A softer light for long learning sessions.
Loading your preference…
Light uses Latte. Your choices stay in this browser.
Volume 03 · 16 flexible weeks
Follow light beyond the first bounce.
Build a CPU path tracer by understanding probability, material scattering, light sampling, acceleration, and convergence. Measure error before chasing a faster image.
Open my weekly workbench ↗Explain your estimator, test its geometry, compare noise fairly, and present an original lighting study.
Your sketchbook / 8 chapters
Two weeks per chapter: understand and build, then investigate and finish. Repeat any week.
Turn a direct-light renderer into a carefully specified light-transport project, beginning with one analytic integral.
A scene sketch, a written estimator contract, and a fixed-seed integral experiment with several-seed evidence.
Give samples explicit state, dimensions, seeds, and reproducible ownership before comparing sampling strategies.
An original sampler plot, deterministic seed tests, and an equal-budget estimator comparison.
Implement a bounded diffuse path loop with a matched cosine PDF, correct throughput, and interpretable termination.
A bounded diffuse path loop, a constant-environment test, and a trace of one sample.
Extend the learner renderer with carefully limited specular reflection and dielectric transmission, without confusing delta events with ordinary PDFs.
A material triptych, reflection/refraction tests, and a statement of supported media.
Extend the supplied diffuse/emission renderer with explicit light sampling and consistent PDF measures, then introduce MIS as a tested learner extension.
An area-light study, a visibility test, a PDF calculation, and a bounded MIS extension or worked estimator trace.
Use the owned median-split BVH companion, compare it with a linear oracle, and design meaningful acceleration evidence.
A query-equivalence report, independent edge cases, and one reproducible timing comparison.
Build progressive linear accumulation with reset rules, useful statistics, and honest convergence comparisons.
An accumulation/reset test, a linear-color test, and an honestly labeled convergence sheet.
Package an original bounded CPU light-transport study with reproducible builds, tests, visual evidence, and an explicit contribution statement.
A finished CPU renderer subset, original room image, tests, build notes, and a five-minute technical walkthrough.
Keep your hands on the work
The companion explains its build steps, tested behavior, and remaining learner work. Keep your own contribution clear as the project grows.
