Deterministic cinema-grade rendering for Blender.
A Python toolkit that turns automotive GLBs into cinema-grade videos — no manual rigging, no Adobe/Substance subscription. Model-Prep, Preflight gates, 6-gate Verify. Reproducible at the byte level.
$ pip install "zer0one-cinema[preflight]==0.2.1"Same GLB + same seed → bit-identical output. K-Means with random_state=0, sorted inputs, no ML in the render path.
Model-Prep automated: wheels detected, origin set, body grouped, ground anchored, materials PBR-normalized.
5 checks + 5 auto-fixes before every full render (saves GPU $). 6-gate CGVF analysis after every render (CI-ready).
GitHub-native, PyPI-installable, GHCR Docker image for RunPod serverless. No proprietary dependencies.
The same 6 CGVF gates verify a cinema-grade automotive render and an abstract Cinema-Web wordmark reveal. No genre-specific code — just a profile switch.

8-second orbit camera, studio set with neon rims, metallic nardo paint, dark wheels. Cycles OptiX 64 samples on RunPod A4000. Profile: night_neon.
case →
13-second scroll-driven sequence: floating primitives materialise into the brand wordmark. EEVEE Next 32 spp. Profile: web_hero.
case →Sprint 19 · v0.2.2 · MIT License · Built in Bogojevo, Serbia.