I work in a studio where we make 3d visualizations of interiors and pieces of furniture in a white background. I have been struggling for a while now with the quality of my renders. Where I work, I’m the only one using Blender. Two others use 3ds Max + Corona and a third colleague uses 3ds Max + V-ray.
We’ve had a colleague that went to another firm last year, and he too used Blender. Coincidence or not, the top bosses weren’t also fond of his renders, although he clearly achieved best results than me.
I know this question was asked before, namely last year, but something could have changed within this year:
In the aforementioned thread, the majority of users said that Corona was the best software in terms of photorealism.
My boss’s boss said he doesn’t like my lighting. I spoke with my boss, and we arrived to four possible solutions:
Some kind of addon that might improve the quality of my renders (doubtful that it may exist);
Not leaving Blender, and learn Octane;
Improve somehow my renders in Blender, with techniques that I’m not aware of;
Model everything in Blender and export to 3ds Max, rendering there with Corona, as my most skilled colleague does.
It doesn’t matter if the evaluation of quality is somewhat subjective (although, to my eyes, the quality of my coworker is clearly leagues above mine). What’s important is to get as close as possible to the renders of my colleague.
So, the question is: what is my best approach to achieve this kind of results?
Blender 4.0 got an upgrade for the realism of its principled shader, so realism really shouldn’t be the problem it once was (as long as you are using the recent Blender versions). I would say Cycles is technically capable of reaching a level of quality very close to other renderers.
When I compare your renders to the example you linked, I don’t really see a difference in renderer capabilities. I see a difference in the quality of modeling, textures and image composition.
In your renders, the couches look low detail and blobby, do they not have fabric seams, wrinkles? Also their fabric texture seems to have the wrong scale to me. If you are using the newer Blender versions, your fabrics could also benefit from using some of the “sheen” parameter (it simulates fluffy or dusty surfaces and makes fabrics more realistic).
Your shelves and cupboards seem to have low quality wood textures. in some images, it seems to be the image quality on the website blurring the wood fibers, though in some of the images you get to see the wood up close and it doesn’t look realistic. Are you using procedural wood? If yes, you might benefit from photo based textures instead.
Is the roughness channel of materials textured? If not, adding a bit of roughness variation to woods could boost the realism (if you don’t have a roughness texture, you can try passing the color texture through a color ramp and play a bit with it to make one).
Are your objects modeled to real world measuments? Some objects seem to me like they could have mismatched sizes, though maybe you did and they are actually that way.
The color choices. One of your image series has intense blue light coming from the outside and orange light inside. The 2 combine and make the images look somewhat purple.
Some of your images have vividly colored and detailed exteriors that steal the show and make the windows almost more spectacular than the inside.
Finally, the image composition. Let’s analyse the example a bit. When I look at that image, my eyes go directly to the cluster of detail in the foreground, as that little table is a nice island of detail in the picture that seems to tell a story. Then, all the lines and shapes in the image guide my eyes towards the other cluster of detail on the main table. I look at this detail and can guess that the scene is portraying some sort of celebration or party in the evening with multiple guests. The image has a clear intention and is guiding my viewing so I can piece that story together easily.
I don’t see that kind of intention in your images. The detail clusters seems placed a bit randomly. The lines and shapes point my eyes towards empty spots or outside the images. There are often many light sources visible in the image, which are all areas that attract the viewer’s eye, competing with each other and taking the attention away from what you actually want to show. You would probably get a better result with fewer but better used light sources.
I’ve used probably Blender 3.4 or 3.5, but my newer renders with 4.0 still look pretty much the same.
You’re right, this couch lacks detail. My newest models are much more detailed, with fabric seams and wrinkles.
I’m using textures, not procedural wood. Most of the textures I use are 2K, while others are 4K. Unfortunately the site over-compressed the JPEGs, and destroyed what texture detail was there to be seen.
I always create a roughness and a normal map from the diffuse map.
