home | download | howto | module docs | git | beta

filmsim: artic's sophisticated spectral analog film simulation

this is an implementation of Andrea Volpato's spektrafilm. for a nice introduction see this post.

this module is best activated by applying the filmsim.pst preset. this will take care of wiring the required input look up table files (shipped with the vkdt installation). if you want to wire yourself, connect data/spectra-em.lut and data/filmsim.lut to the spectra and filmsim connectors, respectively.

for self contained documentation, i'm summarising from arctic's post here.

the true color of film negatives

when researching film, the key takeaway is that the final colors depend heavily on the second stage of the imaging process, whether it's the scanner's color processing or the analog RA4 color reversal printing process. analog printing seemed like the most authentic way to define the look, especially since companies (primarily Kodak) spent decades refining it.

there are nice book chapters on simulating the full analog pipeline of color photography [1,2,3]. film emulsions are quite sophisticated, relying on finely tuned chemistry with silver halides, several dye couplers, and a pinch of magic.

for anyone interested in film manufacturing, check out the series of videos by SmarterEveryDay on Kodak ( How Does Kodak Make Film? series of 3, The Chemistry of Kodak Film, Kodak's Film Quality Control Process).

goal and motivation

the goal is to simulate the entire analog photographic process, from film capture to the final print, using only the datasheets and basic knowledge. to capture the look of products from Kodak and Fujifilm starting from publicly available spectroscopic data. for example, Portra film and its matching paper are designed to deliver subtle hue shifts and perfect contrast for skin tones, while consumer films and paper are more saturated and versatile. how much of these characteristics can we recreate from scratch?

while film simulation LUTs share similar goals, they often lack the flexibility to be fine-tuned. in contrast, a fully physically based pipeline can better reproduce the real-world versatility of the negative plus RA4 printing process by offering adjustable parameters to tailor the final look. naturally, this approach also brings along the inherent limitations of analog photography, so you need to appreciate (or be nostalgic for) the analog process to embrace these constraints.

negative and print exposure

parameters: set process to expose and scan negative to output the negative, ev film and ev paper are the negative exposure and print exposure

here are some test-strips to introduce the capability of the simulation. the overall imaging process is split in two steps: negative and print. two different exposures can be controlled, and color filters in the enlarger can balance the colors of the print. here are virtual scans of Kodak Gold 200 at different exposure compensations of the negative.

the following strips are virtual prints on Fujifilm Crystal Archive Type II at different print exposures (and constant good negative exposure).

raw file taken from this Play Raw Two Taiwanese uncles playing chess, thank you @streetfighter.

grain

parameters: grain size uniform

the simulation builds three sub layers for each channel, imitating modern color negative films where each color layer is composed by 2-3 sublayers with different sensitivity to increase latitude. the stochastic properties of each layer and sublayers are imitated keeping into account that faster layers are more noisy, i.e. they have larger particles.

these above are a few strips of Kodak Portra 400 printed on Kodak Portra Endura with vertical size of 1 mm. The average particle areas of the virtual silver halide particles, then converted in dye clouds, is changed. in first approximation, the area of the particles should be roughly proportional to the ISO. in consumer films particles are in the range 0.2-2 micrometer diameter, i.e. 0.03-3.2 micrometer squared.

here is an example with higher magnification crops with Kodak Portra 400 and Kodak Portra Endura.

print	negative

saturation with DIR couplers

parameter: couplers

the level of saturation of the negatives is controlled via developer inhibitor release couplers (DIR couplers). when substantial density is formed in one layer, DIR couplers are released and can inhibit the formation of density in nearby regions, both in the same layer and nearby layers. the diffusion in nearby layers of DIR couplers produces increased saturation (loss of density on the other channels, i.e. purer colors), also referred as interlayer effects. here is an example using a signatureedits.com raw file, using Fujifilm C200 and Fujifilm Crystal Archive Type II.

