IRAY Photorealism?

1484951535468

Comments

  • bluejaunte said:

    DOF is completely variable on a real camera. There is no "off" other than maybe from a point of view of what a photographer would consider beautiful, which is subjective in any case. Or maybe he just meant the blur per se isn't as refined as on a real camera?

    To quote my brother's thoughts: 

    Nice chair.  Although you know that chair is kinda over looked in the image.  I really like the image and the composition, placing the mirror behind her head like a halo around an angel's head.  That said there are things about computer renders that I struggle with.   They're always so close to real.  The closer they get the bigger anomalies stand out.  So what gets me is depth of field.  I know that it's a computer image and not a camera, but since it looks like a camera image I notice the difference.  For example, the cane is blurred and the floor is reasonably blurred but the chair legs are super focused.  Above that her lingerie is sharply focused into the depth of her belly button but her breasts aren't in the same plain of focus.  Then lastly as you put the ring around her head, you'll notice that the eyes have a greater focus then the lips or nose, but then the hat on head is focus.  So for me it reads funny as it feels random.  In photography, sure you can do anything after the photo is taken but there are norms. 

    tl:dr—it's not the amount of blur, it's where it occurs. Apparently, it being beyond saving with correct depth of field wasn't something he cared to mention…a clear oversight on his part. smiley 

  • MasterstrokeMasterstroke Posts: 1,985

    Once apon a time during the 80s ;-)

    Party Girl.jpg
    1648 x 1648 - 941K
  • aaráribel caađoaaráribel caađo Posts: 686
    edited January 2021

    @Masterstroke, that's really good, enough that I can focus on all the tiny things because overall you've nailed so much of it. Looking back at your early posts, you've taken this character so far, it's truly impressive. 

    The main thing I notice that doesn't quit land is the lighting/toning. For the snapshot look I assume you're going for, with one of those cheap flash cameras with a built-in flash or even flash cubes (ah, old photography!), this misses in two ways. First, there's not enough ambient light. The setting doesn't look like a room where the lights are mostly out, so I'd expect a bit more fill light from the room. The second is the light from her face should be more intense, probably even slightly blown out. Cheap flash cameras didn't allow you to change the amount of light you got and close up like this, you're likely to get some blow out. You might try a film LUT or Photoshop action to get the color of ’80s film, too. Chemical film has a distinct look to it. The square format suggests a Polaroid. You should be able to get a free Poloroid LUT or PS action. I don't recall much grain to Polaroids, but if this is a film camera, it's probably shot on ISO 400 film, so you'd get some visible grain, too. (Unless the noise in the image is grain—maybe you don't need noise)

    Then there's some super nitl-picky stuff that only matters if you pay more attention than people normally do. The sweater she's wearing ends abruptly and looks two-dimensional. While it has a boarder at the top, it's thin and doesnt look wide or thick enough. The cable knit sometimes has gaps, and sometimes has black something, yet she doesn't appear to be wearing something under it. I'd commit to having the same transparency between the cable patterns and put something under it if you want to avoid exposure. Of course, there are a lot of tops in the world and I could be totally wrong about this, so if you have a reference, ignore this entirely! 

    I love the George Carlin Whiskey—great little Easter egg! But it looks like they're drinking some kind of creme liquor like Irish cream. Which totally could happen in a real photo (the bottle they poured from is elsewhere), but for trying to trick the eye, I'd go for something that looks more like whiskey. 

    Post edited by aaráribel caađo on
  • charlescharles Posts: 847

    I'm experiment with IES lights to reproduce camera flashes and ambient lights (which is really cool btw.) You can change stuff in tone maping section of render, but honeslty most of that can also be done in post as well.

    78 Retro Filter Add-ons Compatible with Adobe Photoshop (envato.com)

    Cheap subscription to grab some cool stuff.

