A portrait is not just an image that is technically correct, it is an image that accurately portrays the subject too. In this tutorial we will show you how to create photorealistic 3D portraits using Maya, Mudbox, and Photoshop.
I will explain the importance of using good quality photos as reference and understanding the anatomy of the body, as well as the composition and lighting. Additionally, we will cover some methods of using nCloth, creating shaders and teach you how to use cameras.
Creating interesting portraits is a long process that requires a lot of back and forth. A realistic portrait requires patience and a good eye. It is important that you take the time to do each step correctly and ensure that if something doesn’t look right, you go back and fix it until it looks right.
Step 01 – Get inspired
In order to create a realistic portrait, you need to start by finding your inspiration. Our inspiration for this tutorial comes from the Renaissance art movement, from painters such as de Goya, Rembrandt and Caravaggio. These masters were able to see the world in a way that others did not. They were able to capture their vision in artwork that, centuries later, still leaves viewers marvelling. Begin by gathering images that speak to you and try to understand what it is specifically about them that you like. Is it the lighting? The colours? The composition? The overall mood? Ask yourself, why was this piece created? Why did the artist use certain colours, angles, contrasts and brush strokes? Once you find an image, always refer back to it, but don’t try to copy it.
Step 02 – Reference photos
Once you have chosen your subject, you will need some good photo references for the modelling and texturing process. We have two different photoshoots, one for the pose and one for the texturing. Set up a space with white walls and plenty of light. Choose a natural pose for your subject, something that also brings out their character. Afterwards, take photos for the textures. This requires good lighting, no shadows and no hotspots – you want the subject to be as diffuse as possible. You will need a full rotation, one shot every 30 degrees, with the subject in a straightforward pose. You will also need photos of the subject looking up, left, right and down; close up of the eyes, ear and hands; and one with the eyes closed. Working with a high-focus length creates less perspective, which can be easier to work with when sculpting. Once you have downloaded the photos, work with the RAW files. Adjust values as needed to bring out the details and remove some of the shadows.
Step 03 – Image planes and base mesh
Choose the master photograph, the one that has the right pose and expression. This will be your main reference throughout the process. Check the focus length under the metadata of the image. Always ensure that your camera in Maya, Mudbox, or ZBrush matches the focus length of the photograph, otherwise the perspective will be off and it will be impossible to get a correct representation. In Maya, create a camera, fix the focus length to match the photograph and import that photo as an image plane. We are working with a pre-existing base mesh that has been set up with a simple rig for quick posing. The model has been matched as closely as possible to the photo, then the camera has been locked so it cannot be changed. From this point on, this photo and camera can be the control variable. You can change the camera later once you are working on the composition while creating the model. Now you need a solid base to work from.
Step 04 – Block out and add cameras
Once you have blocked out your model in reference to the photo, you will need to bring in more cameras to double check your proportions and angles. Try to get the head lined up as closely as possible. It may be hard to match the cameras up perfectly, so go back and change them once in a while. This requires a lot of back and forth between changing the model and the camera angles (though these are never locked down). These cameras are going to be the cameras you import into Mudbox or ZBrush, and the Maya camera and Mudbox camera will be in sync. The hardest part about making a portrait is that you can’t work in real-time with the final render. You can sculpt but you will not have a true representation of what the final render will actually look like. Having cameras synced throughout each program will enable you to take test renders at anytime during the process and compare them directly to the photographs that you have as reference. You want your final render to reflect the original photo and the only way to do this is to make sure that each step along the way is a perfect representation.
