SU Podium Forum
Sign up Calendar Latest Topics

  Author   Comment  

Posts: 93
Reply with quote  #1 
I've been working on my skills, techniques, workflow lately and thought I'd try using a picture of a kitchen I liked and modeling it by eye and then rendering through ProWalker GPU. I prefer to light without fake "fill" lights and like my light sources to be actual fixtures as much as possible. My modeling/rendering is focused on being a balance between time/quality. I'm not interested in spending days trying to render a scene just for a client to look at it for 5 minutes and say "OK," or "can we change the (---) to (---)".

Looking for constructive criticism and any helpful tips, taking into account the workflow comments above. Cheers!

- One thing I tend to struggle with is whether to color my LEM's and Omni's in Podium light settings or in post.
- It's looking a little flat to me now that I look at it more.
- I tend to also like to add a line overlay but it's an extra step that can be time consuming in post. Cabinetry can benefit from it here I think. No overlay is in this render.
- ProWalker GPU can look a little muddy in the details. Higher resolution seems to help but still not super crisp. Noticeable around the cabinetry handles in the island.

Lighting Notes:

Ceiling Can's
- Omni  inside (white)+ transparent LEM surface/lens.
    - You can see the omni shadows from the can's on the cabinets. Not sure if I like this or not. Seems to be a limitation of ProWalker GPU where omni shadows can be very hard. I've played around with using spots instead but I just don't know what works best after a lot of trial and error...
- LEM (white) strips just behind the front's.
Exhaust hood
- Spots (used a color just to match what was shown in the picture). Spots used because of the ability to control light spread pattern.
Suspended Ceiling Fixtures:
- 4x omni bulb's (white) per fixture
Fill LEM's
- One behind camera (white- low powered) and one in the room behind the stove)
    - Again, not my preferred method but sine I only modeled part of the house as a scene and not the entire actual house, I needed some "cheat," light.

Post processing is done in Lightroom them combined in Photoshop for the comparison attached.

*Rendered in Pr mode with de-noise ON, 3128x1486 resolution (that's 2x the window size) and 800 samples.




Avatar / Picture

Posts: 11,128
Reply with quote  #2 
That's pretty good actually!

There are a few things I noticed right away though.
  1. The post-processing isn't great. Even without the reference image, you can see that the contrast is too low.
  2. The lighting is not 'right'. You can see that the light from the doorway at the back of the room on the left is over-illuminated, which is causing the refrigerator to look wrong because it's reflecting the  light from that area. The underside of the wall cupboards is over-illuminated. It's brighter than the doors at the front which is not logical and can't be correct. The light output from the bulbs on the hanging lights is too low. They are throwing out a lot more light in the reference image, which is changing the overall illumination. Because their light output is too low, you have added fill lighting which is making the rest of the lighting elsewhere look wrong. Don't use fill lighting unless you absolutely have to. In this scene, you haven't spent the time to really look hard at what the lighting is doing, and try to simulate it. Because of this, you resorted to artificial means to try to compensate, and it hasn't worked particularly well.
  3. The type of recessed ceiling fixture you have used is not a like for like equivalent to the reference image. It looks to me like those have a slightly projecting diffuser, which would explain why these lights don't have the scalloped effect that the ones in the render do. Rebuild this from scratch as a square fitting to match the original, and you'll learn a lot more!
  4. The Material settings on the knobs and handles and on the suspended light fitting are wrong. They are not the same at all.
  5. The modelling is not quite accurate. The proportions of the panels on the doors and the depth of the recesses, the type of handles etc are all not quite right.
  6. The viewpoint in the render is too low.
These types of render are difficult, because you are forced to consider everything in great detail and identify how light and materials settings work in real life. It's the best and possibly the most difficult exercise you can do. Most of the time we do them as learning exercises, exactly as here, to improve our skills.

However, you have cut corners ๐Ÿ˜‰

The point of learning exercises is that get as much out of them as you put in. Every time you are cutting a corner you are denying yourself an opportunity to learn ๐Ÿ˜‰

I look back over almost all my old renders and I can pick out the faults, even when I have tried to do the best job I could. We are constantly improving, and the key is more practice. 

If you have no time or budget constraints for this exercise I would advise you to keep going. Be ruthlessly self-critical. Sweat the small stuff you think might be stupid, and you'll be really surprised what you learn from it [thumb]


That which does not kill us makes us stronger
-Friedrich Nietzsche


Posts: 93
Reply with quote  #3 
Fair points made. Allow me to respond. Keep in mind part of this exercise is to sort out what items to focus on that produce better results- with the investment of time kept at bay.