     

  • MasterstrokeMasterstroke Posts: 1,985
    edited January 2021




    Thanks for the feedback :-)
    So, I tried again.
    What did I do?
    DAZ Studio
    There is only one HDRI map for the main light, that shines through the windows of that room and the camera headlamp , that I cranked up to 4.
    I don't know, I just like that headlamp, that everybody tells you, not to use. ;-)
    Photoshop
    I corrected the dress a little bit with the PS liquifier tool.
    This polaroid film camera look is not made with some photoshop plugin, I just tuned dials in PS, until it looked o.k. to me.
    I added dust to my scene by creating a second layer with a slightly blueish color, and added a point light effect right in the center with more gaussian noise to it.
    The layer opacity is set to Linear Lightning dialed down to 16%.
    Played with brightness, contrast and gamma, until it look right.
    Rectangular selection of the center with a soft selection (100), inversed selection and decreased again the contrast about 25.
    Duplicated and merged layers and applyed lens correction. In this case about -4.
    Added some Polaroid frame and that was pretty much it. ;-)

    80's Party Girl lo.jpg
    1401 x 1600 - 591K
    Post edited by Chohole on
  • The dust works! However, that still feels not Polaroid to me. I took your file and applied a Polaroid 689 LUT to it. Attached is a sample of the skin. To me, it's more convicing with the LUT applied. The LUT I used came from this free pack, They work in any applicaion that can use LUTs (Color Lookup in Photoshop).

    Screen Shot 2021-01-25 at 8.32.31 AM.png
    586 x 612 - 641K
  • MasterstrokeMasterstroke Posts: 1,985
    edited January 2021

    aaráribel caađo said:

    The dust works! However, that still feels not Polaroid to me. I took your file and applied a Polaroid 689 LUT to it. Attached is a sample of the skin. To me, it's more convicing with the LUT applied. The LUT I used came from this free pack, They work in any applicaion that can use LUTs (Color Lookup in Photoshop).

    thanks for the link. Looks very cool. :-)
    So how do you use it?
    What programm do you use for it?
    How does it have to be installed in?

     

    Post edited by Masterstroke on
  • TugpsxTugpsx Posts: 738
    edited January 2021

    @Masterstroke Great job on the composition, Image looks great and I agree with aaráribel caađo assesment. MCasual has some of the best scripts for Daz. Here is a link to a LUT addon for DAZ Studio you can use the LUTS right inside Daz. Also if you are using a reference image there is a Tone Mapper to adjust your image to match the reference,

    Post edited by Tugpsx on
  • MasterstrokeMasterstroke Posts: 1,985

    Thanks everyone, I've figured it out now. Cool stuff :-)

  • The toning to match historical looks is one of my favorite parts of the process. Have fun with it, @Masterstroke.

  • charlescharles Posts: 847
    edited January 2021
  • lilweeplilweep Posts: 2,489
    edited January 2021

    Veins - which channel do maps with veins/blood vessels go on?

    Im aware some of you people dont use translucency maps in the translucency channel, but was thinking that's where they belong?

    (Side note: Most of the G8M maps dont have veins on them, except ironically the G8 base male map actually has some nice veins.  Are there any good maps for veins?  I could just Multiply the G8 base male maps onto other char's maps if need be, but kinda lazy.)

    Post edited by lilweep on
  • With fresh eyes, what do you think gives it away? I tried my best with the socks but left one stil doesn't quite fit and casts a shadow. Morphing doesnt help beyond this point so I suppose next step is geometry manipulation. And then his face is perhaps a little bit too perfect. I wouldn't know how to improve skin rendering so for now it's just default everything. I'm new here but I had a go at this with Poser 10 years or so ago. And boy it takes ages to render anything with my PC that is, well, still the same. GPU rendering doesn't work, it seems support for my card was dropped a short while ago. At least my CPU was one of the better ones at that time.

    Out of the box colors came a bit too saturated so I applied a LUT which helped a lot but I wanted a black and white picture anyway.

