Kodak 2383 Print

Original (research-backed). See the notes, the notes (Kodak's curve traced and measured), the cinema README, and the Knowledge base.

What it emulates

Kodak Vision 2383 is the colour release-print film most cinema prints are struck on. The camera negative (e.g. Vision3) is low-contrast; when printed to 2383 for projection, contrast and saturation increase and the image takes on 2383's characteristic curve. Deep, cool-leaning shadows and warm, luminous highlights/skin. This "print" rendering, amplified by modern digital grading, is the ubiquitous teal-and-orange cinematic look.

The grade in one line

A contrasty, saturated, teal-shadow / warm-skin theatrical print look, the classic Hollywood colour grade.

Why Classic Negative (not Eterna) here

Unlike the Vision3 negative recipes (Eterna, low contrast, neutral shadows), the 2383 print look needs teal shadows + added contrast + saturation. Classic Negative natively provides teal-leaning shadows and hard tonality, so it reproduces the print characteristic far better than Eterna. The warm WB shift supplies the "orange" side; Classic Neg supplies the "teal."

How it's built

Classic Negative for teal shadows + print contrast. — + Shadow —deepen and enrich (print contrast). Warm WB (—/—) pushes skin/highlights warm for the teal-orange split. Color —+ CCE — give rich, controlled print saturation. Clean fine grain + — keep the projected-print smoothness.

Datasheet: measured, not paraphrased

Kodak's curve for 2383 has been traced from the archived datasheet. Full method, calibration and validation in the notes. The headline figures, all this project's own measurement:

	Measured	Meaning
Gamma (D 0.5 → 2.5)	≈ 3.44	~6 × a camera negative's ≈ 0.55–0.60
D-max	4.09	the projected black
Usable scale (D 0.2 → 3.8)	≈ 4.4 stops	vs ~13 for the Vision3 negative
~~Red vs G/B gamma~~	~~3.76 vs 3.28~~	⚠️ RETRACTED — estimator artefact (why)

Cross-checked independently, and the negative side is now measured too. The original check assumed a camera negative at γ 0.55–0.60. The four Vision3 stocks have since been traced and measure γ 0.52–0.55, slightly below that assumption. Using the measured value: 0.526 × 3.46 = system γ 1.82, landing squarely on the classic ~1.8–2.0 dark-surround target. Both ends of the photochemical chain are now measured from Kodak's own charts rather than one measured and one assumed, and the agreement tightened when the assumption was replaced by data.

Three consequences, each landing on a setting:

• The short, steep scale is the design constraint. 4.4 stops at γ 3.4 is a narrow hard window, that is why the negative is deliberately flat. — + Shadow —+ — reproduces the print side of the chain; — would model the negative instead.
• ⚠The teal-orange split is NOT established as intrinsic, an earlier claim here has been retracted. It rested on the red channel measuring ~15 % steeper, but that ordering flips under a different, equally valid gamma estimator, so it is an artefact of sampling rather than a property of the film (full account). The recipe is unchanged. It targets negative → grade → print, and —/— with CCE Strong deliberately produce the graded look, which was always the design and never depended on the retracted claim.

⚠This section previously claimed the datasheet confirmed "a high-contrast print stock", "a steep tone scale calibrated to LAD" and "deep D-max via intragrain absorbing dyes". All three overstated the source, the datasheet contains the word "contrast" zero times, LAD is a colour-timing reference with no numbers given, and Kodak attributes the absorber dyes to sharpness and halation, not D-max. Corrected in full in the notes.

vs the other cinema recipes

• Vision3 250D / 500T: the negatives. Low contrast, clean, gradeable. This is the print, the finished, contrasty, teal-orange look.
• Related in mood to Mumbai Monsoon, which also uses a Classic-Neg teal/warm split (for rain rather than Hollywood).

Original (research-backed). See the notes, the notes, the cinema README, and validation-methodology.

Status: datasheet-measured. No real-footage reference frame yet, see Why a scan is structurally blocked.

