I Spent $1000 to Make Digital Look Like Film

Summary of technological concepts, features, and analysis

• Why different “Roger Deakins looks” happen even with the same camera

  • Films shot by cinematographer Roger Deakins (e.g., Skyfall, Sicario, Prisoners) can look markedly different from each other—even when using an Alexa.
  • The video argues that an overlooked factor is the look-up table (LUT) / transform pipeline used to map digital camera output to the photochemical (film) system, rather than focusing only on the camera capture.

• How Deakins’ LUT approach is described

  • The video claims Deakins’ LUTs are based on empirical characterization of the film system:
    • Characterize negative and print (not merely rely on a manufacturer spec sheet).
    • Measure color patch return across the full chain (negative → print).
    • Build a print film pre-visualization LUT.
  • The LUT is then tested, with the reported result that theater output matches expectations.

• The creator’s project: build a custom film print emulation (≈ $1000)

  • The speaker wanted control over the film-like look without depending on someone else’s proprietary process.
  • They spent about $1000 to build a print emulation derived from measured data, rather than subjective “looks.”

• Core workflow: model the photochemical pipeline using printed film (not just scanned negatives)

  • Instead of scanning the negative and trying to match digital footage to it, the workflow emulates what matters in cinema:
    • In photochemical workflows, the negative requires interpretation via the printing process.
    • The projected look comes from the print, not the negative itself.
  • The speaker contracted Motion Picture Lab to print the negative onto 2383 positive (a specific film stock/print type).
  • They then:
    • Scanned the print
    • Used a spectrometer
    • Combined the scan with Kodak spectral dye density data (from the Kodak data sheet) to reconstruct the print’s spectrum
  • Outcome: thousands of matched samples across the exposure range, with corresponding samples existing on both the:
    • digital side
    • print side

• Color transform / matching tool

  • The speaker uses Color Match to derive/apply the transform based on the sampled dataset.
  • Reported results include:
    • convincing highlight roll-off
    • accurate skin tones
    • nuanced color rendering

• Validation method and claim

  • The video presents a validation frame comparison:
    • Some colors were not included in the training dataset
    • Yet the transformed digital result matches the actual print validation frame
  • This is presented as evidence the emulation generalizes, rather than merely overfitting.

• Color science vs color grading

  • A key takeaway: color science (the underlying foundation) is different from color grading (the finishing layer).
  • The video emphasizes:
    • Color science provides “weight and nuance”
    • Without a solid color-science foundation, grading can’t easily recreate that nuance
  • Historical contrast:
    • Photochemical systems had strong built-in color science
    • Grading was often simpler, relying on limited controls (e.g., adding/subtracting RGB and adjusting exposure up/down)
  • The speaker frames color science as knowledge cinematographers should understand—not only colorists.

• Education and resources

  • The creator offers free classes on color science and profiling film stocks / building transforms.
  • They also reference a 2383 color transform linked in the description.

---

Main speaker(s) / sources

• Main speaker: The video creator/cinematography educator (describing building their own transform and teaching color science).
• Subject/source referenced: Roger Deakins (as the cinematographer whose LUT/approach is discussed).
• Labs / tools referenced: Motion Picture Lab, Color Match, Kodak data sheet, spectrometer, and the 2383 print/emulation workflow.