    I hope it's OK to post a picture like this, checking the guidelines I believe it passes.

    thomas portrait environment DOF LUT BW.jpg
    1000 x 1500 - 773K
  • j cadej cade Posts: 2,310
    edited January 2021

    alright I have a hairless unedited test render comparison for you fine folks, sometimes its good to stop and take stock if the things you're doing are actually improving or you're just getting obsessed with random irrelevant minutia.

    each image is the same material in 2 different lighting situations

    a

    b

    they're obviously different but is one more realistic or are they simply different with neither more obviously real looking?

     

    (side note one thing i definitely do like about the new shader is you can bring it into shader mixer and add things without breaking the autohide features)

    m81compl2.jpg
    1400 x 840 - 694K
    m81compl1.jpg
    1400 x 840 - 690K
    Post edited by j cade on
  • j cadej cade Posts: 2,310

    naprejfx said:

    With fresh eyes, what do you think gives it away? I tried my best with the socks but left one stil doesn't quite fit and casts a shadow. Morphing doesnt help beyond this point so I suppose next step is geometry manipulation. And then his face is perhaps a little bit too perfect. I wouldn't know how to improve skin rendering so for now it's just default everything. I'm new here but I had a go at this with Poser 10 years or so ago. And boy it takes ages to render anything with my PC that is, well, still the same. GPU rendering doesn't work, it seems support for my card was dropped a short while ago. At least my CPU was one of the better ones at that time.

    Out of the box colors came a bit too saturated so I applied a LUT which helped a lot but I wanted a black and white picture anyway.

    I hope it's OK to post a picture like this, checking the guidelines I believe it passes.

    Shoulder bend is what I notice - but theyre an absolute pain to get right without morphing

    on the much easier thing that I thing would improve thing the hair looks pure black while nothing else is including the shadowed area on the chair and others very few things tend to be pure black in reality if it were a bit lighter I think it would blend better

     

    on the positive he really looks like hes sitting on the chairtheres a very good sense of solidity there

  • nonesuch00nonesuch00 Posts: 18,131

    j cade said:

    alright I have a hairless unedited test render comparison for you fine folks, sometimes its good to stop and take stock if the things you're doing are actually improving or you're just getting obsessed with random irrelevant minutia.

    each image is the same material in 2 different lighting situations

    a

     

    b

     

    they're obviously different but is one more realistic or are they simply different with neither more obviously real looking?

     

    (side note one thing i definitely do like about the new shader is you can bring it into shader mixer and add things without breaking the autohide features)

    Personally I think the side lit ones look more realistic but maybe it's more accurate to say I like the 'look of the renders' better if I study them too much they are realistically the same. It's mostly down to the lighting as the evenly lit from above in nature is a situation that one is almost never in.

  • j cadej cade Posts: 2,310
    edited January 2021

    nonesuch00 said:

    j cade said:

    alright I have a hairless unedited test render comparison for you fine folks, sometimes its good to stop and take stock if the things you're doing are actually improving or you're just getting obsessed with random irrelevant minutia.

    each image is the same material in 2 different lighting situations

    a

     

    b

     

    they're obviously different but is one more realistic or are they simply different with neither more obviously real looking?

     

    (side note one thing i definitely do like about the new shader is you can bring it into shader mixer and add things without breaking the autohide features)

    Personally I think the side lit ones look more realistic but maybe it's more accurate to say I like the 'look of the renders' better if I study them too much they are realistically the same. It's mostly down to the lighting as the evenly lit from above in nature is a situation that one is almost never in.

    Soory may have been a bit unclear I meant the other way - shader option a vs b 

    I stuck the 2 lighting options because as you say the side lit one is more nice but the front lighting shows off the difference in shaders better

    Post edited by j cade on
  • nonesuch00nonesuch00 Posts: 18,131

    j cade said:

    nonesuch00 said:

    j cade said:

    alright I have a hairless unedited test render comparison for you fine folks, sometimes its good to stop and take stock if the things you're doing are actually improving or you're just getting obsessed with random irrelevant minutia.

    each image is the same material in 2 different lighting situations

    a

     

    b

     

    they're obviously different but is one more realistic or are they simply different with neither more obviously real looking?