Step 05 – Start the model
Beginning the process of working on the model can be challenging, even with good photo reference. You have to be patient and pay attention to every detail and knowing the anatomy for modelling is a must. Do not guess at anatomy. Make sure that every part of your model is intentional: every bump, every wrinkle, every muscle and every fold must make sense. There are many books and courses on anatomy out there and even if you think you know it, there is always more to learn. Start by blocking in the overall shape in Mudbox. Keeping in line with the photo reference, go through each camera, matching it closely. It is easy to get caught up in the details at this point, but try to keep working on the model as a whole. Do not work in sections at this point. Focus on making the proportions correct and placing the eyes in the right spot by looking at the photo reference and sculpting in the anatomy. There will be a lot of information that is hard to understand – use your knowledge of anatomy to fill in any unknown parts. Anatomy remains the same for everyone just with changes in proportion. Once you have blocked most of your model, start working in smaller and smaller sections, leaving fine details until the last moment. Never, at any point, lose focus of the overall shape.
Step 06 – Match the eyes
The eyes are the heart of a portrait. The slightest change to the eyes will make the subject’s expression change. Take the time to properly block out the anatomy and match it up with the photos. Be warned that no matter how much time you spend working on the eyes at this point, you will have to come back and change them once you start texturing and lighting. Beginning at a very early age, humans learn the subtleties of facial expression, much of which revolves around the eyes. Even someone who has not studied art or anatomy will notice right away if they are wrong. When creating a portrait you should spend the most time working on the eyes since this is what really brings a model to life.
Step 07 – Create clothes
For the clothes, start by making them on top of the T-posed original base mesh. Build them to a low-resolution mesh that you are happy with. Then make a blend shape from the posed character model to the T-pose model, and wrap the clothes onto the body and turn off the blend shape. The clothes follow the mesh into the posed shape. With a little clean up, the clothes are a good base for adding the details. This process can be easier than making the clothes asymmetrical on the posed model. You can add a cloth simulation on the clothes using nCloth, which is pretty good for getting basic shapes and forms in, but not for a high level of detail. When sculpting the clothes pay attention to the form of the body underneath. Even though you will not see the body itself in the final render, it is important that you know its form since the muscles might create stretches in the fabric. To create a highly realistic model it is essential that the clothes interact with the body, which is what causes the shapes, the folds and the stretching in the fabric.
Step 08 – Work on the details
At this point the model should be coming together and it is time to start adding the small details. Detail is what creates a true feast for the eyes. Even if the viewer is not actively aware of observing particular features they will notice if there is a lack of detail, since it’s the detail which adds that extra layer of realism. Be careful not to overdo it however, as you may end up drawing attention away from what you really want the viewer to focus on. In our portraiture work, the main focus is on the eyes and face. Everything else, like the detailing and the background, are meant to showcase and support these elements. If something is distracting and takes attention away from what is important, then get rid of it.
Step 09 – Texture by hand painting
Texturing involves projecting the photographs you took onto the model. We have used Mudbox to project the images. All the cameras that were previously lined up with the model during the modelling stages can now be used to project the images onto the model. Do your best to make sure that there aren’t any stretch marks and use as many cameras and images as needed to check this. It will take time to make this look right and it will require a lot of hand painting to make it seamless, too. Do not paint in textures of any shadows or highlights as you will want the skin to be only diffuse at this point. You may have to go back and change the UVs as they may have been laid out incorrectly the first time round. At this point, you will be glad for all the work that you did earlier on to ensure that you had a good photo reference before starting.
Step 10 – Move on to texture maps
To create the other texture maps you need to use the Diffuse map you painted. Start by desaturating the Diffuse map to create a base for the Spec map. You will have to add some white to the wet parts of the face, such as the lips and tear ducts, to create a higher spec. Look closely at your photo references to see where exactly the white is needed on the face, then create the Gloss map using the same process as above. To create the Bump map you can create a High Pass in Photoshop. Do these maps quick and dirty to begin with to get to the process of the test rendering stage as soon as possible – these maps will change no matter how much work you put into them at this stage so don’t worry too much. Note that you may need to change the Diffuse map many times throughout the process, which means that these maps must also be altered to match. Once the Diffuse map is finalised you can then start really polishing out the others.