the filmsim data

to run this module, you need the filmsim.lut data file. it is shipped with vkdt git and installed by default. the following steps are not necessary, but if you want to create it yourself (and maybe play with different data points), do this:

cd
git clone https://github.com/andreavolpato/spektrafilm
python -m venv spektra
source spektra/bin/activate
cd spektrafilm
pip install -e .
pip install dotmap  # used by mklut-profiles.py, not a spektrafilm dependency
cd src/spektrafilm/data/profiles
wget https://raw.githubusercontent.com/hanatos/vkdt/refs/heads/master/src/pipe/modules/filmsim/mklut-profiles.py
python ./mklut-profiles.py
tar cvJf filmsim.lut.xz filmsim.lut
mv filmsim.lut.xz path/to/vkdt/src/

the compressed file in src/ is the one vkdt ships; make lut unpacks it into bin/data/filmsim.lut, so there is nothing to copy by hand.

and in any case wire an i-lut module with filename data/filmsim.lut to the filmsim input connector.

to update the film stock to new upstream data from spektrafilm, a few steps are necessary:

• the new stock needs to be listed at the top of mklut-profiles.py before running the python script,
• params.ui should list the film and paper entries in the same order as in the script,
• the precomputed white balance values for the enlarger filters have to be computed for the new stock. see the top of wb.h for instructions on how to run the optimiser,
• if the new stock is a positive/reversal film, its direct-scan neutral also has to be fitted into pos_wb.h (same optimiser, see the top of that file); everything else gets a zero row there and is left unfiltered when scanned directly,
• the fit runs against the module defaults, couplers included; it assumes g film at 1.0 and preflash off,
• glare is pinned to 0 in the fit config and defaults to 0 in the module,
• if the total number of film stocks changed, this needs to be reflected in head.glsl, the line const int s_paper_offset = ...; // first paper in data list/lut has to equal len(film_stocks) in mklut-profiles.py (because the papers come right after the films in the same LUT). the film/paper counts in main.c derive from the wb array dimensions, so they follow automatically.

the LUT stores three 256-wide rows per stock, in this order: log sensitivity, dye density, density-curve model. a mono (b/w) stock has its log sensitivity replicated across rgb but its dye density goes in one channel with the other two zeroed, since the spectral integral sums the three dye channels. both film and print paper are developed through the parametric model, so no baked density curve is shipped. The setup shader derives the row addresses from the stock index and row type.

the regenerated filmsim.lut is checked into git as the compressed src/filmsim.lut.xz, see the packing step above.

connectors

• input scene referred linear rec2020 (after the colour module)
• output the exposed, developed, and printed film simulation (or negative)
• filmsim wire data/filmsim.lut with the film data
• spectra wire data/spectra-em.lut, the spectral upsampling table for emission

parameters

this module has a lot of parameters. they are grouped into film options (first block) and print paper options (second block).