*Updated render at at the bottom of this post* - 800 samples 3,764 x 1870 resolution -edit in Lightroom

1. Not much headroom in a 1.5MB PNG file, unfortunately. As I noted, details get get muddy quickly if I try to manipulate it too much. I do agree with you that the image is flat and underexposed (more-so due to some lighting intensities in the model that need changing). Fine balance between getting what I would call "too crunchy" (contrast) vs natural- which is to my taste.

- There is most definitely a window in the room behind the stove that is producing light in towards the refrigerator in the reference image. Easily seen in the white light on the cabinets to the right of the refrigerator. My solution was to use a fill light in that room (rectangle LEM) pointed at an angle towards the refrigerator that I've since removed and added a LEM window on a wall instead.
- No fill lights are present except behind the camera and as I noted in the room behind the stove. I've removed those in the latest update.
- I disagree that I didn't take time to look at the actual lighting when trying to simulate it. The can's I had on my to-do but hadn't got around to it yet is the most glaring thing. The suspended fixtures are a work in progress for sure and are flat surfaces instead of tubular frames, which was a stopgap in order to get the to at least match in design somewhat quickly since none were available in the Browser that matched. Again, part of the purpose of this is to keep my eye on time vs result. in that, I need to learn to balance what might give a greater result vs time invested. In that regard, I can better be able to choose what I should be focusing my learning/skills on to be better for my workflow.
- Might be getting some reflective bounce underneath because there is tile and quartz there. One thing I can't sort out is the way the counter on the left's LEM is brighter than the others, when they are the same texture LEM. I had catered to matching that really bright light on the left side because it stood out so much. I think I'd need to make them all individual.

3. I always have struggled with getting a circular can to NOT show the shadow line so harshly. I'll have to dig into using a specific lens later on. In the meantime, I've had the time to model a 4" square recessed fixture with a reflector trim kit. Looks better, I think.

4. Yes, the knobs and handles were slightly different materials. Corrected.

-Yes, it's definitely modeled too long. I need to take out some length in the center island and the center of the cabinets on both sides. Handles are close enough, IMO. Books are different colors, coffee maker is a different brand, oven is different, flowers are the wrong color, floor is too red... if we go down this path. I'm more concerned with lighting first. I can sort those out as I go.
- I design a lot of custom cabinetry and something I'm working on is methods to better show shaker style doors with better shadowing and detail without having to do a line overlay or give it too much geometry. They change WAY too frequently for me to detail out perfectly, especially when a very select few people actually take the time to notice.
     - As an example, for shaker doors and giving cabinet doors some depth, I sometimes rely on a "fake" shadow by using a black color instead of offsetting a door from the frame. Again - Design time vs ...payment. Or more likely, actual deadlines and a workload of sheer scale if one things needs to change, let alone multiples.

6. The view in the render is what I would expect the photographer to take. This is one area where I decided to differ a bit from the picture because whoever took the reference picture is a) not a photographer and b) not good at taking pictures (lol). Architecture photography should have strait vertical lines- even though it bothers me that they don't match, that's the decision I made.

kitchen 2.jpg 


Avatar / Picture

Posts: 11,128
Reply with quote  #4 
This is better, as you can see, removing the fill lighting has produced a much more natural feel to the image.

The absolute best way to deal with interiors is to use the HDR format and presets. They allow much better and more subtle pp.

Although this is better - it depends what you want to achieve. If it's a learning exercise, the whole business of "I didn't have time" doesn't really apply.  For exercises to develop your skill, you really ought to only stop when you feel you have reached the limit of what you can do. If you keep doing this, it will raise the quality of everything you do, because the little improvements will become habits, and you'll find that you do these things automatically, and that there is probably no time penalty. Even the process of intense scrutiny of your work will improve what you do

I remember when I did karate, my teacher would keep shouting at me to straighten my back leg in front stance. I really didn't understand, as far as I was aware, it was as straight as it physically could be. Then he came up to me and pushed my knee back 2 inches with his finger! It could indeed move, but it was now in a position which wasn't comfortable. What I learned was that I literally had to force myself to deal with the discomfort to improve my form. Eventually the discomfort became normal, and I had to deal with more and more discomfort elsewhere to improve. Until I improved my form, I would get nowhere near black belt level.

That's the great thing about traditional Japanese martial arts - it's all about continual focus on the little details to keep striving for perfection. What I learned from that, I tried to apply in other areas of my life, particularly the professional ones.

So if we use this as a principle in your render, you can see that although it's improved, there is still further room for improvement. 

You have remodelled the recessed ceiling lights, but I suspect the output is too high. Normally their light output isn't awesome. The scallops are gone, which matches the source image.

