presets are like a Nikon Zf. They evoke analog photography, but it's just folklore.
The funny thing is- Film is digital-
Go shoot film- If its doesn’t go directly to print, you are manipulating the colors, the hue, the contrast - All the things- It’s actually why I stopped screwing around with film because it all had to be manipulated to taste to get the end results- I wanted my film to look like the film, not the film to look the interpretation of the scanner and the lab guy as to how that film should look.
Currently having Claude build on his code to build a little app for my Mac that also handles the RAW conversion bit
Did you read about what Cobalt is doing?
Elite Kodachrome by Cobalt
Kodachrome, not “Kodachrome-inspired”
Kodachrome Elite is our highest-end Kodachrome emulation, built to reproduce the distinctive colour relationships and tonal behaviour of real Kodachrome slides with a controlled, repeatable workflow. This is designed for photographers who care about consistency, not quick stylistic presets.
Built on access to authentic Kodachrome slide development
After years of work, we gained access to a privately preserved K-14 Kodachrome slide-development workflow tied to genuine Kodachrome slide processing, maintained within a family lineage connected to a former Kodak employee and carried forward by a professional chemist. This was not a modern “Kodachrome-style” experiment or a commercial approximation. It enabled us to work from Kodachrome slides developed specifically as controlled reference material for measurement and modelling. Kodachrome Elite exists in this form only because we were able to ground it in true Kodachrome slide development, rather than imitations.
Precision in colourimetry and tonal response
Kodachrome Elite is engineered for film-referenced colourimetry and a purpose-shaped contrast response. Contrast is intentionally designed to reflect Kodachrome’s characteristic separation and presence, rather than a neutral raw rendering.
Three profile variants: NB, B, and PR
Each Kodachrome Elite profile is provided in three variants, representing three authentic ways Kodachrome is experienced in practice:
NB (Non-Balanced): Scan-realistic. Preserves the natural scan bias and mood you typically see when the grey axis is not fully normalised. This is the “as-scanned character” option.
B (Balanced): Grey-axis aligned. Balanced through the full greyscale for a neutral, predictable foundation and colour-critical consistency.
PR (Projected): Slide-projection interpretation. Tuned to evoke the perceptual experience of a living transparency on a screen, with the punch and presence associated with projection rather than a flat scan.
p.5 #7 · New Cobalt Kodachrome 25 and 64 film profiles
RustyRus wrote:
Did you read about what Cobalt is doing?
Elite Kodachrome by Cobalt
Kodachrome, not “Kodachrome-inspired”
Kodachrome Elite is our highest-end Kodachrome emulation, built to reproduce the distinctive colour relationships and tonal behaviour of real Kodachrome slides with a controlled, repeatable workflow. This is designed for photographers who care about consistency, not quick stylistic presets.
Built on access to authentic Kodachrome slide development
After years of work, we gained access to a privately preserved K-14 Kodachrome slide-development workflow tied to genuine Kodachrome slide processing, maintained within a family lineage connected to a former Kodak employee and carried forward by a professional chemist. This was not a modern “Kodachrome-style” experiment or a commercial approximation. It enabled us to work from Kodachrome slides developed specifically as controlled reference material for measurement and modelling. Kodachrome Elite exists in this form only because we were able to ground it in true Kodachrome slide development, rather than imitations.
Precision in colourimetry and tonal response
Kodachrome Elite is engineered for film-referenced colourimetry and a purpose-shaped contrast response. Contrast is intentionally designed to reflect Kodachrome’s characteristic separation and presence, rather than a neutral raw rendering.
Three profile variants: NB, B, and PR
Each Kodachrome Elite profile is provided in three variants, representing three authentic ways Kodachrome is experienced in practice:
NB (Non-Balanced): Scan-realistic. Preserves the natural scan bias and mood you typically see when the grey axis is not fully normalised. This is the “as-scanned character” option.
B (Balanced): Grey-axis aligned. Balanced through the full greyscale for a neutral, predictable foundation and colour-critical consistency.
PR (Projected): Slide-projection interpretation. Tuned to evoke the perceptual experience of a living transparency on a screen, with the punch and presence associated with projection rather than a flat scan....Show more →
Yes I did. It is a long and windy way of saying "we built a 3D LUT". It fails the same way all other pre-sets fail, no magic here.
p.5 #9 · New Cobalt Kodachrome 25 and 64 film profiles
Cobalt falls into the pompous "we build PROFILES, not presets" camp, but at the end of the day, it's simply curve and color manipulation just like everything else. There is no accurate way to fully emulate film on digital in Lightroom, full stop.
I find that the in-camera jpeg processing of Fuji and Nikon cameras comes closer than anything I've done in LR, and that jpegs from old CCD sensor cameras come even closer, thanks to their huge pixels.
p.5 #10 · New Cobalt Kodachrome 25 and 64 film profiles
RustyRus wrote:
The funny thing is- Film is digital-
But completely different digitally, without a defined, uniform grid and without silver crystals that are 100% identical in size and shape. RustyRus wrote:
Go shoot film- If its doesn’t go directly to print, you are manipulating the colors, the hue, the contrast - All the things- It’s actually why I stopped screwing around with film because it all had to be manipulated to taste to get the end results- I wanted my film to look like the film, not the film to look the interpretation of the scanner and the lab guy as to how that film should look.
That wasn't the point. It was about trying to make fully digital image files look like something they are not.