---

What was measured, and how

Source: X-T5/_reference-sources/datasheets/kodak-motion/vision-print-2383-3383.pdf, KODAK VISION Color Print Film 2383 / 3383, Kodak Publication H-1-2383, revised 3-22 (March 2022). Archived in this repo.

The datasheet's Sensitometric Curves chart (p. 4) was rendered at 300 dpi and traced programmatically.

⚠Correction (2026-07-26, later the same day). This paragraph originally said the technique had "failed on the Vision3 negative datasheets" and that "the Vision3 sheets use dashed and overlapping curves on inconsistent templates" That was too broad, and later work in this repo disproved it. The failure was specific to the Vision3 granularity charts, which are dashed and overlapping. The Vision3 sensitometric charts, a different chart on a different page, traced cleanly on all four stocks at 100 % ordering validation. The distinction is per-chart, not per-datasheet.

Calibration (from tick detection, not assumption):

Axis	Ticks found	Scale
Log exposure	7, evenly spaced	128.0 px per 1.0 log E (−3.0 … +3.0)
Density	7, evenly spaced	128.33 px per 1.0 density (0.0 … 6.0)

Tick spacing was uniform to better than 0.3 % on both axes, so the calibration carries no meaningful error.

Trace validation, the checks that had to pass before any number was kept:

Check	Result
Curve ordering B < G < R through the rising band	✅ 93 / 93 columns correct
Visual overlay of trace onto the original chart	✅ all three sigmoids covered cleanly
~~Independent sanity check — system gamma~~	⚠️ WITHDRAWN — the argument was circular; see below

⚠The original "system-gamma cross-check" was CIRCULAR: withdrawn

This section first argued: a camera negative (γ ≈ 0.55–0.60) printed onto a release stock should give a system gamma of about 1.8–2.0 for a dark-surround cinema, and 0.55–0.60 × 3.44 = 1.89–2.07 brackets it, so the trace is confirmed.

That reasoning does not hold. A literature check found that the ~1.8 figure in circulation is ARRI's print-through gamma, which is itself defined as negative × print (≈0.6 × 3). Using it to validate the product of negative × print is validating a quantity against itself. It was never an independent direction.

Worse, the perceptual dark-surround figure, the thing that number was mistaken for. Is lower, not higher: R. W. G. Hunt's The Reproduction of Colour is reported to give ≈1.5 for movies, and Bartleson & Breneman's widely-quoted 1.5 is a ratio between light- and dark-surround preferences, not an absolute system gamma. No SMPTE standard quoting 1.8–2.0 for dark surround was located.

What replaces it, a genuinely independent check that the method passes: the Double-X five-point calibration, where Kodak prints the gamma of five curves and the tracer reproduced all five to mean error 0.016. That is ground truth, and it never needed the system-gamma argument.

And an independent check this figure arguably FAILS. Eastman Kodak's own patent language (US 5,891,607 / EP 0 902 323 A1) gives conventional mid-scale contrast as ≈2.5–3.1 for print films and ≈0.45–0.7 for negative films. The VISION3 negatives measured here land at 0.52–0.55. Comfortably inside Kodak's negative range. But this 2383 figure of ~3.46 sits above Kodak's print range, and the sliding-window variants (3.94–4.57) sit far above it. That is consistent with the estimator over-reading at high gamma, exactly as §3 predicts, and it means the true 2383 gamma is more likely near 3.0 than 3.5. (Patent figures are snippet-level, patents.google.com returns 403 here, so they are corroboration, not a citation this project has read in full.)

---

The numbers

= this project's own measurement from Kodak's published curve. Not quoted from the datasheet, the datasheet publishes no numeric tone-scale data at all.

