In 2015 we took part to the competition for the Varna Regional Library with Architects for Urbanity, a team of Rotterdam-based young architects. However, rather than working directly on the competition images, we helped the team making their own visualizations (and eventually, winning the competition) through consulting and knowledge sharing. We always regretted not having the chance to work on the images from start to finish, so in summer 2017 we decided to undertake this little side project and work on our take of the Varna Regional Library.
In this case study, we’ll illustrate the concept, the decision-making process and the techniques employed in an average production scenario here at The Big Picture.
Ladies and gentlemen, my name is Federico and this is the Varna Library scene breakdown.
Composition and concept
Architecture photography is a major source of inspiration for us: we’re huge fans of the work of Adam Mørk, Hufton+Crow and Fernando Guerra, but this time we followed a much more simple approach. The Varna image is based on a previous render that featured an already great-looking central perspective, so we just lowered the camera height by bringing it to eye level, rearranged some of the composition elements (more on this later) and put some extra detail in the scene.
The most significant change from the original image is the mood. We wanted to do a little personal research and experiment with something that we don’t often get the chance to work on: a really cold urban scene. Initially, we thought to go for a sunny freezing day, and using a palette of colors going from blue to silver.
However, we realized pretty soon that this concept wasn’t going to work: just like in the original, the key feature of the image should have remained the interior of the library. A sunny day would have shifted away the focus from the interiors, due to the heavy reflections on the glass and on the metal mesh facade. We needed artificial lighting, and to use its hue as a key element of the image. Hence, the dark overcast version with a palette going from grey to yellow, with a hint of blue.
With this initial concept, we started gathering references and inspiration and, once again, photography was our primary source: even though we study and admire the work of other archviz artists, we never use other people’s work as reference during production, in order to shape and strengthen our interpretation of reality.
The 3D file
We don’t have a rule of thumb when it comes to the amount of work to put into the 3D model. We usually adopt a case-by-case approach according to the complexity of the project, its features and the time needed to complete the job. For instance, in the past few months we worked on projects in which the greenery was a key feature of the design and a lot of work went into the selection and the placement of the 3D vegetation models. However, for fast, conceptual projects, we prefer a a more post-production oriented approach.
This image was rendered in Corona Renderer, and slightly tweaked with the LightMix feature to dim the intensity of the ground floor hall lighting. The light setup was very basic. The interior is lit up by simple Corona Rectangular lights to simulate ceiling lamps, with a subtle glow given by the Bloom and Glare controls integrated in the Corona Frame Buffer: given the long distance between the camera standpoint and the building, for this one we didn’t create an array of light fixtures. Since we went for an overcast scene, there was really one way to set the environment light: HDRI! To be specific, we used a good old PG Skies HDRI: the 1008 – Cloudy.
The raw render turned out to be somewhat detailed, but still with a lot of space for post-production work. The 3D file comes from a pretty well-made Rhino model imported into 3DS Max, with the only major change from the original geometry being the replacement of furniture blocks with something more detailed and fancy.
It’s probably worth to spend a few words about the bookshelves. For this camera angle, creating arrays of 3D books would have been an unreasonable amount of work. However, since the Rhino file included placeholder planes for the books, we used a simple workaround to speed up the workflow. We collected a few book array textures, ending up only with a handful of variations, we imported the placeholder planes in 3DS Max as separate pieces of geometry and we applied to all the planes a single Corona material, with a MultiTexture node plugged in the Diffuse slot. MultiTexture plugin is simply amazing for randomizing texture on separate geometry: in our case study, we also heavily used the hue and gamma randomization inbuilt features, to obtain several different variations for each texture.
Here comes the fun part! You might have guessed it already: the image was post-processed in Adobe Photoshop 😉
Choosing the right sky is very important in exterior shots: Alex Hogrefe even wrote a best-practices article with some really useful tips to deal with this task: it’s definitely a must-read post and I strongly recommend taking a look at it.