     

    (side note one thing i definitely do like about the new shader is you can bring it into shader mixer and add things without breaking the autohide features)

    Personally I think the side lit ones look more realistic but maybe it's more accurate to say I like the 'look of the renders' better if I study them too much they are realistically the same. It's mostly down to the lighting as the evenly lit from above in nature is a situation that one is almost never in.

    Soory may have been a bit unclear I meant the other way - shader option a vs b 

    I stuck the 2 lighting options because as you say the side lit one is more nice but the front lighting shows off the difference in shaders better

    Oh, sorry!

    In that case a) is more realistic. b) almost looks like watercolor (very detailed watercolors be watercolors). Although I could see b) occuring if the person were perspiring alot, it sort of gives me that feeling when I switch back & forth between a) & b).

  • a and b are so close, it's hard to tell which is more photoreal without looking closer, but b had better shadows, which for me suggests realism. That's particularly true with the flat ring lighting on the left—the nose shadows on a are too hard for such flat, broad lighting. Assumign b is the new version, it's a small but significant advancement. 

    More significantly, it suggests flat ring lighting is a poor choice for CGI. I suspect it's because we so rarely see such "clean" environments, so shadows are rarely so throughly banished and unform in what remains. For what it's worth, when I do studio lighting, I use an HDRI turned way down (like 0.25) as fill to get some randomness in the fill lighting. But I also never use ring lights (I don't like the eye reflection almost as much as I like contrasty lighting), so it might be that trick falls apart with a ring light.

  • lilweeplilweep Posts: 2,489

    j cade said:

    alright I have a hairless unedited test render comparison for you fine folks, sometimes its good to stop and take stock if the things you're doing are actually improving or you're just getting obsessed with random irrelevant minutia.

    each image is the same material in 2 different lighting situations

    a

     

    b

     

    they're obviously different but is one more realistic or are they simply different with neither more obviously real looking?

     

    (side note one thing i definitely do like about the new shader is you can bring it into shader mixer and add things without breaking the autohide features)

    i like a-right and b-left.  sorry.

    (while youre at it can you add an opacity channel and missing SSS settings to the shader?)

  • ebergerlyebergerly Posts: 3,255

    Not sure if this fits in with the previous 50 pages of this thread, but for what it's worth regarding realism:

    One thing I've found is that if I want something to look realistic like a photo, a very helpful exercise is to actually take a photo of something, then see if I can duplicate it exactly in a render. Very enlightening and a great learning experience, IMO, and surprisingly something I rarely see others mention. 

    Makes you realize all the little aspects of realism that we often miss, like shadows and dirt and dust and reflections and haze and dents and non-sharp edges and skin translucency and wrinkles and so on. Sounds obvious, but if you actually have to go thru the exercise and match a real photo, it's almost guaranteed you'll have many "oh yeah, never thought about that" moments. 

    As much as I want to believe I obviously know what real life/photorealistic looks like, I realized long ago I'm kidding myself. There's always those many details that are present in the real world that we forget to add to our renders.

     

  • lilweeplilweep Posts: 2,489

    ebergerly said:

    Not sure if this fits in with the previous 50 pages of this thread, but for what it's worth regarding realism:

    One thing I've found is that if I want something to look realistic like a photo, a very helpful exercise is to actually take a photo of something, then see if I can duplicate it exactly in a render. Very enlightening and a great learning experience, IMO, and surprisingly something I rarely see others mention. 

