This week, I set the lighting of the model in Maya to prepare it to render in UAL Render Farm
I started to play with the lighting of the model in Maya and also added a background. I wanted a simple background as the model is crowded enough and I did not want it to take off the attention from the model. Therefore, I found a stone-like texture in black mostly. Then, following the colour palette from the model, I added a back blue and purple light to separate the background from the model and to highlight the texture of the background.
Since the space background in the interior of the dome looked a bit crowded, I did some colour correction in it so the stars looked dimmer and reduced the quantity of them.
Once I got the light set up, I started to test the render settings that would take the maximum detail but at the same time that will not take too long to render.
Render settingsRender sample
Also, since I wanted to import it into Nuke to add shadows and some extra illumination, I decided to export it in different layers, like the main solar system structure separated from the dome, dome neon half ring and background. In order to separate the dome neon ring from the dome for rendering, I added a plane in between so when rendering the alpha the rest of the dome would not show.
Plane to render dome neon half ring
After everything was ready to render, I transferred the project folder to the university computer and set the UAL Render Farm (Deadline).
Submitting to Render FarmRender FarmBackground layerNeon half ring layerDome layerMain model layer
Main model render without dome and background
These layers are going to be put together in Nuke and I will add shadows and extra light reflections.
In this lesson, we saw the different techniques that can be used for 3D project, such as patch projection, coverage projection, or nested projection, and we also analysed how to add texture and lighting onto a 3D object as well as the general problems we can encounter with this.
In 3D tracking, we need to try to avoid to include the sky, as it would give us problems later on, in the same way that we avoid objects that move or reflections in roto.
When adding a ‘rotopaint’ to a card in a 3D space, we need to first freeze the frame with a ‘frame hold’ node at the best position in the sequence for visibility and tracking a specific point. Then we add the ‘rotopaint’ or the patch we need, and add another ‘frame hold’ to ‘unfreeze’ the frame. Then we premultiply it to create an alpha and use a ‘project 3D’ node to project it in our card (the ‘project 3D’ node must be connected to the projection camera and another ‘frame hold’ node). Lastly, we connect our card to the ‘scanline render’ node which will be merged with the main plate.
In order to add texture to a ‘card’ in 3D space, we will use the same method as before, but this time we will take the texture or picture that we want to add which we can ‘colour correct’ and ‘grade’ if needed, to then ‘roto’ the part we want to add from it, premultiply it, and with ‘corner pin 2D’ we will place it in the perspective we desire. Then we will ‘transform’ it to the dimensions we want and ‘merge’ it to the main plate after adding a ‘frame hold’. Lastly, we need to ‘copy’ the roto and premultiply it so we can project the alpha to our ‘card’.
If we want to roto something in the scene to change its features (colour correct, grade, etc), we can do the same as we did with the ‘rotopaint’ but in this case we adjust the roto every 10 or 20 frames. We do not need to adjust the roto every frame as it will follow our match move previously done so just a few adjustments should be sufficient.
When we have several 3D projections that we want to put together, we can use ‘Merge mat’ node, as if we use a regular ‘merge’ node, the quality of the image can decrease and look different.
After seeing these 3D projection techniques, we were asked to practice them using the following a footage of a street provided by the lecturer. For example, we could add something on the wall or floor, change the windows texture, colour correct a specific element of the scene, etc. This is the result of my practice:
Rotopaint on wallGraffiti on wallFloor cracks and failed attempt to add a window
When 3D projecting on top of a 3D object or artefact, the types of projections we can use are:
Patch projection
Coverage projection
Nested projection (projection inside another projection)
We can find some issues when doing artefact projections that can be solved we the following techniques:
Stretching problem: texture is stretched and not showing in the correct place. This issue can be fixed adding a second camera projector on top.
Doubling problem: texture is doubled. We can fix it doing two separate projections.
Resolution problem: texture look pixelated. We can use ‘sharpen’ node to solve it, however, we can also use a more efficient solution which is adding ‘reformat’ node and set the ‘type’ as ‘scale’, to then link node to ‘scanline render’ which would be the connected to a second ‘reformat’ node with the resolution of the original plate.
Lastly, we also saw how to build a 3D model taking as a reference a 2D image. Using ‘model builder’ node, we can create and adjust cards following the perspective of the 2D image, to then ‘bake’ this geometry into a 3D space. We can add ‘point light’ nodes to set illumination with different intensity, colours, and cast shadows. Another illumination node is the ‘direct light’ which is used as a filling light directed to a specific point or direction.