Yes, one of the things we must absolutely make it perfect is to model all the furniture to the milimeter.
The colors have always been a big issue for me. I thought that adding a blue cast to the outside sun light and a yellow cast inside would be a good choice, because they are complementary colors, but I’m probably wrong.
Regarding the detailed exteriors, probably you’re right. However, my boss always asks for photorealistic images placed outside, which I then brighten up in an attempt to simulate an overexposed picture and reduce their detail.
Yes, I agree that I need to improve my composition; do you happen to know any resources that address this issue?
The lights are my main problem, I agree. I often turn on all the interior lights, because my top boss said at the begining that my renders were too dark. However, I do know that an excessive amount of light sources reduce the volume of the scene and take away the attention from what I want to show, like you said.
My immediate superior says that my images look a little plastic, and I must agree with him. The problem is that I don’t know if I can achieve the results of my colleague, even taking into account all the things you said…
Just to clarify I’m not in archviz sector, I’m a VFX compositor for movies and episodics. Learning more about lighting will improve your renders. Cycles isn’t designed for production, yet your work is quite astonishing. My first suggestion is that you have to avoid trying to get the final image straight from renders, retouching or/and compositing is fine tuning, I believe you’re already doing some of that. Adding lens effects will help you to achieve photorealism, depth haze, defocus, chromatic aberrations, lens distortion, lens flares, bloom, etc. Color management in ACES or new AgX will instantly improve your work, I prefer ACES. As this is your work I would invest money and time in a new render engine. Renderman is more difficult to use, but its quality is undeniable, and Pixar has a lot of free tutorials at their website and YouTube. For fast results (3 months of study), switch to ACES and learn more about retouch and compositing.
Those are just suggestions from my experience using Blender in post production.
Maybe look into trying Luxcore and if possible Octane. They both work with Blender. I am not suggesting you switch to them but try experimenting with them in your free time just to see if you prefer them to Cycles.
Luxcore gives very realistic lighting compared to Cycles but its development is not reliable in the foreseable future. You can even try out 3dsmax and Corona since the tools are being paid for by your studio. The most important thing is your bosses are happy with the end result.
Most clients do not care how you go about making your work. What they care about is the end result.
Hi @lightwriter
It’s like in the architectural world - there are two approaches, some like Archicad, others like Autocad. And you won’t convince these people that the second tool is better. It largely depends on the PR of software producers.
Many business owners have a hard time believing that free software can achieve similar results to paid software. I have also encountered such ignorance and disbelief many times that Blender is not able to meet the industry requirements in the field of architectural visualizations.
Of course, it is best to get to know and learn how to use many programs and choose the tool that suits you best.
Of course Corona, VRay are industry standard but that doesn’t mean that you can’t have similar result in Blender Cycles or Blender Octane. Even on this forum you will find dozens of posts from people who show that Blender is a powerful tool and Achieve beautiful results.
If you have already chosen your path with Blender+ Cycles as a tool, it is just a matter of work on your skills. There is rarely time to refine the details in commercial projects. Working with light, composition, color correction - these are things that require some time to explore and practice, practice, practice. You can find in internet many tips for improve skills in those fields.
Maybe You should use some fake light to improve your lightning, create or buy better shaders and study a bit of composition.
Although there is nothing wrong with your render compared to your friend’s render. There are also many aspects that could be criticized in your coleague’s render, e.g. the fact that there is no appropriate focus point, because the curtains seem more important than everything that appears in the foreground.
But remember, there’s also nothing wrong with improving yourself every day, even outside of working hours if you to make renders as a work. And of coure each version of Blender is better and better.
You can use some tips from Chocofur, you can use some knowledge about lightning from CreativeShrimp.
Are you seriously recommending Renderman to an an 3D artist who is doing archviz?
As Melvi said, give Octane a testrun, and maybe test out Corona too.
