Post your Firefly Renders



  • @Bytefactory3D these are great Firefly renders! To my eye the only real "tell" as to which is which, is that SuperFly has the "edge" in fresnel effect specularity when the surface is close to parallel with the line of sight.


  • Poser Ambassadors

    @anomalaus said in Post your Firefly Renders:

    @Bytefactory3D these are great Firefly renders! To my eye the only real "tell" as to which is which, is that SuperFly has the "edge" in fresnel effect specularity when the surface is close to parallel with the line of sight.

    No it doesn't. That's just how he coded his emulator.



  • @bagginsbill meaning more work on the FF shader and you wouldn't tell them apart?


  • Poser Ambassadors

    @anomalaus said in Post your Firefly Renders:

    @bagginsbill meaning more work on the FF shader and you wouldn't tell them apart?

    No - for that to be true there would need to be no discernible differences.

    What I'm saying is that I disagree with your assertion that the only difference (real "tell") is the fresnel effect at grazing angles.

    1. The difference in Fresnel is the result of the shader configuration and if configured properly FF and SF produce and draw identical Fresnel coefficients

    2. There are other differences that I can clearly see but you somehow dismissed

    So - I am saying that multiple differences exist, the particular difference you identified can actually be removed, but that others remain, and some of those cannot be removed.



  • Neither shader is important. The real problem is that Poser does a TERRIBLE job at representing people, something that should be its' strength. The main problem is it's inability to have muscles span joints and then to have skin flow around the resulting underlying object. The skinnier the model, the more you can't tell it does a terrible job, but try an athlete and it looks terrible. You can even see this in movie animation, where the Hulk is just, well, wrong. Sigh.



  • @bagginsbill that's fair. My admittedly hyperbolic comment was intended as encouragement, rather than critique, as I wouldn't know where to begin to offer advice on improvement of PBR shaders without having seen the nodes (and deferred my own offerings in favour of those who have actually levered their lids off).


  • Poser Ambassadors

    Here are renders, from each engine, of a shiny prop in a furnace (pure white environment sphere).

    There is no discernible difference in the Fresnel effect. However, the reflections differ in edge pixel detail, where SF is largely good, but FF produces artifacts (aliasing) that I can't justify or eliminate, and FF misses some or all of highlights on nearly sharp edges.

    If the prop had a softer reflection, even more differences would manifest. I will let you guess which is which.

    0_1509636308015_FT1.png 0_1509636311618_FT2.png



  • When images are shown side-by-side, a person (with stereo vision) can do a cross-eyed focus —overlaying the two images as one, and then the brain can easily notice each of the the differences between the two images. (They are seen practically animated, and draw attention.)

    This works on anything similar; including typed pages of text, identical (in appearance), but for a few worlds here and there; they stand out immediately.



  • @bagginsbill
    I can see the FF render is on the right, seeing the difference in the specular reflection and the aliasing.

    @Forum
    I can also of course see the differences and imperfections in my own comparison renders below. But as a content creator, I needed something which could allow me use PBR texture sets not only in both render engines, but also backward to at least Poser 9, because I didn't want to leave these customers behind. Going for perfection, or for mathematical correctness, in my case would have led to end up with nothing, because I could achieve neither the one, nor the other.

    As a content creator, what I needed was some solution which would satisfy my clients, telling my content 'looks good' for them. To achieve this, it was very necessary for my to use PBR (because I wanted to use Substance Painter, with all the special effects like surface roughness, scratches and so on), without individual adjustments in the Firefly materials, if possible even without a need to create different materials to load for different render engines.

    So in order to come up with something practicable (in terms of resources, render times, backward compatibility with at least Poser 9, etc) I needed to sacrifice some accuracy and correctness. That is why I called it 'Emulator', and not 'Converter', and I always made it clear it is not mathematically correct, but only intentends to 'come close' and keep the cost (resources and render times) as low as possible the same time.

    I assume that few customers will ever render content in both render engines with identical lighting, in order to compare deviations in the fresnel effect. Mostly what they want from me is an asset which 'looks good' in their perception, and if they are users of older Poser versions and can usually not benefit from using PBR textures, they will even be more grateful for something which 'comes close' and still doesn't grind Firefly to a halt during renders.

    As for fellow content creators, as long as they individually can live with the differences in appearence between a Firefly and a Superfly render in their content, for the advantage of doing the texturing only once for both render engines and all Poser versions 9+, I guess it is also an advantage doing so. So my reference renders are not intended to claim 'there is no difference', but instead 'see the difference and judge for yourself if it can be acceptable enough for you to use it in your content, or not'.