Quantity	B (yellow dye)	G (magenta)	R (cyan)	Mean
D-min	0.09	0.09	0.09	0.09
D-max	4.09	4.09	4.10	4.09
Gamma, D 0.5 → 2.5	3.28	3.28	3.76	3.44
Gamma, D 0.5 → 3.0	3.44	3.37	3.86	3.55
Steepest tangent (0.30 log E window)	4.64	4.14	4.54	—
Usable scale, D 0.2 → 3.8	4.59 stops	4.46 stops	4.15 stops	4.40 stops

Three findings follow, and each one lands on a recipe setting.

1. Gamma of order 3: the number behind a word the datasheet never uses

The datasheet does not contain the word "contrast" even once (nor "steep", "D-max", or "saturation", all verified by search across the full text). Kodak's own phrasing is "excellent tonal scale" and "rich blacks and neutral highlights" So every previous contrast claim in this recipe's notes was an inference dressed as a citation.

Measured, it is γ of order 3.5, but see the caveats above and in §3: this film sits outside the method's calibrated range, the value moves between 3.46 and 4.57 depending on estimator, and Kodak's own patent range for print stock is 2.5–3.1. The honest statement is therefore "of order 3, roughly six times a camera negative's measured 0.52–0.55", an order-of-magnitude result, which is robust, rather than a precise figure, which is not. Even so, that is a number derived from Kodak's own curve rather than an adjective, and the ratio is the part that matters: the print stock does essentially all the contrast-building in the photochemical chain.

2. The usable scale is only ~4.4 stops: this is the real design constraint

D 0.2 → 3.8 spans just 1.33 log E ≈ 4.4 stops, and the steep middle (D 0.5 → 3.0) is barely 2.3 stops. Against a Vision3 negative's ~13 stops, the print stock is a narrow, hard window.

This, not the colour, is what makes a print emulation hard. The whole photochemical design exists to capture wide and then throw most of it away through a short, steep transfer. An X-T5 emulation has to do the same thing: a contrast-building recipe on a restrained dynamic-range setting, which is exactly what — + Shadow —+ — is. A — version of this recipe would be modelling the negative, not the print.

⚠3. RETRACTED: "the teal-orange split is intrinsic to the stock"

This section originally claimed that 2383's red channel runs ~15 % steeper than green and blue (γ 3.76 vs 3.28); that because a slope difference survives printer balance where a speed offset does not, this puts cyan in the shadows and red in the highlights; and therefore that the teal-orange split is intrinsic to the film.

It has been withdrawn, the same day it was published.

Why. Those figures come from the density-anchored estimator (average gradient between D 0.5 and D 2.5). Re-measured from the identical trace with a sliding-window estimator. Equally valid, and the one calibrated against Double-X's five published gammas, the channel ordering changes:

Estimator	B	G	R	Steepest
Density-anchored D 0.5→2.5	3.32	3.31	3.76	R
Sliding 0.8 log E	4.03	3.71	4.08	R ≈ B, G lowest
Sliding 0.1 log E	4.75	4.32	4.65	B

The ordering is an artefact of where each curve is sampled, not a property of the film. 2383's curves are steep, strongly offset horizontally, and have short straight sections, so the sampling point dominates. (The same test applied to the VISION3 negatives leaves their channel pattern completely intact, their curves are near-parallel. So this is a property of this chart, not a flaw in the tracer.)

Compounding it: at γ ≈ 3.5 this film sits three to four times outside the method's calibrated range (Double-X, γ 0.50–1.05), where a fixed log-E window spans most of the entire density scale and cannot avoid toe and shoulder. Full analysis in film-curve-measurement.md.

Status: whether the teal-orange split is intrinsic to 2383 is NOT ESTABLISHED. Nothing here shows it is, and nothing here shows it is not. It is not measurable from this chart by this method.

What survives, because none of it depends on inter-channel structure: D-max 4.09 · D-min 0.09 · usable scale ≈ 4.4 stops · γ of order 3.5. Direct readings or order-of-magnitude results, stable under every estimator tried.

The recipe is unchanged and undamaged. Its settings were never derived from the retracted claim. It was offered as after-the-fact support for a design that already existed, which is exactly why losing it costs nothing.