Once we finished reviewing this week’s theory, we were also asked to make the roto of the hole in the scene of the Garage project and to remove the markers with patch projections. I made the roto pretty quick and had no issues with it, but I struggled with two specific markers clean up: in the two markers positioned by the hole in the wall, when I added the roto, the patch made with rotopaint was showing outside the roto boundaries (right on top of this roto), so it was showing the wrong patch.
Roto errorFailed attempt to fix the roto on two markers on both sides of the holeInitial node graph set up
After asking the professor for some help, he figured out that I missed the lens distortion node on both the beginning and the end of the clean up set up (to undistorted the scene and the redistort it back).
Final nodes graph set up with correctionRotoscoping node graph of the hole on the wallProjection cards place in 3D space with point cloud as referenceProjection cards placed in 3D spaceRoto split in sections (bottom section)Roto split in sections (right section)Roto split in sections (left section)Roto split in sections (rope section)Roto split in sections (lamp section)
Another issue I noticed is that the patches added on the floor marks were showing through the roto of the wall. I asked the professor again and found out that this part needs to be merged differently as it is outside the roto. So added a ‘merge (stencil)’ just to these part of the clean-up, then ‘shuffle (alpha-alpha)’ and connected it to the roto ‘scanline render’ node. This will create an stencil of the patches taking the roto as reference and it will not show through the wall.
Clean-up cards visible through roto error‘shuffle’ node set upError corrected with ‘merge (stencil)’
Final clean-up + roto
I had a lot of troubles with this homework and spent a lot of time trying to figure out why it was not working, but I feel that this struggle was useful to familiarise a bit more and feel more confident towards the nodes system used in Nuke.
In this class, we set up the lightning and colour corrected our model to get it ready to render and compositing later on After Effects.
To create the illumination of my scene, I wanted to recreate the real lighting of the scene I took as reference. I added 3 spotlights to my scene: one is in front of the model to illuminate the face, and the other two are behind as backlights.
Spotlights positioningInitial illumination set upInitial illumination preview
Since the lip sync looked a bit odd without a tongue, I modelled one, added a standard surface material with some shiny highlights (to give a wet look) and animated it in the parts with ‘L’, ‘D’, ‘S’, and ‘T’ sounds.
Tongue
Using the ‘Hypershade’ editor, I added two colour correct nodes: one linked to the base colour and coat of the skin material, and the second linked to the specular colour of the skin material. I added a soft yellowish base colour to the skin as is shown in the reference clip but then added a blue/purple highlight with the specular colour to make it a bit more interesting. I also added a subsurface scattering to make the skin a bit more translucent, so it looks more realistic.
Colour Correct 1 – linked to base colour and coat colourColour Correct 2 – linked to specular colourSubsurface scatteringHalf resolution render image to check lighting
After I was happy with my lighting, I set my project render details to half resolution to have a relatively quick render to review the final look. Thankfully, I did not have to change anything as I like how it ended so I set my project at HD 1920×1080 resolution and render the final lip sync animation as an alpha.
Final render
The final render turned out to be good and it is ready for compositing with a background and some nice VFX. So I imported it to After Effects and searched for a proper background for the scene. I found this already blurred background with a futuristic look so I added it to the comp and colour corrected it to make it darker, more saturated and showing more blue hues (I added colour balance and brightness and contrast effects).
Then I also colour corrected the face as it seemed a little flat so using the ‘Levels’ effect increased the contrast and added a bit of more shadows and saturation.
Face without colour correctionFace with colour correction
Finally, I also scaled up the face and I scaled down the background towards to end to make the scene more dynamic and simulate the camera movement.
Scale and position of face and background
Final Comp
I think that the final scene turned out good, as I like the contrast between the blue/purple background, and the yellowish skin colour of the face with the subtle blue/purple highlights on the head and forehead. Overall, I think the scene looks realistic and the lip sync and head movements looks pretty natural.
References
topntp26, freepik. Blurred abstract background interior view looking out toward to empty office lobby and entrance doors and glass curtain wall with frame. [Online] Available at: Available at: https://www.freepik.com/free-photo/blurred-abstract-background-interior-view-looking-out-toward-empty-office-lobby-entrance-doors-glass-curtain-wall-with-frame_1254627.htm#query=building%20indoor&position=36&from_view=search&track=sph [Accessed 12 December 2022]