Step 11 – Create the hair
The hair is definitely a challenge and will take time. There are many ways to make hair and you will have to try different methods until you find the one that works best for you. The way we’ve done it is by creating NURB curves following the photo reference, duplicating each one and having complete control over each. Then convert these to nHair and add a V-Ray hair shader. Next, adjust each hair separately to match the photo. This is a tedious task, but it is important that the hair perfectly matches the photo reference of the subject. Creating hair for all places where there is hair on the subject (like the eyebrows, arms, beards and armpits, for example), will yield the most realistic results.
Step 12 – Use shaders
Once you have brought your textures to a comfortable level start to add them to the shaders. For the skin shader we used the VRayFastSSS2. Start with the pink Preset setting and add maps to the Diffuse, Subsurface colour, Spec, Gloss and Bump. We don’t need a Displacement map as we are rendering the hi-res object. We are constantly changing the model and need to make quick changes so when creating shaders, take a test render before adding another map to see exactly what you are doing. It may be handy to place the shader onto a sphere and render it next to the model. Sometimes the details on the model can cause confusion as to what is actually happening and doing this just simplifies it.
Step 13 – Light with V-Ray
For this portrait, we used VRay Dome Light with an HDR, a key light with a plane in front to create the shadow on the subject’s face, a fill light from the right, and a light to fill the background. It is important to watch for any areas where the light gets too hot as this makes the image appear less realistic. The lighting is key to creating the mood of the piece. The difference between a soft fall-off light and a hard light can change the mood entirely. The work of Caravaggio is a perfect example of how a hard light can change an image. Caravaggio’s work is high contrast, which makes it more dynamic, whereas Rembrandt uses a softer light to give his work warmth. Test out different styles of lighting from a variety of angles. Save a new file each time you do a render so that you can always go back to the one you like best.
Step 14 – Test render
The only way you that will have any idea what your final image will look like is by testing. Next to the eyes is where we will spend the most time. Usually by the first test render you can see a lot of things that need to be changed. You may need to go back and change the model and all the texture maps. This is the point where you can finally begin to work on everything all at once – it is the most frustrating part of creating the image. Take the test renders to Photoshop and do some draw overs, colour corrections and lighting changes. Then go back to make all the corrections. Make sure that you are constantly comparing the portrait to the photo references and art reference.
Step 15 – Get new eyes
It is essential that you take a fresh look at your image – this is the easiest way to notice anything that may be wrong with the image. There are a couple of ways to get new eyes on the work. One is to make the image black and white because the colour can distract you. Also, try playing with levels in Photoshop and go to the extreme with this. Find out where the hotspots are, and see if they are where they should be. Flip the image and right away you’ll notice if the composition is off. Do something radical with the colour levels. When you come back to the original levels you might see that the colour needs slight adjustment. Lastly, get someone’s feedback and ask them to be critical. They will likely spot things that you have missed.
Step 16 – Final render
Your work will never be complete, there will just come a time when you stop working on it. When rendering the final image, make sure you have all the passes that you will need for the touchups in Photoshop: Diffuse, Specular, Reflection, Z-depth, a multimap for the body, raw lighting, Bump Normals, and the Alpha. Having all the passes you need will enable control and the option to make any changes that you want to the final image. Even if you do not make that many changes, it is better to have these passes than not. Double check that all the lights have a good amount of samples so that there is as little noise as possible while maintaining a good render time. If you plan to print out your image keep in mind that most printers require 300dpi. It is better to render out larger otherwise your image will print much smaller than you thought. We will be rendering with a height of 5,000 pixels.
Step 17 – Touchups in Photoshop
Now take all of the renders into Photoshop. Having rendered out a couple of different passes we are able to bump up the spec in parts, add Z-depth, and make any colour corrections as needed. Add a little warmth to the colour of the skin. Go back to the eyes and touch them up again – creating a realistic reflection is important. We can never get the results you are after in the render, so do this in Photoshop. It’s important that the background complements your model but does not stand out too much. The CG world will be one of your biggest critics and nothing will go unnoticed. Take time to go over the entire image to make sure that you haven’t missed anything.