60 product records. 68 references. 14 approximate overlays; no verified six-hue measurement sets.
What the evidence supports
Reported, not reproduced
The video reports saturation decreases of 21% for Color '92, 18% for Portra 400 and 16% for original Phoenix. ColorPlus's 8% and Lomography 400's 8.2% have unresolved signs. UltraMax's spoken +7.15% conflicts with the ranking's unsigned presentation. Source comparison, 28:52.
Overlay retained, measurements withheld
The comparison shapes remain useful, so the 14 original overlays are retained as approximate video transcriptions. They are stored separately from measurements, excluded from calculations, and visibly labeled. Their coordinates still lack patch-level extraction records; decimal places do not establish confidence or error bounds.
Specifications are not color measurements
Characteristic curves describe density against log exposure. Spectral sensitivity describes wavelength response. Neither is a finished sRGB palette: development, illumination, dyes, scanner response, inversion and grading all matter. Portra 400 technical data.
Brand and generation matter
The tested Candido 800T is separate from CineStill 800T. Original Phoenix is separate from Phoenix II; Color '92 is separate from Sun-kissed. WOLFEN NC500 is ISO 400, despite its name. NC500 manufacturer sheet.
Calculation rules
Relative saturation change
100 * (S_film / S_reference - 1). Undefined when reference saturation is zero; unstable near zero. This is not the same as a percentage-point difference.
Aggregation
The mean of patch ratios is generally not the ratio of mean saturations. An unweighted average of six hue labels cannot reproduce an undocumented aggregate of chart patches. No aggregate is calculated from the retired profiles.
Hue direction
A hue error wraps at 360 degrees: 359 to 1 degrees is a 2-degree change, not 358. Fixed-axis saturation plots discard this dimension and cannot establish color accuracy.
Density
D = -log10(T). Density is logarithmic transmission, not an RGB channel value or a saturation multiplier.
Color-reference accuracy
The red, yellow, green, cyan, blue and magenta interface colors are category labels, not measured SpyderCHECKR swatches. A ratio-one baseline is a mathematical reference, not a calibrated physical target. HSL is not perceptually uniform. W3C color specification.
Use the correct target edition and patch IDs, documented reference values and illuminant, a color-managed scan pipeline, and matched comparison conditions. Evaluate lightness, chroma and hue, with a validated color-difference implementation such as CIEDE2000 where appropriate. A screen capture cannot establish a film's spectral response. Datacolor target guide / Colour implementations.
What is still needed for independent verification?
Original per-patch reference and film measurements, with patch IDs and aggregation formula.
Target edition, lamp spectrum, exposure selection, film batch and processing records.
Camera/scanner characterization, inversion settings, working color space and normalization steps.
Repeated captures and processing runs to estimate variability, plus a documented digitization procedure if source graphs are used.
The creator's project page offers project resources, but no reproducible raw-patch table was located. Presets and modeled profiles must not be relabeled as independent measurements.
Reference library
68 references
Dates record research access, not publication or retail availability. Historical sheets may list obsolete formats. Source counts include derived datasets and multiple documents per film, not independent experiments.
2025 catalog-number transition. Record backing generation and processing; not every current VISION3 roll can be assumed to carry the old rem-jet backing.
Indexed first-party page; direct fetch failed. Covers Chrome 100, Elektra 100, Aurora 400, Apollo 18 and Psychedelic 400. Product names do not imply separate base emulsions.
Manufacturer-authored reference table hosted by a third party. Confirm patch identifiers, target edition, illuminant and encoding before numerical use.
Datacolor; hosted by InterfotoResearch / third-party host
Spectral film/print/scan model with digitized and modeled data. Some dye spectra are approximations. Model output is not a reproduction of PARVEC's experiment.
Derived profile collection, including monochrome and printing materials. Not an independent experimental dataset; inspect each profile's source metadata and license.