Despite never having used either of them in a meaningful way I would bet money on the idea that I can get a better picture out of both if rendering an older Blender project I once did in Cycles.
And I think there are many who would place a similar bet on themself or others trying to do the same.
There is a reason why Octane, Corona (and V-Ray) are dominating when it comes to archviz.
I believed I mentioned I’m not in archviz sector, you skiped that part. Yet from my experience working in post production studios I still recommend Renderman for anyone working in CG at professional level. And also Arnold, but there isn’t a Blender version. Those render engines open doors for other roles, considering a career path in CG.
I am assuming here that you are using the principled bsdf for you materials and not some custom material group. If you are using a special setup, you can ignore what follows.
If you were to make the same render in each Blender version, you might notice that non-metallic materials with low roughness look different. Before, the reflections at glancing angles were too bright. This likely affects your shiny counter tops and reflective floors, making them more realistic in the newer versions.
By the way, make sure your materials are set to “multiscatter ggx”. It’s more realistic, especially for glass and high roughness materials and has now been made safer to use (it used to create noise). It’s a bit of a subtle effect in most cases, but it doesn’t hurt to use every realistic option.
Maybe it’s not a matter of brightening them, but rather making them less saturated? Though, I have to comment on the city one. There is no way the outdoor buildings in full daylight would look darker than the inside of the room.
It could in fact be a good idea, though if the colors are too evenly used everywhere, they will mix together and make your image purpleish. Maybe it would work better if the blue from the outside was less saturated? Or if the colors were more confined to separate zones of the image (again, this is the issue with having a bunch of lamps everywhere)?
The thing with brightness and darkness in renders, is that it isn’t accomplished by actually making the image bright or dark. Rather, you want to suggest daytime or night time, to trick the viewer into believing in a specific time of the day.
A “dark” image is done by using artificial light colors (like a night time lighting), by having intense bright zones surrounded by a sea of darkness, like lamps in the night, and by having points and spot lights cast sharp shadows across the scene.
A “bright” image is done by having wide light sources that wrap the scene in gentle gradients, with natural looking colors. Avoid the small lamps, sharp shadows and artificial colors of the night.
If I were making a daytime render, I would start by lighting my scene as well as possible using only the light from the windows. Just try it: in one of your scenes, remove the lamps inside the room and just try increasing the light coming from the window until it gets to a realistic level. I would say go brighter than what you have in your images, a room lit by giant windows would be bright enough that artifical lights are useless because they are barely visible by comparison. If the outdoors need to not be overexposed, Use area lights to boost the window’s contribution (though they would be overexposed in real life).
You will see that using only this one, wide light source gives a rather gentle and pleasing result, which you don’t want to flatten too much by adding other lights.
That sounds like words from someone who doesn’t know how to explain the problem.
It’s hard to tell just from the images, but maybe you have some shading problems? Make sure that every object is set to shade auto-smooth and has the correct edges marked sharp, or else the shading might appear a bit too rounded and bubbly on objects where you don’t intend this to be the case. That’s a common cause of plastic look.
Also, If you have beveled objects that are shaded smooth, you should learn about the “weighted normals” modifier. It fixes the shading of bevels, preventing the rounded shading from leaking on the flat faces.
I can’t point you to a specific ressource, though you might be interested to know I learned a lot of it not from 3D ressources, but rather from photography and drawing classes (I learned 3D animation in a college program and we had photo and drawing classes as part of the course). So you might want to widen your search. Image composition can be learned from more traditional art forms. Or i’m sure there are good tutorials, just don’t limit yourself to Blender for this topic.
In the “film” section of the render settings, you will find the “filter size” setting. This decides how blurry the render is. The default value of 1.5 is slightly on the blurry size. An architectural photo would probably be taken with a high resolution camera, so a smaller value might look better, like 1.1 or 1.2. I am not an expert on this, so if anyone knows what kind of value would be realistic, feel free to correct me.
