Re-creating Disney's sodium vapor process [video]

Original Article ↗ Hacker News Discussion ↗

Why the sodium vapor process faded

  • Required a custom dichroic beamsplitter prism; only a few were ever made and the tech was proprietary with low “bus factor.”
  • Setup was complex: two cameras, special lamps, a dedicated screen, and very controlled lighting to avoid sodium spill onto actors.
  • Couldn’t easily be used outdoors or in full-body shots where the ground would also need to be the special screen.
  • Traditional optical matte work was even more expensive; sodium vapor was a cost-saving step then, but still more demanding than later chroma-key.
  • Studios ultimately favored cheaper, more flexible green/blue screen workflows plus manual post-production labor, which scales with underpaid VFX work rather than specialized hardware.

Technical characteristics and challenges

  • Core trick: narrow-band sodium light (~589 nm) and a beamsplitter/filter that isolates that band to generate a high-quality matte.
  • Low-pressure sodium is extremely narrowband; that makes filtering easy but demands strict control of any 589 nm contamination.
  • LEDs are much broader spectrally than sodium; suitable notch filters and dichroic coatings exist today but were harder/rarer then.
  • Precise alignment and timing between two cameras is important; genlock is noted as effectively required for moving subjects.

Comparison with chroma key and digital keying

  • Sodium vapor excels with translucent materials, motion blur, and fine edges; chroma key often struggles here.
  • However, chroma key:
    • Needs only one camera and normal lighting gear.
    • Is more tolerant of color spill (even if that spill is often still visible).
    • Is supported by advanced digital keyers (e.g., IBK) that handle thousands of shots in big productions.
  • Some commenters feel the video overstates current keying pain points relative to high-end tools.

Alternative spectral approaches (IR, UV, other bands)

  • Several people speculate about doing a similar trick in near-infrared or near-UV:
    • Pros proposed: full-spectrum visible lighting, easier and cheaper broad IR filters, cheap NIR LEDs, and less visible on-set contamination.
    • Cons raised: natural IR from some sources, safety and intensity concerns, chromatic aberration and lens mismatch, and some plastics being opaque in NIR.
  • No clear evidence in the thread that such IR-based systems are actually used; status is “unclear.”

Historical and cultural notes

  • Discussion branches into Disney’s broader technical innovation (multiplane camera, early rotoscoping, Xerox animation) and how some expertise was later “forgotten” or became opaque.
  • Sodium streetlighting is compared visually to the process; people recall its eerie monochrome feel and note its role in light-pollution control near observatories and wildlife areas.

Current experimentation and AI

  • Some experimenters report using band-notch filters at other frequencies for machine vision tasks, achieving near-perfect object isolation.
  • Idea raised: high-quality sodium-style mattes could improve training data for AI-based keying models, which currently inherit artifacts from green-screen-based datasets.