BTW: Incidentally, I photographed exclusively with analog cameras until I bought my first and so far only digital camera in 2023.
p.5 #12 · New Cobalt Kodachrome 25 and 64 film profiles
RustyRus wrote:
Good thing you have Claude I guess
No it's good that we have people like Andrea Volpato who is thinking this through and able to map a proper imaging pipeline that simulates the way how film and paper interact with light.
The only thing Claude is doing is building a nice application around Andrea's code and some open source libraries that handle RAW files.
p.5 #13 · New Cobalt Kodachrome 25 and 64 film profiles
fjablo wrote:
No it's good that we have people like Andrea Volpato who is thinking this through and able to map a proper imaging pipeline that simulates the way how film and paper interact with light.
The only thing Claude is doing is building a nice application around Andrea's code and some open source libraries that handle RAW files.
Awesome- let’s see some examples how you use this imaging pipeline and it’s accurate film representation. -
Apologies that this seems to be way over your head and you feel offended by it.
If someone likes the Cobalt profiles and wants to buy them despite the high price tag that's perfectly fine with me. I just think it's a shame that there is no commercially available app that does what this guy has built with Python. E.g. it takes 7 minutes to simulate a 45mp image from the Nikon Z7, clearly there is room to optimize performance a lot.
Apologies that this seems to be way over your head and you feel offended by it.
If someone likes the Cobalt profiles and wants to buy them despite the high price tag that's perfectly fine with me. I just think it's a shame that there is no commercially available app that does what this guy has built with Python. E.g. it takes 7 minutes to simulate a 45mp image from the Nikon Z7, clearly there is room to optimize performance a lot. ...Show more →
Lol-
So what you are saying is you have a bunch of data and can't get any pictures to look decent from all of it.
Who is in over their head here?
Its so funny we are talking about profiles from a company that has clearly done its homework and made it commercially available. Yet here you are saying its hogwash but can't make a decent profile from the smarter, more scientific guy on Github.
Have fun playing with AI and some other guys research!
p.5 #16 · New Cobalt Kodachrome 25 and 64 film profiles
RustyRus wrote:
Lol-
So what you are saying is you have a bunch of data and can't get any pictures to look decent from all of it.
Who is in over their head here?
Its so funny we are talking about profiles from a company that has clearly done its homework and made it commercially available. Yet here you are saying its hogwash but can't make a decent profile from the smarter, more scientific guy on Github.
Have fun playing with AI and some other guys research!
Wow you're really a special type of.. You realize there is no pre-built app for that github code and that I am adding a complete RAW develop module, right How easy do you think that is to do from scratch?
Here's a screenshot from the prototype I built *today* and I'm not a programer. Obviously not everything is working as smoothly yet as I'd want it too.. Sample attached too as a proof of concept.
p.5 #17 · New Cobalt Kodachrome 25 and 64 film profiles
Mark Sisco wrote:
Thanks for sharing these comparisons. I would love to see more Portra v Kodachrome from you or others that include more reds, skin tones, and vibrant colors.
Interesting observation … my Cobalt Portra emulations render blues much more teal than what I see here (M11-P and M10-R files).
p.5 #18 · New Cobalt Kodachrome 25 and 64 film profiles
fjablo wrote:
Wow you're really a special type of.. You realize there is no pre-built app for that github code and that I am adding a complete RAW develop module, right How easy do you think that is to do from scratch?
Here's a screenshot from the prototype I built *today* and I'm not a programer. Obviously not everything is working as smoothly yet as I'd want it too.. Sample attached too as a proof of concept.
Thats great you want to be a programmer in your spare time and build an app with Claude from a GitHub project-
The output I see is just another photo that we can argue about all day long if it actually looks like Porta 400- Is yours better than what Cobalt would have delivered?
Who cares though- I don't want to do anything in your workflow and just want to apply the preset matched to my sensor in LightRoom- That is what this threads about by the way. Sounds like you aren't into it-
p.5 #19 · New Cobalt Kodachrome 25 and 64 film profiles
RustyRus wrote:
Thats great you want to be a programmer in your spare time and build an app with Claude from a GitHub project-
The output I see is just another photo that we can argue about all day long if it actually looks like Porta 400- Is yours better than what Cobalt would have delivered?
Who cares though- I don't want to do anything in your workflow and just want to apply the preset matched to my sensor in LightRoom- That is what this threads about by the way. Sounds like you aren't into it-
Feel free applying mediocre pre-sets to less-than-mediocre images if you're into it. Don't forget to pay a lot for a bunch of marketing bla bla along the way, too!
There was a lot of debate in this thread about whether these profiles are accurate or not. I have laid out why they cannot be accurate from a technical perspective, no matter what the Cobalt guys are writing in their marketing material.
For negative film there is no single point of truth - the paper, the enlarger filter settings, light source for printing etc all play a role. But at least this pipeline is physically correct in the way it replicates the analog process. I shoot quite a bit of film still, including Portra 400 in Venice and it looks pretty good to me. Sadly I had black & white film loaded that morning so I can't offer a 1:1 comparison.
If Cobalt has such great training data available in the form of Kodachrome slides, I'd appreciate if they took this route to build more accurate profiles. Should be easier with slide film as it has more of a fixed target outcome.
p.5 #20 · New Cobalt Kodachrome 25 and 64 film profiles
fjablo wrote:
There was a lot of debate in this thread about whether these profiles are accurate or not. I have laid out why they cannot be accurate from a technical perspective, no matter what the Cobalt guys are writing in their marketing material.
Thanks man- I think we all realize its not the actual film here- We like the profiles. Again sounds like you don't. Why not start a thread about your github guy and your claude process and go crazy with it!
How easy do you think that is to do from scratch? 
