You wanted to know how Blitter creates his amazing photoreal renders in Blender, so we contacted the artist to talk about his workflow, discuss lighting, composition, and more.
Getting Started with Blender
I have been using Blender since version 2.80, so I guess it has been around a year and a half or so. I spent a good number of months going through tutorials and just becoming familiar with the interface and features. I have tried most of the common 3D toolsets including Maya, 3ds Max, Cinema4D, and a few others, however, Blender strikes the best balance for me between ease of use and power.
‘Photograph’ realism, to me, differs from the common photorealism term in that the goal of photograph realism is exactly that – to make people believe they are looking at a photograph (i.e., with all the distortions, color variance, pixelation, blur, etc). As with most artists/hobbyists, I long had the desire to model/render beautiful things and so it is no surprise I tried my hand at modeling and rendering the female form. I then eventually ran across Poser (way back, probably 15 years ago) and found it fun to use. I recall trying to composite a render of a female with a photograph of my living room. At the time it was fun (though in retrospect, looked terrible) and I think from there I kept trying to make more and more believable renders that mimicked real photos.
As 3D technology improved over the years, particularly when unbiased renders began to be released, I was able to forgo compositing renders with photos and instead use a fully 3D/rendered workflow (i.e., 3D characters and 3D background/environment). Fast forward to today, I have probably logged 20,000 hours trying to create a truly believable rendered scene that mimics a photograph. I do not always hit the mark, but sometimes I get pretty close.
Setting Up Characters
Unfortunately, as with most things in life, there is no single solution that supports all my needs. I firmly believe that if your goal is to focus on 3D characters then there is no better platform than Daz. I know Daz can get a bad rep from hardcore modelers and animators, but their characters have matured to be incredibly detailed with thousands of Morphs/Shape Keys that truly make it incomparable. I mention this only because Daz is always at the center of my workflow – regardless of the tool I use to actually construct the scene and render it. As such, I exclusively use the Daz 8.1 family of characters as the basis for all my characters.
Today, my general workflow is as follows:
- Modify the base 8.1 characters in Daz Studio to my liking.
- Export the modified character into Blender and use the Sculpt tools to perform additional tweaks. This includes adding some asymmetry to the model to add to the realism.
- Bring the base texture set in from Daz to Blender and perform a lot of manual texturing painting to add bumps, variation of specularity and reflectivity, adding blemishes, color patches, moles, etc., to again mimic how the skin looks on the real human form. I would say this is one of the most critical steps to achieve my realism.
- I then have a vision in mind of how I want the scene to feel, what pose/poses the character will be in.
- I compose the scene, being sure to pay a lot of attention to the environment and lighting – which I discuss more later.
- Choose a camera angle that matches the vision/mood I have in mind. I find that tiling the camera, oddly cropping the figure out of frame, etc., all adds to the realism of the final render.
Also, I should say that I make heavy use of the amazing Diffeomorphic Daz->Blender plug-in by Thomas Larsson – it is a super robust, free tool that is frequently updated and covers the gamut of model, texture, morph imports.
Working on Environments
Environments are extremely important for my work. Much like with my characters, I try to re-use as much quality content as I can. This is a good point to state, even though it is obvious, that I do not possess a strong modeling skillset. I have a good working knowledge of modeling and can create simple items, but modeling has never been my main interest or forte.
Back to environments, I generally look for re-use opportunities (BlenderKit is a go-to for me) and mash-up various items to create the most realistic environment that I can. For me, a key rule of the road here is to ensure a lot of clutter, variation, imperfections, etc. Additionally, I always adjust tables, chairs, etc., to be rotated slightly off any perfect angle; i.e., I never use 90 or 180-degree rotations, instead, I use 93.32 degrees or 175.4 degrees, for example. As for clutter, I add cables, socks, bottles, laundry, and other items liberally to ‘busy up’ the environment – anything to make it look lived-in. Just as with human characters, environments and props vary greatly in quality and realism, so I always look for items with realistic shaders, reflectivity maps, etc.