    Makes you realize all the little aspects of realism that we often miss, like shadows and dirt and dust and reflections and haze and dents and non-sharp edges and skin translucency and wrinkles and so on. Sounds obvious, but if you actually have to go thru the exercise and match a real photo, it's almost guaranteed you'll have many "oh yeah, never thought about that" moments. 

    As much as I want to believe I obviously know what real life/photorealistic looks like, I realized long ago I'm kidding myself. There's always those many details that are present in the real world that we forget to add to our renders.

    using reference photos isnt exactly revolutionary. i just assumed everyone did that.

    Then again, lots of Daz content creators really phone it in when texturing/modelling things. Various content in store just looks like the platonic idea of something, but not like anything that would exist in real life.

  • After many attempts, I've given up on this new skin shader with DS4.15. J cade, if it can be edited in shader mixer, can scatter and transmit be put into it? That is the only way I'd give this shader a try again. I do not use scatter only in any of my setups. I use scatter and transmit with tinting and sometimes I use diffuse overlay as well. If you can put that into this thing and post the shader, I'd mess with it again.

  • aaráribel caađo said:

    Hey y'all smart explorers of CGI photorealism, I just sent a render to my brother, who is a painter and photographer with a great eye. The thing that stood out for him in the render was the depth of field, which he felt was off. 

    Render of woman with top hat and large cane in front of mirror

    (This is a cropped version to comply with nudity regulations. The original, which shows the bottom of the cane and chair is link removed.)

    I don't take or examine photographs often enough to have a real sense of this, but his criticism doesn't feel off. Which lead to the question: how accurate is Iray lens simulation, and what can be done to make it more accurate? For this image, I matched the sensor width with the sensor width on the camera I was trying to emulate (a Mamiya 645 medium format camera) and a lense length to a real lens (the 150mm f/2.8), but I have no idea how to accurately set my depth of field to match that lens. The camer F-stop settings don't feel representative to me (It's usually way more shallow than I believe would be true).

    This is kind of rambling, but I'd love to get your thoughts on the camera settings side of this. (Not that it matters since the tone mapping settings are disconnected from the focus settings, but I always render out to EXR and ignore tone-mapping altogether. My monochrome conversions are done with film emulation LUTS.)

    I'm a Commercial and Fine Art photographer, 3D is just my fun side thing I've been doing since the early 2000's. There is some issue with how DAZ calculates, even with the 35mm camera package. The thing a lot of 3D artists need to understand about DoF is that focal length, f/stop and distance of subject from the camera, all matter in how shallow the DoF may be. If you're shooting at f/4, with a 50mm lens, and the subject is 6 feet from you, it's not going to have a narrow plane. However, if you're shooting with the same set up and same settings, but the subject is 1 ft from you, it's going to be more narrow and it will fall off pretty quickly. It quickly changes when you shoot wide open (like f/1.2) on a longer lens such as a 200mm, but that also depends on camera in relation to the model.Another thing to keep in mind is how far the background elements are from the model as well. Here's a quick example I've shot to demonstrate the shallow DoF and fore/background elements. I shot this with a longer(105mm) lens. I can give you actual examples from my own work if you need, just shoot me a DM since I don't think I can post my website.

    From what I *can* see, it doesn't look weird, honestly. If you're not shooting wide, and you shoot more stopped down(like f/8) all the time, a more shallow image is going to look weird. What happens when you shoot(or even render) with a wider f/stop, you're going to have separation from the subject and the background. Where your focal point lies, the fore and background will gradually become more 'blurry', aka bokeh. You see her hand and the stick are oof(out of focus), she is in focus, and then the back of the couch is starting to become oof. With it being a 150mm at f/2.8, it may need to be more shallow, however,that honestly is all relative to where the camera is and the distance to the model, and how far the background elements are from the model. 

    Now, translating this to the 3D world is and isn't trickly. The calculation that DAZ does is not 100% correct, but it's *close* to it. What also needs to be brought up is that some people like the separation that a shallow DoF gives, and others like it all stopped down so you see verything in focus. It's all a personal preference, and whatever you choose is awesome for your work.