Picking a sky (or building it with photobashing) is a crucial part of a post-production oriented approach and really sets the way and the tone of your image. Choosing the right color, light, and even the right shape and direction of the clouds for your sky can really make the difference. So, our advice is not to stop at the first couple of pretty pictures! Keep going and test your ideas with several options.
When we light up a scene with a Sun we always replace the backplate, but most times we end up switching sky even with an HDRI in place, just because we want to be completely in control with the composition. With time and experience, we learned that when replacing the sky from an HDRI lit scene, it’s very important to bring the photo assets to a similar luminosity value, especially when you have 3D vegetation, to avoid weird artifacts at the edges of the mask.
Back to our image, we blended three different pictures to get our sky right.
We wanted something that looked natural but had some perspective in it. The main layer (1) is the central perspective sky, which complemented very well the camera angle by using the shape of the clouds as not-so-obvious lines leading the eye towards the building.
After tweaking the luminosity of the photo with the Curves adjustment, we added some subtle texture by blending the second layer on top of it (2) in Linear Burn mode (25% opacity), with really low saturation. Finally, we added some horizontal clouds (3) close to the horizon.
This is just a rough description of the process we followed: to get to the final appearance of the sky I assure you there was A LOT of masking and layer adjustments involved!
Sidewalk and street
Getting a good feeling and story for the image foreground was also an important part of this job. Since we had decided to go for a wintery-snowy mood, we started collecting photo assets in order to mask them and paint snow in the scene.
In the Varna Library image, a lot of work went into brushing the snow through layer masks. The foreground snow on the sidewalk, for instance, was painted out of this texture, set in Hard Light blending mode.
This texture, in combination with the blending mode, produced a pretty good result out of the box, without the need to refine too much the mask: all we did was roughly brushing out the sidewalk of the original texture, and putting a Curves adjustment on top to blend the photo with the rest of the image.
We used a slightly different approach for the snow on the asphalt. Sometimes, finding the right texture, with the right angle takes way too much time, and we had to use a bit of lateral thinking to paint the snow trails. Instead of using a photo, we created a simple Color Fill layer filled with a dirty white color and masked it out; then we “revealed” the snow trails by brushing them on the layer mask with a combination of custom brushes, to get exactly what we wanted. We used some custom-made tire mark brushes and the well-known Jonas de Ro brush set.
Once we were happy with the snow marks, we added the bus lane mark from an on-site Google Street View snapshot (sometimes quick and dirty just does the trick!) and some snow stacked on the edges of the street by adjusting with curves and masking the same texture we used for the foreground.
To finish off the photobashing, we painted some extra details on the library roof, the traffic divider and we sprayed some snow on the passing cars: all these tiny things may seem trivial, but a good image is nothing else but the sum of small details 😉
Trees are not just pretty things to add to a scene because the client wants a happy image with a lush landscape. In fact, the Varna image is far from being happy, and the trees are there to help balancing the composition. This is particularly clear from the position of central tree of the middle row. This tree is not equidistant from the other two, but it is slightly shifted to the left of the image: this positioning helps to frame and counterbalance two very prominent features of the building, the stairwell and the elevator core.
Speaking of composition, have you noticed how many renders feature tree branches in the corner? The Varna image makes no exception: it’s a simple way to “frame” the image, and it’s a very popular trick coming from photography. By blocking other parts of the image, the attention of the viewer is brought back to the subject (in this case, the building).
Picking bare trees for the scene was not just a matter of coherence with the mood of the scene, but also a way to avoid interfering with the transparent nature of the building.
People in renderings is a major point of discussion between artists and clients. The general consensus among clients seems to be that images need to be filled with people and life, populated by cheerful crowds of individuals from all social group and ages. Not saying that crowding an image is generally wrong and to be avoided; some projects definitely require this kind of approach.
Truth is that most images don’t need to rely on crowds to inspire and strike a chord in the viewer: they just need a few, accurately selected “actors”. Moreover, just like vegetation, people are a key composition element in the image and, if used correctly, they can increase engagement and enhance the sense of architecture.