    I am grateful for the discussion, because it shows that basically both sides of the medal have their pro's and cons. And differences in judgement about reference renders basically result from different expectations about goals to achieve. Like I said, i would have been too happy (from the point of view of content creation) if Smith Micro could have come up with some 'upgrade node' like i.e. the P11 Physical Node offered for Poser 9 and Poser 10, but they haven't. I would be too happy if some 'perfect' or 'mathematically correct' shader exists for Firefly to use PBR, provided it is practicable in terms of resources and render times for the 'average Joe's expectations', which I have to respect as a content creator.

    Unfortunately I don't know any and never found one anywhere, so I went ahead and created my own practicable solution, that is the story behind my Firefly PBR renders.



  • @Bytefactory3D for the record I've noticed your emulator before but just haven't had the time to fart around with it. Its cool to see that people still bother with FF. Firefly is not the most perfect renderer but it suits me fine for what I do. I like it and I'll probably be using it for a while to come.

    6_1509646573608_Lali03.jpg 5_1509646573608_xtina03.jpg 4_1509646573608_Sunny days 01.jpg 3_1509646573608_Tactical Lali 02.jpg 2_1509646573608_Tactical Lali 03.jpg 1_1509646573608_Tactical Lali 04.jpg 0_1509646573608_Tactical Lali 05.jpg



  • 0_1510753760662_Nightgaunt.jpg



  • @bagginsbill said in Post your Firefly Renders:

    FF produces artifacts (aliasing) that I can't justify or eliminate, and FF misses some or all of highlights on nearly sharp edges.

    Did you check the "HDRI output" option? If so, uncheck it and try again.



  • SuperFly has a nasty tendency to break skin, especially around the neck edge of a character where it joins the chest, but Firefly doesn't. Weird.



  • @tburzio Got image of that you can show?



  • @eclark1849
    Freak3, no change in pose or lights, Medium Auto FireFly, Medium Detaults Superfly

    0_1510767269405_sfff1.jpeg



  • @tburzio
    Don't quote me on this, but I suspect it has something to do with polygon smoothing. FireFly has polygon smoothing, whereas SuperFly does not. What happens when you subdivide the mesh once, does it improve?



  • @Deecey said in Post your Firefly Renders:

    @tburzio
    Don't quote me on this, but I suspect it has something to do with polygon smoothing. FireFly has polygon smoothing, whereas SuperFly does not. What happens when you subdivide the mesh once, does it improve?

    SuperFly has per-actor polygon smoothing.
    https://forum.smithmicro.com/topic/2432/not-smoothed-polygons-in-superfly/8

    Does skinning method set to Unimesh? If not, and especially if mats have displacement, SuperFly will break the mesh.



  • Interesting information about breaking of a mesh. I'm going to experiment, and see if I can fix things.

    A primary problem is the ability to break skin. There is no reason whatsoever (short of a Fantastic 4 character) you would ever want skin to break. Skin is interesting. You can actually have skin from one area pull on skin in another, such as when you bend over. This is a critical failure in how characters are represented in 3D programs. Skin is more a shrink wrap, not a covering, related somewhat to methods in the cloth room. Oh well, some day. The boffins at SIGGRAPH are working on it... :-)



  • @tburzio said in Post your Firefly Renders:

    Interesting information about breaking of a mesh. I'm going to experiment, and see if I can fix things.

    A "Unified" mesh or a sub-d mesh should correct it. Superfly, aside from the micropoly displacement issue, also suffers from the split-vertice displacement issue, I guess. (Just a guess, but a Unified or Sub-D command forces an "artificial" weld between groups. That's the "V3" series collar group breaking against the chest group in the render due to split-vert displacement. Fix one thing (collar/breast deforms)... bork up another. :) )

    This is a critical failure in how characters are represented in 3D programs. Skin is more a shrink wrap, not a covering, related somewhat to methods in the cloth room. Oh well, some day. The boffins at SIGGRAPH are working on it... :-)

    It can be either a rigging or materials/mapping issue for simple figures. For truly complex production/movie quality figures, it's generally "taken care of", but with very complex thingies.

    Game figures have to be as lightweight as possible. The gamer expects immediate feedback, not a slideshow. In short, standard figures for general animation have to be fairly lightweight. The more processing power we get, the heavier these figures can be without resulting in a "slideshow" effect that's supposed to really be a real-time animation effect...



  • @phdubrov Wow, that's valuable information. Thanks.