ReBlender: Caustics, GPU Path Guiding & Nested Dielectrics have all arrived.

Product reviewed

A paid custom build of Blender 5.2 (developer: Light Wrangler) that adds major rendering upgrades directly into Cycles core, focusing on:

• GPU-accelerated path tracing / path guiding
• Native priority-based nested dielectrics
• Advanced caustics via VCM (Vertex Merge and Merge) with bidirectional techniques

---

Key features (and what they do)

1) GPU-based path guiding / path tracing controls

• Implements GPU-accelerated path tracing improvements to reduce noise and speed up convergence.
• Learns where light comes from and directs rays toward likely light paths, instead of relying only on random sampling.

UI behavior

• In official Blender 5.2, path guiding is CPU-only.
• In this custom build, path guiding remains available in GPU mode.
• Includes a new sampling method: path tracing and bidirectional (bidirectional discussed later).

Observed results in demo scenes

• Simple room with two window lights:

  • Turning on path guiding → faster, less noisy, and brighter viewport.
  • Comparison claims:
    • Non-path-guided: ~1.5 minutes and “quite a lot of noise”
    • Path-guided: same render time, “significantly cleaner”
    • Example comparison: 1550 samples vs 850 samples while showing clear improvement, implying better efficiency at fewer samples.

• Soap bars / nested glass-like refraction (subsurface scattering inside refraction):

  • Path-guided renders show cleaner results in the same time, improving denoiser-friendliness.

• Interiors (indirect light / walls / bathroom cabinets):

  • Examples show less noise and cleaner cabinets.
  • Claimed to create a “better dataset to pass through the denoiser in less render time.”

Limitation mentioned

• If light sources are wide, the advantage is less pronounced.

Pros

• Less noise and faster convergence in challenging lighting.
• Works in GPU mode (unlike official Blender 5.2 path guiding).

Cons

• Not every scene benefits equally (wide/soft sources can reduce impact).

---

2) Native priority nested dielectrics (architectural update, not a hack)

Problem addressed

• Standard Cycles assumes refractive index transitions via air (IOR = 1.0) when refracting through refractive surfaces.
• This can break physics for real cases like glass ↔ liquid transitions (e.g., perfume bottles, ice in whiskey).

Before this build

• Accurate embedded/nested dielectrics required tedious geometry hacks:
  • splitting surfaces, removing overlaps, or manually setting artificial relative IORs.

New approach in this build

• Uses priority-based nested dielectric tracing:

  • Model liquid as slightly overlapping glass.
  • Assign IOR per material naturally and set priority integers.
  • Example priorities (from the video):
    • Glass priority 7
    • Liquid priority 2 (less than 7)
    • Ice cubes priority 4
    • Bubbles priority 3 (with additional logic to avoid showing bubbles inside ice)

• The tracer maintains a priority stack:

  • When transitioning between overlapping volumes, it decides which medium is “active” based on priority.
• Supports higher-priority volumes inside lower-priority ones (e.g., liquid → ice, liquid → bubbles).
• Can stop certain effects when blocked by higher priority.

Pros

• Easier setup (no internal geometry cleanup).
• More physically correct refraction between complex overlapping media.
• Enables placing volumes (e.g., scatter/volume effects) inside fluids while respecting priority boundaries.

Cons / tradeoffs mentioned

• Requires careful priority assignment so nested objects render correctly (e.g., preventing bubbles inside ice by giving ice higher priority).

---

3) Powerful caustics: VCM (Vertex Merge and Merge) + bidirectional path techniques

Goal

• Produce clear, bright, fast caustics, which are hard in standard Cycles.

Context

• While path guiding improves caustics versus random sampling, clear caustics were still “out of reach” for complex transparent cases.
• This build adds advanced caustics solvers based on research.

What’s added

• VCM (Vertex Connection/Vertex Merge) combining:
  • bidirectional path tracing
  • photonic mapping principles
  • vertex merging in 3D space

Claims

• Clear caustics with “a fraction of samples and time”
• Supports caustics across rough, glossy, and metallic materials, not just diffuse targets (unlike traditional photonic mapping limitations).

Caustics workflow modes mentioned

• Photonic mapping mode

  • Uses “photon caustics” with parameter like photons per pixel (lower = lower resolution).
  • Increasing photon density improves detail but takes longer.

• Bidirectional mode

  • Caustics appear “instantly” after switching.
  • Main tuning concept: blend radius (default 0.5):
    • Lower (e.g., 0.25) → coarser caustics
    • Higher (e.g., 1) → sharper/high-resolution caustics but slower rendering
  • Advantage stated: no need to tweak “photon variables.”

• VCM / Vertex merging extension

  • Vertex merging can better handle complex interiors.
  • Tradeoffs:
    • May require more memory
    • Some cases (example given: nested subsurface scattering) may render worse

Demonstrated comparisons

• French press interior caustics

  • Bidirectional alone: less light detail preserved inside (e.g., “inside the glass,” “inside the spirals”).
  • VCM with vertex merging: preserves/captures more caustic detail inside the complex glass interior.

• Swimming pool with deformed/wave surface

  • Bidirectional barely handles caustics through the water from above.
  • Vertex merging immediately generates causticity.

Performance note

• Tests on an old MacBook Pro (M1 Pro) without hardware-accelerated ray tracing:
  • “Everything works pretty well,” described as fast.
• On newer hardware with hardware acceleration, caustics described as “incredibly fast.”

---

Overall pros & cons (from the video)

Overall Pros

• Big quality + speed improvements for difficult lighting via GPU path guiding
• Eliminates nested dielectric setup pain with native priority-based overlapping refractive media
• Strong caustics solution with VCM, improving clarity and performance, with better complex-interior handling than basic bidirectional

Overall Cons / limitations

• Path guiding benefits can be scene-dependent (wide light sources may reduce gains)
• VCM vertex merging:
  • can be more memory-intensive
  • may produce worse results in certain complex setups (e.g., nested subsurface scattering)
• Requires understanding/choosing priorities for nested refractive objects

---

Comparisons made

• Official Blender 5.2 vs this build

  • Official: path guiding in CPU-only mode
  • This build: path guiding available in GPU mode

• Path guiding vs no path guiding

  • Clear noise reduction and faster convergence (example: 1550 vs 850 samples)

• Bidirectional caustics vs VCM/vertex merging

  • VCM preserves more interior caustic detail and can handle complex scenarios better

• Photonic mapping vs bidirectional/VCM

  • Photonic mapping can work, but may be slower; bidirectional/VCM described as more consistently usable

---

Verdict / recommendation

Recommended if you want significantly improved rendering in Cycles for:

• noisy/bottleneck scenes (better GPU path guiding)
• accurate nested glass/liquid/refraction setups without geometry hacks
• high-quality caustics, especially in complex transparent interiors

The build appears to be worth it for artists focused on realistic light transport, with main “gotchas” being scene dependence and the memory/performance tradeoffs of advanced caustics options.

---

Speakers / contributions

• Presented primarily by Leonard Altman (Light Wrangler) as the developer implementing the technologies.
• A narrator/reviewer explains UI behavior, demos, comparisons, and setup steps (no separate additional speakers explicitly introduced in subtitles).