diffuse(albedo) and specular should be rule-based with no artist input/interpretation of the material.
specular is a flat value (unless you have different surface/material types on the same map)
diffuse(albedo) is a mean range based on the albedo of the material type.
roughness/glossiness are based on the surface detail on the material (which can vary quite a lot depending on if it is polished/ scratched/grained etc. but these effects have no inpact on the albedo/specular unless dirt has collected or a different material is underneath (incase of scratches) but then this means you have a different material type and should basically do the same for that and then merge them).
the only real challenge for any artist to create a 'realistic' material now is them understanding how the material they are creating is made and what properties ( the makeup of the material, is this wooden plank a mixture of different woods pulped together, is this gravel made of different types of stone... is this metal painted, what type of paint is it? etc) it needs.
depending on how complex the shaders in the engine are, you may need to understand the Index of reflection/refraction of a material (metal/glass specifically) for a more realistic conservation & Fresnel response. but most engines ignore these and have specific setups for glass and metal which follow the same rules but give a general imitation of that type of materials conservation properties.
if you follow the rules then you will have no issues relating to your material setup, only engine/implementation/lighting issues. There should be no need to 'fudge' a material if the other parts of the scene are setup correctly (Lighting&Post specifically).
metallic != pbr, it's just one approach to material authoring. it should be noted that this method was used before pbr as well.
Exactly. I don't disagree with the problems surrounding the metalness workflow, but the metalness workflow does not = PBR. Issues related to the metanless workflow are not caused by or directly related to PBR.
Physically based rendering is a holistic rendering principal that is not locked to any specific technique or method.
Metalness is a reflectivity input type
Specular is a reflectivity input type
Index of refraction (IOR) is a reflectivity input type
All three could be used in a PBR system, all three could be used in a NPR system or any other rendering system you like.
your statements made it sound like you have experienced more issues than in this single particular project that you cannot show due to NDA.
PBR implementations are advancing, both for offline and online renderers. A benefit is that the assets (calibrated, measured...) are usable with the renderer getting better and improving what they can represent. Pre-PBR the assets were less portable due to lack of standardization. Some assets looking good in a particular lighting condition but not another, or not surviving when advanced HDR effects are used. I disagree that this is a false impression when the industry moves towards this on a grand scale. They wouldn't do it if they weren't getting better quality per dollar overall, or if the previous system had worked better.
Replies
metallic != pbr, it's just one approach to material authoring. it should be noted that this method was used before pbr as well.
specular is a flat value (unless you have different surface/material types on the same map)
diffuse(albedo) is a mean range based on the albedo of the material type.
roughness/glossiness are based on the surface detail on the material (which can vary quite a lot depending on if it is polished/ scratched/grained etc. but these effects have no inpact on the albedo/specular unless dirt has collected or a different material is underneath (incase of scratches) but then this means you have a different material type and should basically do the same for that and then merge them).
the only real challenge for any artist to create a 'realistic' material now is them understanding how the material they are creating is made and what properties ( the makeup of the material, is this wooden plank a mixture of different woods pulped together, is this gravel made of different types of stone... is this metal painted, what type of paint is it? etc) it needs.
depending on how complex the shaders in the engine are, you may need to understand the Index of reflection/refraction of a material (metal/glass specifically) for a more realistic conservation & Fresnel response. but most engines ignore these and have specific setups for glass and metal which follow the same rules but give a general imitation of that type of materials conservation properties.
if you follow the rules then you will have no issues relating to your material setup, only engine/implementation/lighting issues. There should be no need to 'fudge' a material if the other parts of the scene are setup correctly (Lighting&Post specifically).
Exactly. I don't disagree with the problems surrounding the metalness workflow, but the metalness workflow does not = PBR. Issues related to the metanless workflow are not caused by or directly related to PBR.
Physically based rendering is a holistic rendering principal that is not locked to any specific technique or method.
Metalness is a reflectivity input type
Specular is a reflectivity input type
Index of refraction (IOR) is a reflectivity input type
All three could be used in a PBR system, all three could be used in a NPR system or any other rendering system you like.
PBR implementations are advancing, both for offline and online renderers. A benefit is that the assets (calibrated, measured...) are usable with the renderer getting better and improving what they can represent. Pre-PBR the assets were less portable due to lack of standardization. Some assets looking good in a particular lighting condition but not another, or not surviving when advanced HDR effects are used. I disagree that this is a false impression when the industry moves towards this on a grand scale. They wouldn't do it if they weren't getting better quality per dollar overall, or if the previous system had worked better.