Draw translucent dice as refracting resin - #376
Merged
Merged
Conversation
The blended (TRANSPARENT) material drew a translucent die as a dusty, faded opaque one: nothing behind it bent and nothing in it deepened. A translucent die is now drawn with Filament's screen-space refraction (SOLID, IOR 1.5, LIT with the clear coat kept), so the felt, its shadow and the dice beside it show through, displaced and tinted, with the colour deeper through the middle than at the edges. The resin is a second material compiled on first use and cached like the opaque one; it replaces the blended variant, which no die needed any more. Refraction requires opaque blending in Filament, and the SUBSURFACE model can neither refract nor carry a clear coat, so it is not used. Printed numbers and artwork transmit nothing and keep the author's roughness, so they stay crisp and opaque. Every resin input is derived from fields a set already writes (translucency, color, size_mm, roughness), so the dice-set format does not change: transmission = translucency, the see-through blur is max(roughness, 0.6 x (1 - translucency)) so low translucency reads milky, thickness is 0.7 of the drawn size, absorption comes from the colour. Opaque surfaces draw the same pixels as before: the only change to their shader is dropping a multiply by a coverage that was always 1. Decision 90 in docs/architecture.md. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Translucency 0.2 and 1.0 (with roughness 0.05) alongside the 0.6 scene, so both ends of the resin material can be judged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On the Pixel 10a a glassy amber die came out near-black and a 0.6 one read as opaque copper, with nothing of the felt visible in either. The screen-space input was not empty: the existing device test already saw 704 pixels of a cream die change between a red and a blue floor, which an empty or black refraction buffer cannot produce. Filament 1.76.1's own evaluateRefraction (in the shader sources libfilamat carries) gives the reason instead: it multiplies the refracted sample by the base colour itself (Ft *= pixel.diffuseColor), and drops the body's diffuse light by the transmission. The resin also passed an absorption worked out from the same colour, so green felt (linear green 0.11) under amber (0.22, and about half again from the absorption) kept a tenth of its green, and that felt is in the die's own shadow. The absorption was not buying depth either: a solid-sphere ray crosses between 0.75 and 1 of the thickness, so it was a uniform second tint. So the absorption is gone and the bare body's base colour is moved towards the square root of the die's colour by as much as passes through: the colour a set writes is what a clear die looks like over a pale table, which is two passes through it, so one pass is the root. Green felt under amber is then olive rather than black. Transmission is now 1 - (1 - translucency)^2 instead of translucency as it stands. The body is lit by the key light while the felt it shows lies in its shadow, so weighed linearly the felt was a few per cent of a 0.6 die. 0.6 now lets 84 % through and 0.2 lets 36 %; both ends stay. A new device test draws a near-white glass d20 over a red and then a blue floor and requires at least 30 % of its body pixels to take each floor's hue, logging the share under dinfinity.startup. Docs: decision 90, "A die you can see into", dice-sets.md translucency. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Step 3 on
feature/realistic-rendering: translucent dice that look like resin.Approach (decision 90): screen-space refraction,
SOLIDtype, IOR 1.5,LITshading with the clear coat kept.SHADING_MODEL_SUBSURFACE, it has no clear coat, and it never shows what is behind the die.translucency,color,size_mmandroughness. On a translucent die, roughness also frosts what is seen through it.First device round: glassy dice came out black. Filament already tints the refracted sample by the base colour, and the material tinted it again through absorption, so the felt kept about a tenth of its colour.
Fixed:
On the Pixel 10a:
FilamentStageTestpasses 13/13.aGlassyDieTakesTheColourOfTheFloorUnderIt: 91 % of a clear die's body pixels take the floor's hue (bar 30 %).Known limit, for the owner's judgement: Filament's shadow maps treat a resin die as opaque, so the felt seen through it is in its own full shadow, and glassy dice look dark.
Docs:
docs/physics-and-rendering.md("A die you can see into"),docs/dice-sets.md, decision 90.Based on #375.
🤖 Generated with Claude Code