Add 2025x spec version

[bsdorra]

• Add thin-film interference
• Add translucency color
• Add 25x parameter list
• Update localization files with placeholder

TODO:

• User Guide update
• Rough transmission energy compensation

[proog128]

This is really nice work and energy compensation looks great. Spec-wise I think it still requires a bit of work to flesh out all details and make the spec self-contained again, moreover it would be nice to explain the derivation of the energy compensation in more depth (similar to Section 6.2.1). Let me know if you need another review!

[proog128]

Hmm, I don't understand it :-). What exactly is replaced? How can I read from this that specular and specular tint are not part of the thin film Fresnel? Thin film is rather invasive, and it probably affects Seq. 1 where we introduce the base BRDFs $\text{Metallic}$, $\text{ThinTransparentDielectric}$, $\text{VolumetricTransparentDielectric}$, and $\text{OpaqueDielectric}$.

[proog128]

The math should really be treated like code. It doesn't help if we simplify by leaving out details, redefining symbols (ambigous symbols), introducing unused functions or calling functions with the wrong number of arguments. Text is just comments, the math must give a correct program even without the text.

[proog128]

Ah, maybe I understand now what you mean. Is it the following?

$$ \begin{align} \text{ThinFilmMetallic} &= (1-i) \text{Metallic} + i M_r R \\ \text{ThinFilmOpaqueDielectric} &= (1-i) \text{OpaqueDielectric} + i (B + s M_r R) \\ \text{ThinFilmThinTransparentDielectric} &= (1-i) \text{ThinTransparentDielectric} + i (M_r R + \rho M_t (1 - R)) \\ \text{ThinFilmVolumetricTransparentDielectric} &= (1-i) \text{VolumetricTransparentDielectric} + (M_r R_d + \rho M_t (1 - R_d)) \\ \end{align} $$

This needs to be fleshed out (add all parameters, format it nicely etc.). Moreover, you need to update the definition of $\text{Core}$ (Eq. 13) and $\text{ThinCore}$/$\text{VolumetricCore}$, and everything that refers to it (Sheen, Appendix, ...):

$$ \begin{eqnarray} &\text{ThinFilmCore}(\rho, \rho_l, s, \eta, \rho_s, i, d_i, \eta_i) = \\ &\quad\quad m \text{ThinFilmMetallic}(\rho, i, d_i, \eta_i) \nonumber\\ &\quad\quad +(1-m)(1-t) \text{ThinFilmOpaqueDielectric}(\rho, \rho_l, s, \eta, \rho_s, i, d_i, \eta_i) \nonumber\\ &\quad\quad +(1-m)t \text{ThinFilmTransparentDielectric}(\rho, s, \eta, \rho_s, i, d_i, \eta_i) \nonumber \end{eqnarray} $$

This will cause quite a lot of non-straightforward refactoring work, I suppose, but it will be worth it. At the moment it's really unclear how exactly all those parts interact...

[bsdorra]

Yes, that's what I had in mind. I found the idea of explicitely rewriting all of the core components just too verbose to communicate the simple idea of replacing the microfacet/fresnel product with a weighted sum. From your reaction it becomes clear that it doesn't really work. I'll stick to the formal approach and rework this based on your suggestion. Thanks!

[proog128]

Thank you for this refactoring, I think it's a lot clearer now. Last time I thought it would require a much more invasive refactoring of the spec, but I think you found now a very good combination of keeping the changes local (inside the thin-film section) and describing its global effects.