I think most of the light is coming from the pendants. I don't think omni lights are the optimal choice here. You could try using LEMs. You would have to make a kind of diamond or 'double pyramid" shape like we use for the omni placeholder, only with faces, and apply LEM to the faces. LEMs have much softer shadows and more accurate falloff. I think they will produce more realistic shadows.

The cupboard doors in the source image are more glossy. This can make quite a big difference. I don't agree with you about the handles I'm afraid. They are nothing like the originals at all, neither in shape nor material, and if you look at the ones on the far right, you can see just how crude they are. The originals seems to have a copper-like tinge. The round knobs are also slightly larger in the source image. The light fittings also have a similar colour, and are glossy.

If you really want to get closer to the source image (it's always a great thing to attempt) you need to increase the width of the margins around the doors, and make the recess slightly deeper. Also, you need to move the drawer fronts away from the the cabinet faces by a couple of mm. You have done this with the doors, and you can see that it looks much more realistic.

Your cafetiรจre glass settings don't look right either. If this is from the Browser, the material settings are wrong, because the interior surface of 'solid' glass needs to be different to the exterior. The interior needs to have zero reflectivity because there is no such thing as interior reflections in solid materials. The vases on the countertop look right though.

The floor texture isn't a close match. Actually I don't like that material in the Browser. We have a better one which is paler. You could use that and create a variant which is darker. That should be a much better match.

I understand your point entirely about the camera position. What I would say though is that it is only when you duplicate it as closely as possible that you can see how far away you are from the original. When you think it's as good as it can be, by all means improve the viewpoint.

I'm being really picky here. The reason for this is that if you want to be able to produce images that look really realistic, it's the attention to all these subtle little details that makes the difference. I'm not sure if you are aware that almost all the IKEA catalogue and website images are CGI. They look real because they have taken into account far more than the things I have mentioned.

You have produced some good initial renders, but they lack the subtlety I think you would like to be able to add to take quality up a notch. If that's what you want to do, then the sort of thing I have mentioned above will be a good start ๐Ÿ˜‰


That which does not kill us makes us stronger
-Friedrich Nietzsche


Posts: 93
Reply with quote  #5 
I agree that moving cabinet doors away from a the box gives a better depth/contact shadow, but it's not completely accurate to how they're built. I've done that to the uppers but didn't do the base because I wanted to see the difference. Then I simply forgot about it lol.
      - I see this tip often and also see recommendations to raise objects above the floor so they create a "better" contact shadow (subjectively). I design cabinetry for our projects and use those models to create basic schematics for the shop to work with. So if I start cheating geometry in order to achieve something that renders better then I'm creating a new problem. This is part of the balance I'm working on achieving in accuracy vs creating a "perfect" render.

Yes, the glass vase is from the podium browser. I need to be able to rely on this.

I like the wood floor, so I suppose that's subjective. Clarity and Texture in Lightroom gave it a bit more definition in the wood grain for sure, though. Easy to take some saturation out of the texture in SketchUp.

I'm interested in creating my own LEM "omni" lights for better shadow fall-off if it is indeed that much better. But if it's a preferred method, I'd like to see it in the browser. That would be useful for the community, wouldn't it?

Recessed ceiling can lights can absolutely generate very good and bright lighting. The suspended fixtures are are giving the most distinguishing and noticeable lighting shadow patters, though. Most casual observers would gravitate to thinking" wow, the shadows on the ceiling even look the same." Focus on the defining features first, I think.

Glossiness/reflectivity is something I have to play with A LOT in ProWalker. I can never seem to get things to gloss vs reflect the way I want. I think I prefer to not use "blurred reflection" in ProWalker now that you mention it.  I don't use Podium to render, but when I do "blurred reflection" always looks like terrible grain on surfaces.

ProWalker GPU is my preferred render method. So no HDR format. I'd need to render 4-5x window size to get more and more detail. Loads of time there twiddling my thumbs.

To be fair, IKEA designers are using 3DMax, Solidworks, VRay, etc. I still believe SketchUp can get most people 95% there.

Now, back to my original goal. Imagine you have spent this time opining about nooks and cranny's of a render. And let's be honest here- MOST people would not notice such things. Then you present it to the client OR suddenly a revision has been made to the design. TIME is very suddenly not on your side- yet the output quality is expected to be consistent. How do we balance being able to keep up the output quality with the need to stay current and keep a current model like this?  We must find balance and yet still keep our sanity ๐Ÿ˜‰ . So I tend to model objects with the intent to change them later - and as fast as possible.
Previous Topic | Next Topic

Quick Navigation:

Easily create a Forum Website with Website Toolbox.