Beware light clamping, it helps avoid fireflies, but it’s not good for realism. You should put the value as high as you can without getting issues.
For a static, realistic render, you will want to use lots of diffuse / glossy bounces in the render settings. The only reason to use a lower number (like the default settings) is for performance while rendering an animation.
You are using lots of white materials. Beware of making them too bright. The default value of 0.8 is on the bright side even for white paint. It’s plausible, but that’s some really bright white paint. It might be easier to light a scene if it uses slightly lower values.
That one is more of a performance than look thing, but I assume you are using light portals for your windows, right? They are really useful if you are lighting a scene using an HDRI, they help guide the light to the window and find the HDRI through it.
The comparison between Cyclesr and Corona seems to be advantageous to Corona.
If you use V-Ray and Corona in the relevant field and there is a positive evaluation of the results, Cyclesr is unlikely to look attractive.
However, for those who are good at rendering, the difference between the two will not be that big.
And Cyclesr also has difficulty handling nodes related to materials to produce good results.
So, talking about Blender Cycles and Corona Render – they both have their perks. Blender might lack some cool ready-made stuff for archviz, obviously , that’s where your colleague has an upper hand; but now we’ve got more converted libraries from 3ds Max( check blender market, 3dshaker models, imeshh , etc,.
Also Max importer in 4.1 is a game-changer in the making ( quite too early tho) but you can take advantage of that.
But here’s the real deal: it’s not about the software doing the magic. You’ve got to know your way around setting up scenes and the basics of making things look good especially light clamps and bounces. Path guiding , MNEE are all algorithm you may want to look at especially if your boss is a fan of fancy archviz renders.
Note that even the competent max users struggle with engines like V-Ray and Corona if they don’t get the hang of it.
Corona is like tailor-made for archviz, especially for newbies. Yet, if you’re rocking Blender and know your stuff, Cycles can give Corona a run for its money. It’s all about using what you’ve got to get the look you want.
Bottom line – it’s not just about which software is fancier. You need skills, good resources, and to understand the nitty-gritty of each engine. Tweak Cycles, play with post-production, and use the viewport compositor, enable developer extras> enable the experimental compositor> change compositor execution mode from tile to GPU for faster/realtime post processing feedback, also hack glass objects to avoid noisy renders and also don’t depend on denoisers they make renders worse.- that’s some secret for making Blender shine in archviz.
I’m a full-time architectural Visualiser,
So I understand these concerns
Portals are only relevant in one specific scenario: light from the background entering through a small opening. They don’t do anything for light objects and they don’t help if you have directional (sun) light creating a bright spot on the floor, but they are great for solid color backgrounds or the ambient sky light of an HDRI.
As far as I know, nothing that was added in the recent versions should make portals obsolete. Path guiding can help find light coming through openings, but portals outperform it in the specific cases I mentionned.
I’ve never changed the default value. I’ll try a smaller value and see if it does change my renders for the better.
Regarding light clamping, you’re absolutely correct. If I set the value too high the highlights become dull. I’ve learned it the hard way.
Yes, I’m often guilty of setting the diffuse / glossy bounces too low more often than not.
Yeah, I avoid too high values of white, as well as black ones; they’re not realistic, as you say.
Yes, I do indeed use light portals for my windows, but I use Nishita Sky. I stopped using HDRI quite a while ago, because I can’t have the fine control of the direction of the sun with them.
Your suggestions are all great, especially the “weighted normals” modifier.
When I add a bevel modifier, I only use 1 segment, because when the value is 2 or more the rounded shading leaks to the flat faces, as you mentioned.
This is an issue that I also saw with my ex-colleague that used Blender and with one of my coworkers that uses 3ds Max.
Than you for your input. I agree with everything you said. However, I’m inclined in learning now Corona. Not because it’s better (it might be, or not), but if I have some kind of question, I’ll have my coworkers near me, which will be able to aid me in a swift manner.
It largely depends on the PR of software producers.