• process determine the input and the output of the processing done here: (0) input raw image and output print on paper, (1) input raw image and output virtual negative (or, for a positive/reversal stock, the final scanned slide), (2) input scan of real film negative and output virtual print on paper.
• film the film id in the datafile
• ev film exposure correction when exposing the film
• g film gamma correction for exposing the film, use to adjust dynamic range
• g fast gamma correction for the fast (highlight) density curve sublayer
• g slow gamma correction for the slow (shadow) density curve sublayer
• exhaust developer exhaustion, reduces effective contrast/density in areas of high overall exposure
• hl boost boosts highlights above a threshold, useful in combination with halation.
• paper the paper id in the datafile
• p base scale on paper's base density in the print
• ev paper exposure correction when sensitising the paper, affects shadows more than the film exposure
• g paper gamma correction when sensitising the paper, affects dynamic range and contrast
• g fast p gamma correction for the paper fast density curve sublayer
• g slow p gamma correction for the paper slow density curve sublayer
• p exh paper developer exhaustion, reduces effective contrast/density in areas of high overall paper exposure
• glare veiling glare in the scanner/viewing optics, in percent of the illuminant
• filter c when exposing the print paper, dial in this share of cyan filter. this parameter is automatically filled by neutral optimisation. set to -1 to fill filter cmy with auto white balance weights for the current film and paper. when directly scanning a positive/reversal stock (process 1), this instead filters the scan lamp so the stock's own neutral point comes out grey, always auto-filled the same way (the slider is only shown for the print-negative process; use tune m/tune y to trim it for a direct scan)
• filter m same as filter c, magenta share. this parameter is automatically filled by neutral optimisation
• filter y same as filter c, yellow share. this parameter is automatically filled by neutral optimisation
• tune m fine tune the magenta filter (or, for a directly scanned positive stock, the scan lamp's magenta share). think of this as a red/green tint
• tune y fine tune the yellow filter (or, for a directly scanned positive stock, the scan lamp's yellow share). think of this as a warm/cold white balance temperature
• preflash switch preflashing the paper on or off, lowering maximum luminance and decreasing contrast
• pf ev exposure of the preflash step
• pf m magenta filtration during the preflash step, analogous to tune m
• pf y yellow filtration during the preflash step, analogous to tune y
• couplers switch developer inhibitor release couplers on or off (affects colourfulness and local contrast)
• cp amt amount of developer inhibitor release couplers
• lang r red-channel Langmuir isotherm coefficient for the coupler inhibition curve
• lang g green-channel Langmuir isotherm coefficient for the coupler inhibition curve
• lang b blue-channel Langmuir isotherm coefficient for the coupler inhibition curve
• cp rad radius of influence of the couplers, in micrometres on the negative
• halation switch halation on or off, causing a slight colourful blur around high contrast edges
• radius radius of the halation effect, in micrometres on the negative
• hal amt scale the rgb strength of the halation effect in lockstep
• hal mids midtone protection for halation. this heuristic lets you gradually fade out the effect of halation for darker tones. useful to preserve some extra sharpness outside the highlights. set to 1 for maximum effect, 0 means all tones are affected equally
• hal bnc number of halation light bounces to simulate
• hal dec decay factor applied to the halation contribution of each successive bounce
• scat amt in-emulsion light scatter: how much of the exposure is scattered before halation
• strength the strength of the halation effect per colour channel / layer in the film
• grain switch grain simulation on or off
• size scale the grain size. physical
• uniform scales the stock's own grain uniformity. 1.0 uses the stock value unchanged; higher uniformity suppresses grain in dense areas
• enlarge upsamples the image before exposing the paper, for a bigger print/export: careful with 4x, it requires a lot of memory!
• scan ill colour temperature (K) of the viewing/scanning illuminant. 5000 (D50) is the neutral reference setting
• scne ill colour temperature (K) of the light the scene was shot under, for correcting an unbalanced capture back to neutral. 0 is no correction; below 4000K it is modelled as a blackbody, at or above 4000K as CIE daylight.
• film ill colour temperature (K) to expose the film under instead of its own reference, for simulating a mismatched film/light combo (independent of scne ill). 0 is no effect (shot under the stock's own reference); below 4000K it is modelled as a blackbody, at or above 4000K as CIE daylight.

licence

spektrafilm profile by Andrea Volpato, licensed under CC BY-SA 4.0. redistribution and derivatives must credit the author, link the project (https://github.com/andreavolpato/spektrafilm), preserve this license, and remain CC BY-SA 4.0. modifications must be noted. full text of the license and attribution requirements

citation

if you use these profiles in your work, please cite the spektrafilm project, see CITATION.cff for details.

datasource

these profiles were created by processing raw measurement data from data-sheets and/or scientific papers. original data are property of the respective holders. film/photo-paper: kodak and fujifilm data-sheets, scientific publications, and technical material. reflectance: hisanari otsu, munsell, human skin, forest colors, japan colors. all data publicly available.

modifications

the vkdt datafile lut has been generated from the spektrafilm profile/*json using a half assed python script found here. the implementation has been done in a best effort kind of sense, with some changes for efficiency (for instance the grain model was swapped out), as well as ui and parameters (e.g. scale of enlarger filters). if you find gross differences between spektrafilm and vkdt please let us know, this is probably not intentional.