This is exactly what we tried to achieve with the Varna image. Libraries are generally not overly crowded, and the weather conditions didn’t really call for a busy exterior: therefore, we decided to place just a few people around the library square and portray an ordinary morning. Furthermore, overcast lighting means no hard shadows on people, so we had to be really careful to pick assets that reflected those conditions.
Placing cutouts with the appropriate lighting (and the appropriate dress code, if we’re placing 2D people) is a very important part of our job, and a cutout that doesn’t match the weather and/or the lighting condition of the scene can easily compromise the final outcome of the image. Once picked and placed into the file, 2D assets also need to be blended and color matched with the rest of the scene.
If people are very distant from the camera, or they are shot from above/below, we usually prefer to use 3D models to save us time in searching for pictures with the right light condition and/or the right angle. However, for this scenario, we wanted to go the extra mile and create a few stories and situations for the interior of the library which could have not been made exclusively with 3D assets. So, we went for a full 2D entourage and we blended the cutouts very quickly: we grouped all the “interior” people, masked them when necessary, and clipped the Raw Reflection pass on top with 80% opacity (in Corona, the Raw Reflection pass is obtained by using a CShading_RawComponent in “Reflect” mode). Finally, we tweaked each cutout individually with the Curves adjustments. We did the same for the people in the front square, of course without clipping the Reflection pass on top.
The girl in the foreground acts as a counterweight to the whole composition and to an otherwise almost void bottom-left corner. We deliberately used a “neutral” actor, which would not draw too much attention from the building. The human eye is quicker to process moving objects, so we avoided placing a still character, as we didn’t want the girl to become the main focus of the image. This intention is also reflected by something else, which has to do with colors… it is not a coincidence that the color palette of the girl’s outfit matches the image!
Since the shot was probably taken in overcast weather conditions as well, the cutout had no hard lights or shadow and the color/light matching process was pretty easy. All we had to do was using a Curves adjustment to clamp some of the highlights and shadows, bring up the blues and paint some soft shadows on the girl’s lumberjack boots.
However, no projected shadows doesn’t mean no shadows at all, especially on the ground, where contact shadows happen. To paint contact shadows, we use a very soft, oval brush (15% opacity) on a new layer. We started by drawing a very subtle hint of shadows beneath the footprint of the character and the more we go closer to the contact point with the ground (in this case, the feet of the girl), the darker we go with the brushstrokes. Finally, we painted a heavier shadow where the feet of the character touch the ground.
Color grading and contrast
Last, but not least, contrast and color grading. When it comes to contrast, we prefer to be completely in charge and be able to fine tune areas and colors. Our main tools are Curves, for local contrast adjustments, and the Black and White adjustment set in Soft Light blending mode, one of the many useful tricks learned at State of Art Academy. When set in Soft Light blending mode (or Overlay/Hard Light, for a stronger effect), the Black and White adjustment naturally contrasts the scene in a soft, natural way, also allowing control of the luminance of each color hue through the sliders. If the blacks in the image turn out to be too strong, we use the Selective Color adjustment layer and tweak the Blacks to make it softer.
Coincidentally, Selective Color is one of our favorite tool for color grading. We absolutely love it: it even surpasses the Color Balance adjustment, since it allows for a complete and more explicit control of the hues in the image. It is complemented very well by the Hue/Saturation adjustment, which also enable us to correct single hues and, unlike Selective Color, can act on color saturation values.
The final step goes through Adobe Lightroom or the Camera Raw plugin in Adobe Photoshop. Processing the final output like a raw photograph opens up several interesting possibilities, such as further color grading options, giving structure to the image, tweaking exposure and contrast. Theoretically, color grading (split toning, hue, saturation and luminance of single colors) could be done entirely in Camera Raw: for the Varna image however, we mainly worked on contrast, structure, shadows/highlights and vignetting. We also added some atmospheric haze directly in Camera Raw through the Dehaze slider in the Effect tab, trading some contrast for extra mood.
And this is the final result! As you can see, playing around in Camera Raw can really improve the final look and feel of an image.