  • latexluv13_65ebcf7300 said:

    After many attempts, I've given up on this new skin shader with DS4.15. J cade, if it can be edited in shader mixer, can scatter and transmit be put into it? That is the only way I'd give this shader a try again. I do not use scatter only in any of my setups. I use scatter and transmit with tinting and sometimes I use diffuse overlay as well. If you can put that into this thing and post the shader, I'd mess with it again.

    Should be. I'm a bit in the same situation. Tried to bring back Top coat Fresnel/Top coat bump mode/Top coat bump I need, but all failures so far lol (looked fine for me, not sure where I fail but I'm not used at all with the mixer). With a bit of luck some cool skin makers will bring back their favorites missing parameters into PBRS.

  • j cadej cade Posts: 2,310

    Do we have confirmation anywhere that the new shader isn't doing scatter and transmit? it could be the same as chromatic sss where there isnt an "option" for it, because it is the only option. Disirregardless it would be a difficult one to bring back is its within a node - (technically you can also adjust things within a node, but that requires messing with the mdl which is waaaaaaaaaaaaaaaaaayyyyyyyyy beyond my capabilities)

     

    On the other hand bringing back the old topcoat should definitely be doable as its its own node and not some settings within a node

     

    (also diffuse overlay functionality is still there just renamed to makeup)

     

    and yes shader mixer tab is painful to work with, even just navigating around is a struggle. I use blender's nodes all the time so I have the advantage of being very familiar with nodes conceptually - but also the pain of knowing how well nodes can work

  • I messed around again with the new shader yesterday and today, and I can say for sure that the transluscency is Scatter only, not Scatter and Transmit. I'm managing a render right now with it set at .76 but then the other problem is how they've set up dual lobe specularity. This shader needs improvement.

  • PadonePadone Posts: 3,700
    edited February 2021

    @jcade @latexluv There's now some support to import pbrskin to blender with diffeo, only basic features for now: translucency, volume, dual lobe, top coat. But that's enough to import Victoria 8.1. The new shader seems the same as chromatic sss in the volume section. At least that's how I converted it.

    https://bitbucket.org/Diffeomorphic/import_daz/issues/349/first-support-for-genesis-81-skin-bsdf

    Post edited by Padone on
  • j cade said:

    Do we have confirmation anywhere that the new shader isn't doing scatter and transmit?

    Confirmation I don't know, but I think it is the case (hence the removal of reflectance tint).

    <--- PBRS | Uber/scatter only --->

    image

     

     

    11111.png
    1920 x 1080 - 1M
  • takezo_3001takezo_3001 Posts: 1,979
    edited February 2021

    j cade said:

    Do we have confirmation anywhere that the new shader isn't doing scatter and transmit? it could be the same as chromatic sss where there isnt an "option" for it, because it is the only option. Disirregardless it would be a difficult one to bring back is its within a node - (technically you can also adjust things within a node, but that requires messing with the mdl which is waaaaaaaaaaaaaaaaaayyyyyyyyy beyond my capabilities)

    Iray settings on the left, PBR settings on the right, both have translucency at 100% and set to white... I also rendered the same figures with diffuse maps in the translucency node with the SSS settings with the color pink, I also rendered with the default and specular rendering mode...

    NOTE: They're dry as I don't have the specular turned on... also, chromatic mode is turned on by default for the PBR shader!wink

    IRAY ON THE LEFT, PBR ON THE RIGHT.png
    1705 x 962 - 1M
    IRAY SETTINGS.png
    813 x 1348 - 149K
    PBR SETTINGS.png
    800 x 1353 - 141K
    DEFAULT LIGHTING.png
    1704 x 956 - 2M
    SPECULAR LIGHTING.png
    1696 x 952 - 2M
    Post edited by takezo_3001 on
Sign In or Register to comment.