This one is simple – I use the basic textures/materials that come with the Daz 8.1 characters. These are created by TexturingXYZ and are extremely high-quality. I then tweak and modify these extensively using Blender texture paint tools and/or in Photoshop. Some of my favorite things to do to increase texture realism are to add very faint color blotches all over the body (e.g., adding blues/green and other non-obvious colors to areas such as the chest, head, and legs) and to add tanlines or some form of saturation variety. Shader-wise, I use a pretty simple shader set-up. Nothing magical or secret here. If I had to summarize I’d say the keys are to adjust shaders to have very subtle specular and reflective skin. Don’t overdo it though, in real life, people just are not that shiny.
Skin Shader – UDIM Material for Entire Body (except for eye parts and teeth):
I think that composition is key. It is probably my strongest skill. I do not use reference photos but instead use the ‘Camera to View’ to look through the eye of the camera in the viewport to try interesting, off-kilter angles. I generally go for a selfie or amateur framing type of look. I tend to use a lot of low-to-the-ground angles with some subtle depth of field in the elements in the foreground. That said, a picture of any kind will not be appealing unless it has a natural balance to it. By that I mean, a good mix of items of focus (like the character itself), some empty space (like a bed in the background), and elements being cropped at the edges of the frame. It is hard to explain, and I can not really speak to my method. It is truly just trial and error and asking myself, “Does that look good? Does that feel natural?”.
Lighting and Rendering
This will probably be a let-down, but my lighting technique is embarrassingly simple. That said, it is not from lack of trying highly sophisticated lighting approaches; rather, it just turns out that after 15 years of experimentation that I have landed on a very basic setup that provides the most realistic result. I generally use a single spot or area light with transforms matched to the camera. In this way, it is as if the light was attached to the camera.
This results in a few benefits:
- It naturally emulates the look/feel of a selfie or amateur photo because the light is coming from the camera angle, just like a flash would.
- It equally lights the character such that there are no strong shadows on the body/face, and while this may seem bad, it greatly helps increase the realism.
While it creates a believable feel, it has some downsides as well, namely:
- Given that the sole light is locked to the camera’s transform, any shadows cast by the light are perfectly lined up behind the object casting the shadow, and as such, you cannot see the shadows. Many people have noted this in my renders.
- It is hard to achieve the same level of realism with other lighting situations (e.g., outdoors, lights at angles). I am working on this shortcoming but I am not there yet.
In terms of the Render Settings, I generally set Cycles to around 600-1200 samples. This leaves some grain which I prefer and reduces render time. Depending on the type of photograph I’m trying to emulate, I will do various post-processing. For example, if I want to emulate a really low-quality photo, I will pull the render into Photoshop and save it off as a low-quality JPEG. Also, I like to use Photoshop’s Selective Color to give the Blacks tones a blueish tint.
Rigging and Rendering Time
I have an AMD 3900X and an RTX 2080ti with 11G VRAM. Hoping to upgrade later this year to something with more VRAM as that is becoming an issue (i.e., without using Blender’s Simplify command I cannot even load two standard female characters into a scene). Hard to cite a specific time from concept to render, but once I have a general scene set up, I usually just use it to make a series of renders, changing the posing/camera angles, etc. A typical 1080x1200 render usually takes around 10 minutes or less at ~1000 samples. By far, the most time-consuming part of my process is creating a new character. This usually takes a few weeks before I am happy with the character and the respective rig, morphs, textures, and shaders are in place.
I think male characters have been my challenge for a few reasons. Firstly, not being a modeler, I have relied on Daz male characters as my baseline. Until Genesis 8.1, the male character faces had an artificial look that I just could not fix regardless of how many morphs or tweaks that I made. They always looked like an action hero. With the 8.1 series of characters and texture sets, I think I may have better luck.
The other reason I have found it hard is likely because I am just not as motivated to create a male character; therefore, when it gets hard or if it is not working out, then I tend to move on. This is a character flaw of my own (no pun intended) that I need to work on!
Below is a sneak peek of a new Asian character I'm working on: