Measured Blue (量出來的藍) — Under the Microscope 顯微鏡下 (Lemo-Opuscar) · LemoLab × Claude Opus 5.5
No AI-generated images, no stock footage. Every pixel of glass and tissue is one real whole-slide scan and the
scanner's own photo of that slide; every line, label and number is drawn in code from the file's metadata.
IMAGERY
- Ventana-1.bif — human liver, trichrome stain, brightfield, scanned on a VENTANA DP 200
(XMP ScanRes 0.25 um/px at level 0; pyramid levels labelled mag=40/20/10/5/2.5/1.25 in their ImageDescription;
the scanner's photo of the whole slide, TIFF page "Label_Image", 21.167 um/px from its XResolution tag).
Credit: PathAI. Licence: CC0-1.0 (public-domain dedication), per the OpenSlide test-data index
https://openslide.cs.cmu.edu/download/openslide-testdata/Ventana/index.yaml
Downloaded 2026-10-16 from https://openslide.cs.cmu.edu/download/openslide-testdata/Ventana/Ventana-1.bif
sha256 2eb05cd8f628c36408c3fde09beeeebaaae90181c4f840d5475ec7594bdcbbb6
Hosted by the OpenSlide project (Carnegie Mellon University).
Modifications (tools/plates.py): crops; one shared colour grade (glass normalised to a warm white, +12 % density,
+10 % chroma); a 0.3 mm alpha feather at the scan edge; on the slide photo: cut out of the scanner tray (tray bars
replaced by the neighbouring frosted glass, tray clips at the far end filled with the glass of the same row),
line-sensor streaks removed column by column, and the slide maker's printed logo between label and cover glass
filled with the surrounding frosted glass. The slide's own label (specimen ID, barcode) is shown as photographed.
MEASUREMENTS (all computed here, from the file; method and values in FACTS.md)
- Scale bar, field of view: level-0 pixel size 0.25 um/px. Magnification readout: the "mag=" of the level on screen.
- Tissue length 4.80 mm, collagen shares 13.5 % / 32.6 % in two 1 mm circles, nucleus 8.5 um.
FONTS (SIL Open Font License 1.1, licence texts in fonts/; Chinese fonts subset with the Google Fonts API &text=)
- Noto Sans SC — Google / Adobe (labels, subtitles)
- Noto Serif SC — Google / Adobe (title)
- Barlow Condensed — The Barlow Project Authors (readings, English subtitle line)
- IBM Plex Mono — IBM (instruments, pixel counts, credit line)
VOICE
- Kokoro-82M (hexgrad), voice af_heart, Apache License 2.0; run locally via kokoro-onnx (MIT) through the
library's core/tts/tts.py. Speech check: faster-whisper small.en (MIT; OpenAI Whisper weights, MIT).
MUSIC AND SOUND (original score and foley, made in code for this film: mix.py)
- Sampled instruments through core/audio/sampler.py, all CC0 1.0:
vibraphone, vibraphone_hard, vibraphone_bowed, marimba, glockenspiel — Versilian Community Sample Library (VCSL)
https://github.com/sgossner/VCSL
glass (water glasses) — FreePats project https://freepats.zenvoid.org
timpani, bass_drum — VS Chamber Orchestra: Community Edition (Versilian Studios / Sam Gossner)
https://github.com/sgossner/VSCO-2-CE
gran_cassa, sus_cymbal — VCSL
- Synthesized in numpy (no samples): pulse bass, objective turret click and servo, pen on glass, scan noise and
counter ticks, focal sweep, whooshes, riser, room tone.
SOFTWARE
- Rendering: the library's core/render (headless Chromium, canvas 2D), ffmpeg / x264, core/render/mux.sh.
- Slide reading: tifffile + imagecodecs (BSD), numpy, scipy, Pillow.