The lesson, recorded so it is not re-learned: a difference between channels is only real if it survives a change of estimator. Mean values were robust here; inter-channel structure was not. Both came out of the same trace, and only one deserved to be published.

---

What the datasheet does say, verbatim

Kept separate from the measurement above, because these are Kodak's words rather than numbers:

• "rich blacks and neutral highlights", the headline claim.
• "Improved blacks, and more neutral highlights on projection" Stated twice, once in each benefit column.
• "With the excellent tonal scale, cinematographers can be more creative with lighting and exposure, and still see excellent results."
• Image structure: "This film's excellent sharpness captures the detail in the printing negative for projection onto the largest of theatre screens. Fine-grained emulsions, an ultra-thin layer structure, intragrain absorbing dyes, and superior halation protection contribute to its performance."
• Process ECP-2E. Control strips are Process ECP-2. (Earlier notes in this repo said "ECP-2"; the datasheet specifies ECP-2E.)
• LAD: "To aid in color timing and curve placement, negative originals should be timed relative to the Laboratory Aim Density, (LAD)" Kodak Publication H-61.
• Reciprocity: "Exposure times may range from 1/10 of a second to almost 1/3000 of a second, with little or no change in tone scale." Flat across 4.9 stops of printer speed.
• Base: 120 µm ESTAR polyester, no rem-jet; antihalation from "proprietary solid particle dyes" coated under the emulsion.
• Archival: "less than 10-percent image dye loss, even after several decades."

⚠A correction to this repo's own earlier notes

The previous "Datasheet confirmation" paragraph in the notes claimed the sheet confirmed "a high-contrast projection print stock", "a steep tone scale calibrated to Laboratory Aim Densities", and "deep D-max via intragrain absorbing dyes" All three overstated the source:

Claim	Reality
"high-contrast" / "steep"	The words do not appear in the datasheet. True, but it was an inference. Now measured instead: γ ≈ 3.4.
"tone scale calibrated to LAD"	LAD is a colour-timing and process-control reference, not a description of the tone scale's shape. No LAD numbers appear in the datasheet — it points to Pub. H-61, which this project does not hold.
"deep D-max via intragrain absorbing dyes"	The datasheet attributes the absorber dyes to "increasing sharpness and further reducing halation" and to controlling film speed — not to D-max. Misattributed.

D-max is now measured (4.09) rather than asserted, and the validation ledger entry has been corrected: it previously read "LAD/tone-scale extracted", but LAD was never extracted.

On the apparent conflict with "neutral highlights"

The datasheet says "neutral highlights"; the measurement shows highlights running slightly warm. These are not in conflict. Kodak's phrasing is comparative ("more neutral highlights on projection", i.e. against the predecessor stock), not an absolute neutrality claim. Worth stating, because the phrase reads at first glance like a contradiction of this recipe's warm WB shift.

---

Why a scan is structurally blocked

2383 is a release print stock, so a faithful reference would have to be a projected print or a print scan. Neither is obtainable here:

• Real theatrical footage is not Commons material: it is copyrighted studio content.
• Frame-grabs of released films are not faithful references anyway: they carry the film's own DI grade, which is the very variable this recipe needs to hold constant.

So the achievable evidence ceiling for this recipe is datasheet + measurement, which it now meets. A user-supplied frame, your own footage graded toward a print look, or a print scan. Dropped into references/ would be the only way to go further.

---

What would change this recipe

Only two things, and neither is available yet:

1. A real print reference frame showing the settings miss in a specific, repeatable direction.
2. Kodak Publication H-61 (LAD), which would give the actual aim densities and let the balance point be modelled rather than just the slopes.

No setting is changed on the strength of this measurement, the measurement confirms the existing design rather than contradicting it. Changing settings on theory alone is exactly the unvalidated churn this bank avoids.

---

Measurement performed 2026-07-26 from the archived datasheet. Calibration, trace validation and all derived figures are reproducible from the chart on p. 4 of the source PDF.

Last updated: 2026-07-26